IEEE Rolls Out Large Language Models Training Course - IEEE Spectrum Opens in a new window Opens an external website Opens an external website in a new window This website utilizes technologies such as cookies to enable essential site functionality, as well as for analytics, personalization, and targeted advertising. Privacy Policy (https://www.ieee.org/security-privacy.html) Accept Deny Non-Essential Manage Preferences Storage Preferences IEEE.org (https://www.ieee.org/)IEEE Xplore (https://ieeexplore.ieee.org/Xplore/home.jsp)IEEE Standards (https://standards.ieee.org/)IEEE Job Site (https://jobs.ieee.org/)More Sites (https://www.ieee.org/sitemap.html) Sign In (https://www.ieee.org/profile/public/createwebaccount/showCreateAccount.html?ShowMGAMarkeatbilityOptIn=true&sourceCode=spectrum&signinurl=https%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fsaml%2Fmain%2Flogin%3Fnext_url%3Dhttps%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fintegrations%2Fieee%2Fchanges%3Fredirect%3Dhttps%3A%2F%2Fspectrum.ieee.org%2Flarge-language-models-ieee-course&url=https%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fsaml%2Fmain%2Flogin%3Fnext_url%3Dhttps%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fintegrations%2Fieee%2Fchanges%3Fredirect%3Dhttps%3A%2F%2Fspectrum.ieee.org%2Flarge-language-models-ieee-course&autoSignin=Y&car=IEEE-Spectrum)Join IEEE (https://spectrum.ieee.org/join) https://spectrum.ieee.org/ IEEE Rolls Out Large Language Models Virtual Training Course Share FOR THE TECHNOLOGY INSIDER Search: Explore by topic Aerospace (https://spectrum.ieee.org/topic/aerospace/)AI (https://spectrum.ieee.org/topic/artificial-intelligence/)Biomedical (https://spectrum.ieee.org/topic/biomedical/)Climate Tech (https://spectrum.ieee.org/topic/climate-tech/)Computing (https://spectrum.ieee.org/topic/computing/)Consumer Electronics (https://spectrum.ieee.org/topic/consumer-electronics/)Energy (https://spectrum.ieee.org/topic/energy/)History of Technology (https://spectrum.ieee.org/topic/tech-history/)Robotics (https://spectrum.ieee.org/topic/robotics/)Semiconductors (https://spectrum.ieee.org/topic/semiconductors/)Telecommunications (https://spectrum.ieee.org/topic/telecommunications/)Transportation (https://spectrum.ieee.org/topic/transportation/) IEEE Spectrum (https://spectrum.ieee.org/) FOR THE TECHNOLOGY INSIDER Topics Aerospace (https://spectrum.ieee.org/topic/aerospace/)AI (https://spectrum.ieee.org/topic/artificial-intelligence/)Biomedical (https://spectrum.ieee.org/topic/biomedical/)Climate Tech (https://spectrum.ieee.org/topic/climate-tech/)Computing (https://spectrum.ieee.org/topic/computing/)Consumer Electronics (https://spectrum.ieee.org/topic/consumer-electronics/)Energy (https://spectrum.ieee.org/topic/energy/)History of Technology (https://spectrum.ieee.org/topic/tech-history/)Robotics (https://spectrum.ieee.org/topic/robotics/)Semiconductors (https://spectrum.ieee.org/topic/semiconductors/)Telecommunications (https://spectrum.ieee.org/topic/telecommunications/)Transportation (https://spectrum.ieee.org/topic/transportation/) Sections Features (https://spectrum.ieee.org/type/feature/)News (https://spectrum.ieee.org/type/news/)Opinion (https://spectrum.ieee.org/type/opinion/)Careers (https://spectrum.ieee.org/topic/careers/)DIY (https://spectrum.ieee.org/topic/diy/)Engineering Resources (https://spectrum.ieee.org/engineering-resources/) More Newsletters (https://spectrum.ieee.org/newsletters/)Special Reports (https://spectrum.ieee.org/special-reports/)Collections (https://spectrum.ieee.org/collections/)Explainers (https://spectrum.ieee.org/type/explainer/)Top Programming Languages (https://spectrum.ieee.org/top-programming-languages)Robots Guide ↗ (https://robotsguide.com/)IEEE Job Site ↗ (https://jobs.ieee.org/) For IEEE Members Current Issue (https://spectrum.ieee.org/magazine/current-issue)Magazine Archive (https://spectrum.ieee.org/magazine/)The Institute (https://spectrum.ieee.org/the-institute/)The Institute Archive (https://spectrum.ieee.org/the-institute/ti-archive/) For IEEE Members Current Issue (https://spectrum.ieee.org/magazine/current-issue)Magazine Archive (https://spectrum.ieee.org/magazine/)The Institute (https://spectrum.ieee.org/the-institute/)The Institute Archive (https://spectrum.ieee.org/the-institute/ti-archive/) IEEE Spectrum About Us (https://spectrum.ieee.org/about)Contact Us (https://spectrum.ieee.org/contact)Reprints & Permissions ↗ (https://www.parsintl.com/publications/ieee-media/)Advertising ↗ (https://advertise.ieee.org/) Follow IEEE Spectrum https://twitter.com/ieeespectrumhttps://www.facebook.com/IEEE.Spectrumhttps://www.instagram.com/ieeespectrum/https://www.threads.net/@ieeespectrumhttps://www.linkedin.com/company/ieee-spectrum/https://www.youtube.com/c/ieeespectrumhttps://www.tiktok.com/@ieeespectrumhttps://spectrum.ieee.org/customfeeds/feed/all-topics/rss Support IEEE Spectrum IEEE Spectrum is the flagship publication of the IEEE — the world’s largest professional organization devoted to engineering and applied sciences. Our articles, videos, and infographics inform our readers about developments in technology, engineering, and science. Subscribe (https://ieee.dragonforms.com/spectrum_subscribe?utm_source=ieee-spectrum&utm_medium=nav&utm_content=hugemenu) About IEEE (https://www.ieee.org/about/)Contact & Support (https://www.ieee.org/about/contact.html)Accessibility (https://www.ieee.org/accessibility-statement.html)Nondiscrimination Policy (https://www.ieee.org/about/corporate/governance/p9-26.html)Terms (https://www.ieee.org/about/help/site-terms-conditions.html)IEEE Privacy Policy (https://www.ieee.org/security-privacy.html)Cookie Preferences (https://spectrum.ieee.org/large-language-models-ieee-course#)Ad Privacy Options (https://spectrum.ieee.org/st/ppid-info) © Copyright 2026 IEEE — All rights reserved. A public charity, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. Enjoy more free content and benefits by creating an account Saving articles to read later requires an IEEE Spectrum account The Institute content is only available for members Downloading full PDF issues is exclusive for IEEE Members Downloading this e-book is exclusive for IEEE Members Access to Spectrum 's Digital Edition is exclusive for IEEE Members Following topics is a feature exclusive for IEEE Members Adding your response to an article requires an IEEE Spectrum account Create an account to access more content and features on IEEE Spectrum , including the ability to save articles to read later, download Spectrum Collections, and participate in conversations with readers and editors. For more exclusive content and features, consider Joining IEEE (https://spectrum.ieee.org/join) . Join the world’s largest professional organization devoted to engineering and applied sciences and get access to all of Spectrum’s articles, archives, PDF downloads, and other benefits. Learn more about IEEE → (https://spectrum.ieee.org/join) Join the world’s largest professional organization devoted to engineering and applied sciences and get access to this e-book plus all of IEEE Spectrum’s articles, archives, PDF downloads, and other benefits. Learn more about IEEE → (https://spectrum.ieee.org/join) CREATE AN ACCOUNT (https://www.ieee.org/profile/public/createwebaccount/showCreateAccount.html?ShowMGAMarkeatbilityOptIn=true&sourceCode=spectrum&signinurl=https%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fsaml%2Fmain%2Flogin%3Fnext_url%3Dhttps%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fintegrations%2Fieee%2Fchanges%0A&url=https://spectrum.ieee.org/&autoSignin=Y&car=IEEE-Spectrum)SIGN IN (https://spectrum.ieee.org/core/saml/main/login?next_url=https://spectrum.ieee.org/core/integrations/ieee/changes) JOIN IEEE (https://www.ieee.org/membership-application/public/join.html?promo=JOINLITE&style=SPECTRUM&joinlite=TRUE)SIGN IN (https://spectrum.ieee.org/core/saml/main/login?next_url=https://spectrum.ieee.org/core/integrations/ieee/changes) Close (javascript:) Access Thousands of Articles — Completely Free Create an account and get exclusive content and features: Save articles, download collections, and post comments — all free! For full access and benefits, subscribe (https://ieee.dragonforms.com/spectrum_subscribe) to Spectrum. CREATE AN ACCOUNT (https://www.ieee.org/profile/public/createwebaccount/showCreateAccount.html?ShowMGAMarkeatbilityOptIn=true&sourceCode=spectrum3c&signinurl=https%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fsaml%2Fmain%2Flogin%3Fnext_url%3Dhttps%3A%2F%2Fspectrum.ieee.org%2Fcore%2Fintegrations%2Fieee%2Fchanges%0A&url=https://spectrum.ieee.org/&autoSignin=Y&car=IEEE-Spectrum)SIGN IN (https://spectrum.ieee.org/core/saml/main/login?next_url=https://spectrum.ieee.org/core/integrations/ieee/changes) The Institute (https://spectrum.ieee.org/topic/the-institute/)AI (https://spectrum.ieee.org/topic/artificial-intelligence/)Careers (https://spectrum.ieee.org/topic/careers/)News (https://spectrum.ieee.org/type/news/) IEEE Rolls Out Large Language Models Virtual Training Course Learn how to design, secure, and deploy LLMs Angelique Parashis (https://spectrum.ieee.org/u/angelique-parashis) 3 minutes ago 3 min read javascript: iStock mailto:?subject=IEEE%20Rolls%20Out%20Large%20Language%20Models%20Virtual%20Training%20Course&body=https://spectrum.ieee.org/large-language-models-ieee-course AI (https://spectrum.ieee.org/tag/ai)type:ti (https://spectrum.ieee.org/tag/type-ti)education (https://spectrum.ieee.org/tag/education)ieee educational activities (https://spectrum.ieee.org/tag/ieee-educational-activities)large language models (https://spectrum.ieee.org/tag/large-language-models)ieee products and services (https://spectrum.ieee.org/tag/ieee-products-and-services)type:ti (https://spectrum.ieee.org/tag/type-ti) Large language models (https://spectrum.ieee.org/recursive-self-improvement) have moved out of the research lab and into engineers’ daily workflow. LLMs serve as reasoning engines that can orchestrate complex tasks including identifying vulnerabilities in source code (https://spectrum.ieee.org/tag/source-code) and transforming fragmented project discussions into rigorous technical specifications. While the general public uses AI tools to write email and plan vacations, technical professionals use LLMs as core architectural elements that are fundamentally changing how digital infrastructures are built and maintained. As the AI models (https://spectrum.ieee.org/tag/ai-models) move into mainstream engineering practice, the demand for technical expertise is rising. The LLM technology market is expected to grow by about 33 percent every year through 2030 (https://www.marketsandmarkets.com/Market-Reports/large-language-model-llm-market-102137956.html), according to MarketsandMarkets (https://www.marketsandmarkets.com/AboutUs-8.html). The rapid expansion suggests that proficiency in implementing and securing the models is transitioning from a niche into a core requirement for technologists. More than just a better search engine (https://spectrum.ieee.org/tag/search-engine) To use LLMs effectively, technical professionals must move beyond treating them as conversational robots. At a fundamental level, the AI systems are built on the transformer architecture (https://ieeexplore.ieee.org/document/10245906), a framework that replaced the older method of processing data in a fixed, sequential order. Unlike earlier models that analyzed information one step at a time, transformers (https://spectrum.ieee.org/tag/transformers) use self-attention mechanisms to ingest vast datasets simultaneously. For technical professionals, LLMs are core architectural elements that are fundamentally changing how digital infrastructures are built and maintained. Relying on such LLMs without understanding their internal logic creates a significant reliability risk. To build tools that work consistently, developers must understand the core principles that govern how the models process information and generate results. By mastering how a model processes information and how its internal settings influence the result, developers can move away from a trial-and-error approach toward a more precise one to ensure the AI tool handles complex data reliably. Four ways LLMs are changing jobs Here are areas that integrate large language models (https://spectrum.ieee.org/tag/large-language-models). • *Moving past basic prompts.**Developers are using application program interfaces (APIs) to connect LLMs directly to their databases (https://spectrum.ieee.org/tag/databases) and software tools. Employing the APIs allows AI to perform work such as executing code or searching through internal repositories. • *Fixing the “hallucination” problem.**LLMs are at risk of hallucinations (https://spectrum.ieee.org/ai-agent-benchmarks), which are generated facts or code that looks correct but actually is wrong or broken. To fix the problem, retrieval-augmented generation (RAG) forces AI to look up information in a trusted source such as a company’s database. • *Prioritizing data security (https://spectrum.ieee.org/tag/data-security).**When using AI with proprietary code, security (https://spectrum.ieee.org/two-new-ai-ethics-certifications) is a major concern. Engineers must learn how to set up “private” instances of the models to ensure that sensitive company data stays within a secure cloud environment and is not used to train public versions. • *The future of collaboration.**By automating repetitive coding tasks and summarizing thousands of pages of documentation, LLMs let engineers spend more time on high-level designs and solving important issues. Online course program helps with mastering the tech The gap between people who use AI and those who understand how to build with it is growing wider. To help technical professionals stay ahead, IEEE offers a five-course online program, Large Language Models Demystified (https://iln.ieee.org/public/contentdetails.aspx?id=B570F53B5DA44B258042A12AE5BD6846), available through the IEEE Learning Network (https://iln.ieee.org/). The program, developed by IEEE Educational Activities (https://ea.ieee.org/) in partnership with the IEEE Computer Society (https://computer.org/), is built for people who want to understand the “how” and the “why” behind the technology. Rather than just teaching basic prompting, the curriculum dives into the engineering behind generative AI (https://spectrum.ieee.org/tag/generative-ai), including: • Evolution, impact, and hands-on exercises:the shift from statistical methods to modern transformers, including hands-on model optimization. • Understanding transformer architectures: the mathematical core of self-attention and positional encoding, implemented in NumPy (https://numpy.org/) and Python (https://www.python.org/). • Architectural analysis and implementation: advanced LLM design with practical model-building exercises. • Training and modeling with PyTorch: end-to-end pipelines in PyTorch (https://pytorch.org/), leveraging parameter-efficient techniques such as low-rank adaptation (https://arxiv.org/abs/2106.09685) and quantization. • Optimization, alignment, and deployment: performance scaling, reinforcement learning from human feedback (RLHF) (https://aws.amazon.com/what-is/reinforcement-learning-from-human-feedback/), group-relative policy optimization (https://cameronrwolfe.substack.com/p/grpo), RAG, and agentic AI. Upon completion of the program, participants earn professional development (https://spectrum.ieee.org/tag/professional-development) credits and a digital badge from IEEE to verify their expertise. Enroll in the course program (https://iln.ieee.org/public/contentdetails.aspx?id=B570F53B5DA44B258042A12AE5BD6846) on the IEEE Learning Network (https://spectrum.ieee.org/tag/ieee-learning-network). Organizations looking to prepare their teams to work on LLMs can connect with an IEEE content specialist (https://forms1.ieee.org/Large-Language-Models-Demystified.html) to discuss group enrollment and tailored training paths. From Your Site Articles • How Large Language Models Are Changing My Job › (https://spectrum.ieee.org/large-language-models-2668430044) • What to Look Out for When Acquiring AI Systems › (https://spectrum.ieee.org/ieee-ai-3119-standards) • Two New AI Ethics Certifications Available from IEEE › (https://spectrum.ieee.org/two-new-ai-ethics-certifications) Related Articles Around the Web • LLM 101: Foundations and Practical Application - Young Professionals › (https://yp.ieee.org/event/llm-101-foundations-and-practical-application/) AI (https://spectrum.ieee.org/tag/ai)type:ti (https://spectrum.ieee.org/tag/type-ti)education (https://spectrum.ieee.org/tag/education)ieee educational activities (https://spectrum.ieee.org/tag/ieee-educational-activities)large language models (https://spectrum.ieee.org/tag/large-language-models)ieee products and services (https://spectrum.ieee.org/tag/ieee-products-and-services)type:ti (https://spectrum.ieee.org/tag/type-ti) Angelique Parashis (https://spectrum.ieee.org/u/angelique-parashis) is senior manager, education marketing, for IEEE Educational Activities. See full bio→ (https://spectrum.ieee.org/u/angelique-parashis) The Conversation (0) https://spectrum.ieee.org/video-friday-agentic-ai-robot Robotics (https://spectrum.ieee.org/topic/robotics/)News (https://spectrum.ieee.org/type/news/) Video Friday: Do Robots Even Need Legs? (https://spectrum.ieee.org/video-friday-agentic-ai-robot) 3 hours ago 3 min read javascript: https://spectrum.ieee.org/agilink-contact-intelligence-robot-manipulation Robotics (https://spectrum.ieee.org/topic/robotics/)Humanoid Robots (https://spectrum.ieee.org/topic/robotics/humanoid-robots/)Sponsored Article (https://spectrum.ieee.org/type/sponsored/) Beyond Dexterity: Why Contact May Define the Next Era of Robotics (https://spectrum.ieee.org/agilink-contact-intelligence-robot-manipulation) 09 Jun 2026 6 min read javascript: 3 https://spectrum.ieee.org/modos-e-paper-monitor Consumer Electronics (https://spectrum.ieee.org/topic/consumer-electronics/)News (https://spectrum.ieee.org/type/news/) Two E-Paper Engineers Keep Turning to the Wisdom of the Crowd (https://spectrum.ieee.org/modos-e-paper-monitor) 18 Jun 2026 4 min read javascript: 1 More from The Institute The Institute (https://spectrum.ieee.org/topic/the-institute/)Careers (https://spectrum.ieee.org/topic/careers/)Article (https://spectrum.ieee.org/type/article/) IEEE’s 2026 Education Week Events Emphasized Lifelong Learning (https://spectrum.ieee.org/ieee-education-week) The Institute (https://spectrum.ieee.org/topic/the-institute/)Article (https://spectrum.ieee.org/type/article/)Careers (https://spectrum.ieee.org/topic/careers/)Climate Tech (https://spectrum.ieee.org/topic/climate-tech/)Climate Change (https://spectrum.ieee.org/collections/climate-change/) Engineering Is Critical to Boosting Food Security (https://spectrum.ieee.org/engineering-critical-food-security) The Institute (https://spectrum.ieee.org/topic/the-institute/)IEEE Member News (https://spectrum.ieee.org/the-institute/collections/ieee-member-news/)Robotics (https://spectrum.ieee.org/topic/robotics/)Careers (https://spectrum.ieee.org/topic/careers/)Profile (https://spectrum.ieee.org/type/profile/) Award-Winning Researcher Trains Robots to Make Educated Guesses (https://spectrum.ieee.org/researcher-trains-robots-to-guess) The Institute (https://spectrum.ieee.org/topic/the-institute/)Careers (https://spectrum.ieee.org/topic/careers/)Article (https://spectrum.ieee.org/type/article/) EPICS in IEEE’s Awards Honor Outstanding Students and Faculty (https://spectrum.ieee.org/epics-in-ieee) The Institute (https://spectrum.ieee.org/topic/the-institute/)Careers (https://spectrum.ieee.org/topic/careers/)Article (https://spectrum.ieee.org/type/article/) IEEE Celebrates Technology’s Brightest Minds at Annual Event (https://spectrum.ieee.org/ieee-celebrates-honors-ceremony-2026) Energy (https://spectrum.ieee.org/topic/energy/)Magazine (https://spectrum.ieee.org/magazine/)Feature (https://spectrum.ieee.org/type/feature/)Climate Tech (https://spectrum.ieee.org/topic/climate-tech/) This 1976 University Experiment Spun Up the U.S. Wind Industry (https://spectrum.ieee.org/william-heronemus-wind-energy) A nuclear submarine veteran turned a truck axle into a turbine Peter Fairley (https://spectrum.ieee.org/u/peter-fairley-a) 15 Jun 2026 16 min read javascript: 2 mailto:?subject=This%201976%20University%20Experiment%20Spun%20Up%20the%20U.S.%20Wind%20Industry&body=https://spectrum.ieee.org/william-heronemus-wind-energy https://spectrum.ieee.org/william-heronemus-wind-energy With a commanding yet kind presence, Captain William Heronemus [left, in hat] inspired his UMass Amherst team to get the job done. Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries • *A half century ago,**a scrappy crew at the University of Massachusetts Amherst erected a wind turbine on Orchard Hill, the highest point on campus. It was a frugal production, cobbled together from the rear axle of a Ford truck, a donated generator and microcontroller, a steam pipe, and various handcrafted steel and fiberglass parts, including its 4.5-meter blades. The team of UMass (https://www.umass.edu/) engineering grad students, faculty advisors, and one precocious undergrad built it to prove that wind energy could keep rural homes toasty in New England’s frigid winters, as a way of trimming U.S. oil dependence—a national imperative in the aftermath of the 1973–1974 energy crisis. To illustrate the point, they also assembled a modular home there on Orchard Hill, and outfitted it with heaters that would be powered by the turbine. In 1975 and 1976, a crew from the University of Massachusetts Amherst designed and constructed the 25-kilowatt wind turbine that kick-started the U.S. wind industry. Sandy Butterfield It worked—too well. “We had to open up the doors in the dead of winter. It was just too damn hot,” recalls Michael Edds (https://www.linkedin.com/in/medds/), who designed the turbine’s electrical system and served as the project’s first resident engineer. Fittingly, they dubbed the turbine the “Wind Furnace.” The turbine maxed out at 25 kilowatts—puny compared to modern machines that generate up to 26 __mega__ watts, but more than most energy experts expected from wind technology in November 1976. Back then, wind power still conjured up images of quaint Dutch mills and creaky prairie water pumpers. Crafty engineers would soon show that wind power could be so much more. And it all began with the brilliant, commanding, and often polarizing UMass professor leading the Wind Furnace project: William Heronemus. A retired U.S. Navy captain, Heronemus had joined the UMass faculty in 1967. He’d earned Bronze Stars for valor in World War II, designed and built nuclear submarines, and liaised with the British Royal Navy on the Polaris missile. UMass had recruited Heronemus to do ocean engineering, but the energy crisis and his growing misgivings about nuclear power shifted his attention to renewable energy. Heronemus, photographed circa 1973, publicly advocated for the buildout of wind turbines, both onshore and off, at immense scale. Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries By 1972, Heronemus was advancing detailed designs to deploy wind turbines at immense scale. That year, at the Marine Technology Society’s annual gathering in Washington, D.C., he presented schemes for building thousands of them across the Great Plains as well as a vast grid of massive floating turbines transecting New England’s continental shelf. Wind power, he contended, could generate nearly a fifth of U.S. electricity needs by the year 2000. Never mind that the technology for such an enormous buildout had yet to be commercialized. Espousing grand schemes made Heronemus a quixotic figure. He also vigorously attacked the commercialization of nuclear power, creating enemies within electric utilities and U.S. government agencies that saw nuclear technology as the future. They didn’t appreciate his claims that a cleaner energy future via wind was ready to be tapped, and that the push for nuclear power and its radiological risks was unnecessary. As author and energy analyst Peter Asmus (https://www.peterasmus.com/) put it in his 2000 book, __Reaping the Wind__: “William Heronemus (https://www.umass.edu/windenergy/about/history/heronemus/index.html) was a dangerous man suggesting an audacious departure from the status quo.” The UMass Amherst wind turbine generated most of the energy to heat a modular home through the cold, windy winters on Orchard Hill. Solar thermal panels provided some heat during windless periods. Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries What happened on Orchard Hill in 1976 marked Heronemus’s turn from provocateur to changemaker. The success of the experimental turbine set off waves of technological and industrial developments that forever changed the energy landscape. Within a few years, the students he trained and the entrepreneurs he inspired were building the world’s first modern wind farms and leading the Great California Wind Rush—the market that turned wind craft into an industry that’s still growing fast half a century later. Globally, annual wind generation more than tripled between 2015 and 2025, according to data from Ember Energy (https://ember-energy.org/), a think tank based in London. It will best nuclear’s global output by the end of this year, Ember predicts. And it all started with Heronemus, says Robert Thresher (https://research-hub.nlr.gov/en/persons/robert-thresher/), longtime former director of wind research at the National Renewable Energy Laboratory (NREL) in Golden, Colo. (a U.S. Department of Energy lab rebranded late last year as the National Laboratory of the Rockies (https://research-hub.nlr.gov/en/persons/robert-thresher/)). “In my mind he was the father of the people that went out and really made the industry what it is today,” he says. William Heronemus and the History of Wind Power I got to know Captain Heronemus posthumously, interviewing his contemporaries and sifting through boxes delivered to the UMass Amherst archival research center’s 25th-floor reading room. During three visits there since 2023, I have discovered clues to his life, thinking, and research process amid the writings where he pitched his big ideas to the world. His papers include proposals to governments, utilities, and deep-pocketed philanthropists and investors, including Jane Fonda and Goldman-Sachs. Papers reveal the internationalism and commitment to service that took Heronemus on renewable-energy consulting trips to Pakistan, Cuba, Côte d’Ivoire, and beyond. Records show meetings with corporate powerhouses like Boeing and Grumman Aerospace and calls on politicians, including the senator and presidential hopeful Ted Kennedy. Postcards from former students exude gratitude. Heronemus sits with a mock-up of a multirotor turbine in his cramped office in Marston Hall, UMass Amherst’s main engineering building. Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries I learned that Heronemus turned his attention from ocean engineering to energy a few years after arriving at UMass, when he saw the growing string of nuclear power plants going up along the Connecticut River, which flows past Amherst en route to Long Island Sound. The U.S. government had picked nuclear power as an antidote to the 1970s oil crises, and Northeast utilities had jumped in big. But Heronemus and other UMass engineers worried that the riverside reactors’ waste heat would threaten the river’s ecosystem and bounty. The advent of cooling towers to blow off heat into the air addressed the thermal pollution concern but created another: water depletion. (Nuclear plants consume about 60 million gallons of water per day, per reactor, on average.) And Heronemus perceived other nuclear power liabilities, stemming from his experience with nuclear propulsion on Navy ships. As a design engineer and head of construction and repair for a shipyard, he valued the military’s zero-accident standard for reactors but also knew the high cost of adhering to it. He argued that building expanded versions of the Navy’s pressurized water reactors to power cities and factories couldn’t be both safe __and__ economical. In 1971, Heronemus designed an offshore turbine with three rotors, but the first big multirotor prototype wouldn’t be built for another four decades. Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries He predicted—accurately, as it turned out—that costs would rise sharply as the nuclear industry addressed safety and environmental concerns. “Each plant costs more than its predecessor. The shipyards involved with nuclear reactors came to that conclusion years ago,” he wrote in a 1973 research proposal. He also argued that the risks inherent in nuclear reactors and their radioactive waste were unnecessary given Earth’s abundant solar and wind energy resources. He broadcast those views wherever and whenever he could: before congressional committees, at U.S. Atomic Energy Commission hearings, at academic conferences, in media interviews, and even at Rotary Club luncheons. At a 1973 licensing hearing for the proposed 820-MW Shoreham Nuclear Power Plant (https://en.wikipedia.org/wiki/Shoreham_Nuclear_Power_Plant) on Long Island, N.Y., for example, Heronemus called affordable nuclear energy a “myth.” He detailed, in its stead, a floating wind power system that could be moored off Long Island and sized to deliver more than four times as much electricity as the Shoreham plant. Each of the 640 floating platforms would carry six rotors and crank out up to 12 MW, some of which would power electrolyzers to generate hydrogen. The hydrogen would be fed to power plants or fuel cells to produce electricity when the wind wasn’t blowing. This seemingly futuristic idea drew on his Navy experience with water-splitting electrolyzers, which supplied the oxygen that enabled subs to remain submerged for months at a time, and NASA’s use of hydrogen fuel cells (https://spectrum.ieee.org/tag/hydrogen-fuel-cells) to power the Apollo missions. More than five decades later, his vision for offshore wind power (https://spectrum.ieee.org/tag/offshore-wind-power) is big business. Floating platforms are now widely accepted as the future of offshore wind, as necessity pushes the industry to build in deeper waters (https://spectrum.ieee.org/floating-offshore-wind-turbine). Testing began on the first floating electrolysis platforms (https://spectrum.ieee.org/green-hydrogen-offshore-wind) in 2023, and multirotor turbine prototypes are in development in China, Norway and Scotland. The UMass Amherst Wind Turbine Legacy Photos in the UMass archives invariably capture Heronemus in jacket and tie, usually standing bolt straight. That commanding affect, plus his World War II veteran pedigree, Cold War engineering credentials, and his informed, pugnacious attacks made him a hard target for his adversaries in the nuclear establishment. He certainly wasn’t your typical antinuclear activist. Wielding his Cold War engineering credentials and often dressed in a suit and tie, Heronemus fought hard against nuclear energy, arguing that wind was a far safer and cost-competitive resource.Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries But brutal candor in public settings probably won him as many enemies as friends. Consider his presentation at the IEEE Power and Energy Society (https://ieee-pes.org/)’s 1974 winter meeting, where Heronemus suggested scrapping the utilities’ then nuclear-focused research arm, the Electric Power Research Institute (https://www.epri.com/). That stance no doubt created discomfort for the engineers in attendance who were involved in EPRI projects, or who aspired to be. It’s hard to say whether Heronemus’s campaign slowed nuclear development. The industry was already struggling with cost overruns when, in 1979, a reactor at Three Mile Island (https://spectrum.ieee.org/three-mile-island) in Pennsylvania partially melted down and slammed the brakes on further expansion. What is certain is that Heronemus spurred investment in wind power. When he started talking up wind in the early ’70s, even fellow travelers in the fledgling renewable energy movement were writing it off. As future White House science advisor John Holdren (https://www.hks.harvard.edu/faculty/john-holdren) opined in a 1971 Sierra Club (https://www.sierraclub.org/) book: “There are few places in the world where the wind is strong enough and steady enough to make harnessing it for the large-scale production of power at all interesting.” Heronemus dreamed up networks of wind turbines over and along highways after driving down the Garden State Parkway to a conference in Cape May, New Jersey. Ellen Heronemus Heronemus countered the naysayers by quickly forging expert consensus around wind power’s immense potential, playing a key role as the sole wind expert on a 1972 federal panel on renewable energy (https://ntrs.nasa.gov/api/citations/19730018091/downloads/19730018091.pdf). That joint National Science Foundation–NASA panel concluded that, in fact, wind could meet up to 19 percent of projected U.S. power demand by the year 2000. Congress listened, sort of. After most Persian Gulf states restricted oil shipments to the United States in 1973, congressional appropriators dedicated US $1.8 million to wind-power research and development for 1974—up from zero—and by 1976 it had bumped that to $22 million. (For comparison, Congress gave nuclear power $714 million in 1976.) Heronemus’s vision for a massive highway wind-power scheme was inspired in part by the wind-power advocate Percy Thomas, who in the 1940s and 1950s “talked a lot about how fresh New Jersey winds are,” he told the New York Times in 1974. “I got to thinking about what Thomas had said and how wind energy could be captured there.” Ellen Heronemus The bulk of the funding for wind power flowed to big aerospace firms and to NASA, financing an ultimately fruitless attempt to leap straight to megawatt-scale wind turbines. UMass struggled to grab a slice of the leftovers to pursue Heronemus’s offshore wind system. Professors and students who worked with Heronemus told me they felt they’d been blackballed as payback for his activism and antagonism. UMass finally caught a funding break when Heronemus dialed back his ambitions and proposed the 25-kW unit for Orchard Hill. A $130,000 federal grant landed in early 1975, and $150,000 more the following year. It was a “trivial” sum, according to team member Sandy (https://www.linkedin.com/in/sandy-butterfield-24b38513/)Butterfield (https://www.linkedin.com/in/sandy-butterfield-24b38513/), who would later become chief engineer for wind-turbine testing at NREL. “They gave us just enough to fail,” says Butterfield. A crane erects the “Wind Furnace” in November 1976. Sandy Butterfield But the project triumphed, resulting in Wind Furnace 1, or WF-1 (pronounced “woof one”). The young engineers behind it credit their success to the confidence, sense of mission, and structure that Heronemus gave them. The self-described “hippies” called Heronemus “the Captain” out of both affection and respect. As team member Edds puts it: “What showed in his demeanor and his actions was discipline, and it sort of rubbed off on us. We didn’t always dress like the Captain, but we knew we had to be disciplined, to be prepared, and just do the job.” From Helicopter Rotor to Wind Turbine Team WF-1 got a quick start, thanks to earlier, privately financed work by a couple of doctoral students, including Forrest “Woody” Stoddard (https://scua.library.umass.edu/stoddard-forrest-s-1944/). Stoddard had been designing helicopter rotors for the U.S. Air Force when Heronemus invited him to come work on wind power in 1972. Stoddard set about adapting helicopter-rotor theory to the closely related wind rotors, and his aerodynamics modeling proved essential to the engineering of the entire machine. Woody Stoddard [far right, in hat] designed the fiberglass blades with Ted Van Dusen. The team assembled the blades in a campus shop, and when it was time to squeegee epoxy from the blades, it was all hands on deck. Robert S. Cox Special Collections and University Archives Research Center/UMass Amherst Libraries As WF-1’s de facto chief designer, Stoddard likely supported the team’s early choice to mimic a helicopter’s ability to “pitch” its blades. To fly forward, a helicopter continuously adjusts the lift created by each blade, turning the airfoil on its long axis to reduce lift as it swings past the front of the aircraft. Doing so tilts the nose down and moves the vehicle forward. In WF-1’s case, blades pitched to regulate torque, helping get the rotor spinning in low winds and then easing off to protect the machine in dangerously high winds. Repurposing a truck axle to mechanically couple WF-1’s rotor and generator was one of several design elements borrowed from engineers at McGill University (https://www.mcgill.ca/) in Montreal. Production of WF-1’s fiberglass blades got started at UMass in 1974 under the direction of doctoral student Ted Van Dusen (https://composite-eng.com/). A competitive rower, he had a side hustle making ultralight composite boats—a trade that had stalled his doctoral work at MIT but was an accelerant for WF-1. The federal funds in 1975 allowed Heronemus to really spin up the project and recruit a squad of students to engineer the balance of WF-1’s components. They made good use of the UMass engineering machine shop and received guidance from faculty, including mechanical engineering professors Duane Cromack (https://prabook.com/web/duane_ellis.cromack/230343) and Jon McGowan (https://scholar.google.com/citations?user=NmB8VIwAAAAJ&hl=en&oi=sra). But it was the dozen or so students who really cranked out the parts. Most were master’s students, like Butterfield, who designed the blade-pitching mechanics. Edds, the team’s only electrical engineer, had come to UMass to learn ocean engineering, only to be diverted into handling WF-1’s generator. Louis Manfredi (https://www.linkedin.com/in/louismanfredi), another ocean engineering student, teamed up with master’s student Jim Sexton (https://scholarworks.umass.edu/entities/publication/0fe58480-7291-449b-ad9e-9b04625a2132) on the nacelle housing the generator and drivetrain. Fred Antoon (https://scholarworks.umass.edu/entities/publication/40f08f39-f951-46ba-9d92-89865a0fe8bb) adapted the truck axle. Brian Kuhn (https://www.linkedin.com/in/brian-kuhn-18616228/) did drawings. WF-1 contained a mechanism that pitched its blades to regulate torque in response to wind speed, a feature that became an industry standard.Sandy Butterfield An 18-year-old freshman, Dan Handman (https://patents.justia.com/inventor/daniel-f-handman), came aboard and soon made himself indispensable. When he approached Heronemus to introduce himself, Heronemus handed him three months’ worth of anemometer readings punched into recording paper, and told him to turn it into 15-minute averages. Figuring there had to be a more efficient method for analyzing wind speeds, Handman asked around and found a wind-averaging machine from an earlier student project. A month or so later, he’d installed it in a cabinet near Heronemus’s office and wired it to an anemometer on Orchard Hill. Handman’s primary role on WF-1 was setting up its computerized control system, which tracked wind speed and sent commands to Butterfield’s pitch mechanism. The controls also tracked the generator’s speed and adjusted the current to its rotor windings, in accordance with calculations by Edds. Tweaking the current ensured that power demand from the electric heaters installed in the home below didn’t stop the rotor in weak winds. Sandy Butterfield, part of the 1970s “UMass Mafia” team that built WF-1, became a wind-power entrepreneur and a top engineer at the National Renewable Energy Laboratory in Golden, Colo. Sandy Butterfield The finished WF-1 really cranked up the heat, some of which was stored by heating water in tanks in the modular house’s basement, to be circulated through baseboards in windless periods. It turned out WF-1 was unusually efficient at capturing wind energy because its rotor could change speed with the wind, keeping the blades close to an aerodynamic optimum. This varying rotor speed meant that the frequency of the electric power WF-1 produced also varied. Turbines linked to power lines must strive for the opposite—a steady output that synchronizes with the grid’s frequency—primarily 50 or 60 hertz. But it suited the home’s low-tech heating scheme just fine. (Electronic converters let today’s turbines have it all by ingesting a variable wave and outputting a new wave that’s synced to the grid.) The Great California Wind Rush In 1977, with WF-1’s success in hand, Heronemus projected that 3 million homes like the one on Orchard Hill could soon slash U.S. heating oil demand by 90 million barrels a year. That never happened, but an industry was born, starting with a Burlington, Mass. startup called US Windpower—the first “credible” U.S. turbine manufacturer, according to Thresher, who is now an emeritus researcher at the National Laboratory of the Rockies. Belgian-made WindMaster turbines erected at Altamont Pass signaled the internationalism of the California wind rush. UMass team member Woody Stoddard conducted engineering analyses of many early designs deployed there.Bettman/Getty Images Boston-area entrepreneurs Russell Wolfe and Stanley Charren launched US Windpower with Stoddard and Van Dusen after visiting Heronemus in 1974 and liking what they heard. They adapted WF-1’s design to make it suitable for grid-connected operation, building and breaking prototypes before erecting the world’s first grid-connected wind farm in 1980—20 turbines on a mountain in New Hampshire (https://granitegeek.concordmonitor.com/2017/11/29/nations-first-real-wind-farm-new-hampshire/). California’s water authority placed an order for 100 MW of wind power, and in 1981 US Windpower began installing hundreds of turbines in Altamont Pass (https://www.nytimes.com/1983/02/14/us/private-investors-selling-wind-power-to-utilities.html), east of San Francisco. As more firms jumped to California, drawn by state government incentives, WF-1’s creators and the next cohort of UMass grads assumed important roles in the nascent market. Seven joined Energy Sciences, a startup cofounded by Butterfield. More joined U.S. Windpower. Stoddard left that company to start a consulting firm and ended up advising some of Denmark’s modern wind pioneers, which rapidly expanded thanks to the California market. Those early Danish firms made relatively simple, sturdy machines that subsequently scaled up and dominated globally for several decades — until China embraced wind power. The California wind power boom peaked in 1986, after which energy prices collapsed and incentives faded. Most manufacturers were bankrupted by equipment failures and financial challenges, making the 1990s a tough time for wind power’s pioneers. Many UMass wind engineers, like Butterfield, joined Thresher’s operation at NREL, culling everything they could from the California experience. “An entire generation of U.S. wind engineers got their graduate training, at least in part, using the Wind Furnace.”—Harold Wallace There, Heronemus’s protégés became known as the “UMass Mafia.” Thresher says it attests to the crew’s impact: “There were others. But that UMass Mafia were really leaders in the field. I think that’s the heritage we got from Bill Heronemus. Those people were so impactful and the education they got [with Heronemus] was the key.” What Heronemus began at the university became the UMass Wind Energy Center (https://www.umass.edu/windenergy/home/index.html), which has awarded over 300 graduate degrees. WF-1 now rests in the Smithsonian Institution’s collections (https://americanhistory.si.edu/collections/object/nmah_1389175) in Washington, D.C. It earned its place there, as Smithsonian’s only modern wind turbine, because it represents wind energy’s revival, according to Harold Wallace (https://profiles.si.edu/display/nwallaceh1102006), Smithsonian’s curator for electricity collections. “An entire generation of U.S. wind engineers got their graduate training, at least in part, using the Wind Furnace,” he says. Heronemus didn’t get to witness the production of the massive offshore machines that he foresaw. He lost his long fight with cancer in November 2002, at the age of 82, even as former students and family members were racing to patent his multirotor and floating turbine designs. Had he lived longer, the Captain would almost certainly have railed against current U.S. energy policy. The U.S. government has never backed wind power as generously as he’d hoped. Wind supplied 10 percent of U.S. generation last year—that’s half the share in Europe—with offshore turbines providing only a tiny sliver. Federal support for wind power has been in a stop-go cycle since Ronald Reagan’s administration, and it’s hit a low again under President Donald Trump, who has vowed to stop wind power cold. As Trump boasted to oil executives (https://www.usatoday.com/story/news/nation/2026/01/09/trump-assails-windmills-and-wind-energy-as-junk-theyre-losers/88108694007/) in January: “We have not approved one windmill since I’ve been in office, and we’re going to keep it that way.” Under Trump, stop-work orders have disrupted offshore projects from Massachusetts to Virginia, contributing to a nearly $600 million loss in 2025 for GE Vernova’s wind business (https://www.bostonglobe.com/2026/01/28/business/ge-vernova-offshore-wind-losses/). GE Vernova is the only major wind turbine manufacturer remaining in the United States, and it too can be traced back to Heronemus via a US Windpower patent (https://patents.google.com/patent/US5083039A/en). In stark contrast, European and Asian countries have been going big on offshore wind and are now developing floating wind farms to push into deeper waters. China might be the one to finally conjure up Heronemus’s favored wind design: floating platforms bearing massive multirotor machines. In 2024, Zhongshan-based turbine maker Ming Yang Smart Energy Group (https://en.myse.com.cn/) deployed a two-rotor offshore prototype. The company says its next iteration will generate a whopping 50 MW (https://www.rechargenews.com/technology/mingyang-building-50mw-offshore-wind-turbine/2-1-1888862)—a twin-headed beast that would be the world’s most powerful wind machine. That will be a bittersweet moment for the U.S. wind industry and Captain William Heronemus’s UMass Mafia, for whom such massive machines are a dream come true. Joanne Carroll, a retired member of the UMass Mafia, says she remembers the very moment, her freshman year, when Heronemus’s dream became hers. While he was lecturing in Introduction to Engineering about the hidden costs of coal-fired power, Heronemus walked to the window and said: “‘But out there there’s wind, and you can harvest that energy,’” Carroll recalled. “And I remember thinking: That’s what I want to do with my life.” The author would like to give special thanks to UMass professor emeritus James Manwell for his assistance with this story. From Your Site Articles • Wind-to-Hydrogen Production Reaches Deep Water › (https://spectrum.ieee.org/green-hydrogen-offshore-wind) • If We Want Bigger Wind Turbines, We’re Gonna Need Bigger Airplanes › (https://spectrum.ieee.org/wind-turbine-blade-transport-plane) • Inside the Global Race to Tap Potent Offshore Wind › (https://spectrum.ieee.org/floating-offshore-wind-turbine) Related Articles Around the Web • The Life and Work of Bill Heronemus, Wind Engineering Pioneer | Wind Energy Center › (https://www.umass.edu/windenergy/about/history/heronemus/) Keep Reading ↓ {"imageShortcodeIds":[]} Energy (https://spectrum.ieee.org/topic/energy/)Sponsored Article (https://spectrum.ieee.org/type/sponsored/) What It Takes for Future-Ready Power Distribution (https://spectrum.ieee.org/distribution-grid-modernization) A bolder vision for distribution for an increasingly complex grid Nick Lehnert (https://spectrum.ieee.org/u/nick-lehnert) Nick Lehnert (https://www.bv.com/) is the Distribution Solution Grid Leader at Black & Veatch (https://www.bv.com/), responsible for strategy, delivery, and growth across large‑scale electric distribution modernization programs. He brings more than 20 years of experience helping utilities improve resilience, system design, automation, and scalable program delivery. 03 Jun 2026 4 min read javascript: 8 https://spectrum.ieee.org/distribution-grid-modernization Black & Veatch sees that leading utilities are no longer debating whether to modernize — they’re deciding how quickly they can do it, and how to do it at scale. Black & Veatch mailto:?subject=What%20It%20Takes%20for%20Future-Ready%20Power%20Distribution&body=https://spectrum.ieee.org/distribution-grid-modernization _This sponsored article is brought to you by Black & Veatch (https://www.bv.com/en-US/projects/georgia-power-grid-investment-plan?utm_campaign=portfolio_for_power_utilities-pp-grid_solutions-noia-26-100223&utm_id=26-100223&utm_source=publication&utm_medium=qr-code&utm_content=power-generation&utm_tactic=na&utm_term=brand-awareness\26-bolder-vision-spectrum-native-article). The biggest challenge facing utilities today isn’t what it seems. It’s not demand, even as load growth accelerates. It’s not extreme weather, even as “major events” become routine. It’s not cybersecurity, even as connections expand across the grid. Nick Lehnert, Associate Vice President, Distribution Grid Leader, Black & Veatch. Black & Veatch The real challenge is this: Distribution systems were designed for a different reality. Long gone are the days of predictable demand, one-way power flow and isolated disruptions. At Black & Veatch, we see that leading utilities are no longer debating whether to modernize. They’re deciding how quickly they can do it, and how to do it at scale. Across grid modernization programs globally, three truths consistently emerge. They define what it takes to prepare the distribution system for what’s next: 1. Outage response is not a resilience strategy Resilience is being redefined in real time. A strategy centered on mobilizing crews and restoring service as quickly as possible is reactive, and increasingly insufficient. Resilience has to shift upstream into integrated system design. That starts with hardening. Stronger poles, undergrounding and structural upgrades all have a role, particularly in high-risk corridors. We’re also seeing meaningful gains from how the network is configured and how quickly it can respond without waiting on manual intervention. This is where distribution automation programs can change outcomes. Strategically placed reclosers, automated switches and fault indicators help contain disruptions before they spread. When combined with feeder reconfiguration and updated protection strategies, distribution automation investments allow utilities to set more aggressive recovery targets and achieve measurable reductions in outage duration and customer impact. 2. Future-readiness depends on DERs at scale Forecasting is less and less reliable. Only 19 percent of utilities report strong confidence in their ability to predict future load growth, according to the Black & Veatch 2025 Electric Report (https://www.bv.com/en-US/resources/2025-electric-report).Distributed Energy Resources (DERs) like solar, storage, EVs and behind-the-meter generation are exciting solutions; but they fundamentally change how the system operates. Power is no longer just delivered. It’s injected, stored and redirected in ways the system was never designed to manage. At scale, these challenges show up quickly — particularly on feeders where distributed generation is approaching or exceeding hosting capacity. Protection coordination becomes more difficult when fault current comes from multiple directions. Voltage becomes less predictable as generation fluctuates throughout the day. And planning models must now account for highly variable, location-specific behavior. Distribution modernization is fundamentally changing how the system is designed and operated so it can absorb disruption, manage bi-directional flows and respond in real time. Adapting to bi-directional power flow requires more than incremental updates. Leading utilities are responding by building flexibility into the system, moving beyond static assumptions toward dynamic hosting capacity and interconnection studies, planning that incorporates DER, EV adoption and localized load growth, and infrastructure aligned with the communications and control needed to manage it. 3. The edge must be intelligent, visible and secure As system stress and complexity increase, utilities need far greater visibility and control over the network. Historically, utilities relied on customer calls, Supervisory Control and Data Acquisition (SCADA) at the substation level and field crews to understand what was happening on the system. That model doesn’t hold up. You can’t effectively manage a system you can’t see. Plus, the most critical events are increasingly happening beyond the substation — on feeders, laterals, and at the edge where DER and customer behavior are interacting with the grid. Grid-edge technologies have become essential. Sensors, Advanced Metering Infrastructure (AMI) and automated switching provide the raw data and control needed to move from reactive to proactive operations. In more advanced deployments, utilities are creating centralized control environments that allow operators to see and manage the distribution system in near real time. That capability is enabled by: • Advanced communications networks to form the backbone of real-time grid visibility • Distribution Management System (DMS) and Outage Management System (OMS) to enable faster, more coordinated system response • Analytics, AI and machine learning to improve situational awareness, anticipate system conditions, and support operational decision-making The same connectivity enabling this real-time visibility and control also introduces new vulnerabilities, blurring the line between physical and cyber risk, yet many utilities manage them separately. Only 22 percent have unified teams in place, even as threats continue to rise, including a 50 percent increase in substation attacks and growing exposure to malware and ransomware, according to the Black & Veatch 2025 Electric Report (https://www.bv.com/en-US/resources/2025-electric-report). Cybersecurity and resilient network design must be embedded into the architecture from the outset—not layered on after the fact. See what bolder vision looks like Distribution modernization is fundamentally changing how the system is designed and operated so it can absorb disruption, manage bi-directional flows and respond in real time. To learn about a successful program, check out Georgia Power’s recent grid modernization program (https://www.bv.com/en-US/projects/georgia-power-grid-investment-plan?utm_campaign=portfolio_for_power_utilities-pp-grid_solutions-noia-26-100223&utm_id=26-100223&utm_source=publication&utm_medium=qr-code&utm_content=power-generation&utm_tactic=na&utm_term=brand-awareness_26-bolder-vision-spectrum-native-article). Black & Veatch partnered with the utility on large-scale infrastructure upgrades. The results? Outages are down 76 percent, restoration times have improved by more than 80 percent and communities across Georgia are powered by a grid built to meet the future head-on. When the state faced the most destructive storm in the company’s history, Hurricane Helene, Georgia Power deployed a rapid response team that utilized its “smart grid” and restored power to more than 1 million customers within days. A grid built to meet the future head-on—that’s the result of bolder vision. Keep Reading ↓ {"imageShortcodeIds":[]} Robotics (https://spectrum.ieee.org/topic/robotics/)Whitepaper (https://spectrum.ieee.org/type/whitepaper/) Defining Autonomy for Wellness Robots in Senior Care (https://content.knowledgehub.wiley.com/wellness-robots-and-the-path-to-full-autonomy-a-new-paradigm-in-ai-powered-senior-care/) What defines a wellness robot as a category Dreamface Technologies (https://spectrum.ieee.org/u/dreamface-technologies) Dreamface Technologies (https://dreamfacetech.com/) specializes in social robotics and AI for the aging population, dedicated to enhancing well-being and supporting caregivers. 11 Jun 2026 1 min read javascript: mailto:?subject=Defining%20Autonomy%20for%20Wellness%20Robots%20in%20Senior%20Care&body=https://spectrum.ieee.org/defining-autonomy-for-wellness-robots-in-senior-care An examination of how socially assistive wellness (https://spectrum.ieee.org/tag/wellness) robots could support the seven dimensions of senior wellness, and how a framework can measure their autonomy. What Attendees will Learn 1. Why the senior care crisis exceeds incremental automation (https://spectrum.ieee.org/tag/automation). Demographic pressure, workforce shortages, and a daily wellness-programming gap all strain traditional care models. 2. What defines a wellness robot as a category. The seven ICAA wellness dimensions and eight properties separate these robots from companion and medical devices (https://spectrum.ieee.org/tag/medical-devices). 3. How autonomy can be measured with CRAS. This six-level scale, modeled on the SAEJ3016 driving standard, evaluates four care dimensions. 4. What maps the road to full autonomy. The paper examines technical capabilities, clinical evidence, and a three-phase roadmap toward the early 2030s. Download this free whitepaper now! (https://content.knowledgehub.wiley.com/wellness-robots-and-the-path-to-full-autonomy-a-new-paradigm-in-ai-powered-senior-care/) Keep Reading ↓ Robotics (https://spectrum.ieee.org/topic/robotics/)Guest Article (https://spectrum.ieee.org/type/guest-article/) What Amazon’s Astro Taught Me About Giving Robots a Soul (https://spectrum.ieee.org/amazon-astro-robot-sound) Character is the difference between a machine people tolerate and a product people trust Mike Forst (https://spectrum.ieee.org/u/mike-forst) Mike Forst (https://www.linkedin.com/in/mike-forst-77823718/) is a Los Angeles-based character director and sound lead with 15+ years of experience shaping how technology feels, behaves, and comes alive. He served as Character and Sound Lead on Astro, Amazon’s first consumer robot, and has spent years designing voices and characters for AI. He has worked with companies such as Google, Microsoft, Meta, and Zoox, and currently designs and consults with teams building the next generation of intelligent machines. 8 hours ago 6 min read javascript: https://spectrum.ieee.org/amazon-astro-robot-sound Looking cute is not enough for a robot to have character. Astro: Amazon mailto:?subject=What%20Amazon%E2%80%99s%20Astro%20Taught%20Me%20About%20Giving%20Robots%20a%20Soul&body=https://spectrum.ieee.org/amazon-astro-robot-sound Amazon (https://spectrum.ieee.org/tag/amazon)astro (https://spectrum.ieee.org/tag/astro)consumer robotics (https://spectrum.ieee.org/tag/consumer-robotics)home robots (https://spectrum.ieee.org/tag/home-robots) In 2018, Amazon (https://spectrum.ieee.org/tag/amazon) brought me in as the lead UX Sound Designer for Astro, their first consumer home robot (https://spectrum.ieee.org/amazon-astro-robot). Astro used cameras (https://spectrum.ieee.org/tag/cameras) and other sensors to map and navigate your home and workplace (https://spectrum.ieee.org/ai-robots), and could proactively patrol, check up on loved ones, and transport small items using its built-in cargo bin. While there was a well-defined feature set and form factor, initially there was no character direction. In fact, even before Astro (https://www.amazon.com/Introducing-Amazon-Astro/dp/B078NSDFSB) had a name, there were two main questions—was it simply Alexa (https://spectrum.ieee.org/tag/alexa) on wheels, or was it a robot with its own character? The Astro team was divided. One option was to focus on Alexa, and treat the mobile robot simply as an added utility. I argued for Astro to not focus on Alexa, along with the majority of the UX team. Our belief was that a thing that moves through your home and turns toward you with intent can never be just an appliance. People would ascribe character to whether we wanted them to or not, and so the only question was whether we shaped that character or let it happen by accident. Ultimately, Astro became Astro rather than Alexa (https://www.aboutamazon.com/news/devices/meet-astro-a-home-robot-unlike-any-other), and user testing backed up our decision. People __didn’t__ see the robot as Alexa. They saw it as its own character, and that’s what they wanted it to be. Alexa on the device felt somewhat strange and creepy, but building Astro its own voice was too slow and expensive in 2018. So, we settled on Alexa as a supporting character that handled any actual talking, while Astro was the main character, communicating as much as it could without words, through sound, motion, and facial expressions. I had been brought on to the Astro team to define the robot’s sound design language and voice. But there was no one to flesh out the robot’s actual character. You cannot make a single real decision about a character without defining it first. Every choice about how Astro moved, sounded, paused, or reacted was a character choice, and those choices required all disciplines working together. As Sound Lead, I was weaving together sound, motion, and character, and how they played together inside each story moment. The animators, who programmed Astro’s motion and facial expressions, were extraordinary at what they did, but the emotional arc they were animating came from the sound (and therefore character) work first. So I stepped into that role, which is where my real work started. What I learned about building character for robots applies to nearly everything being built in embodied AI right now. Character Is a Design System Developing a character for Astro meant answering questions that had never been asked about a product at Amazon: What is the emotional range of this robot’s baseline state? How does this robot communicate uncertainty without eroding trust? Where is the line between being expressive and annoying? What are the vulnerabilities of this device’s character? These are design questions. They have real answers, and every team working on the product has to build from them. For example, Astro’s emotional range was designed to be relatively small at first. We never wanted Astro to get too sad or too angry. It could play sad, but would snap out of it quickly and end the reaction on a high note to keep things positive. Character leaks out of every seam and can create a disjointed experience if not defined correctly. Even if it’s just animation timing that’s slightly off, or a response that’s technically correct but contextually tone-deaf, users feel every one of these inconsistencies, even if they can’t name them. Watch what happens at the beginning and end of this Sing sequence: Astro goes from nothing, into the emotional moment, and then lands back on nothing. No build up, no cool down, no sense that the feeling came from somewhere or had anywhere to go. I pushed hard for better character stitching, the transitions in and out of expressive moments that make a performance feel continuous rather than assembled, but it never got implemented. The moment itself works. But without the stitching, it reads as a clip playing on a robot rather than coming from within the robot character itself. Story and Sound at the Beginning We had decided that Astro would have no spoken dialogue, but it had something that functioned the same way: a vocabulary of sounds, tones, and rhythms that acted as its voice. This vocabulary became the leading output of the character’s personality. The robot’s motion and facial expressions were built around it. Astro’s wake-up sequence is a great example. Waking wasn’t just a boot animation on the screen; it was an entire performance. Slow and humble at first, the robot oriented itself quietly, then stretched its screen, checked its wheels, and finally, with an upward gesture toward its telescoping mast, it popped it up slightly, and did a little dance of joy. Sound, motion, and eyes hit every beat together in full choreography. The character’s output in that sequence was first written as a story. Astro is waking up in its new home for the first time. Its main aspiration is to be part of a family, so this is the moment it has been waiting for, this is its purpose. Being the responsible character that it is, it wants to make sure everything is good to go before it introduces itself and starts learning its new home. This narrative came first because it drove every other decision that we made. After the story was written, sound gave that story a metaphorical voice: the excited tones, the pacing as it checked its wheels, and the bright melodic phrase as Astro looked up at its new family for the first time and introduced itself. Once the sound was laid down, animation did their thing with motion and facial expressions, taking cues from the emotional arc the sound had established. Motion didn’t lead—it followed the feeling of the story and the sounds, the same way an animator follows a recorded vocal take. That wake up sequence became one of the most-discussed moments in early user testing. People described it as “alive.” What they were responding to wasn’t any single element. It was all three channels (sound, motion, and facial expressions) expressing the same defined character in harmony. Context Is Where Character Becomes Real The most compelling characters are defined not by a fixed disposition but by how they respond to their environments and the people in them. They’re still recognizably themselves even as they adapt. This is what I call contextual character. A robot living in a home doesn’t occupy a single emotional state. It moves through rooms with different energy, encounters people in different moods, operates at different times of day, and responds to an endless range of social situations it was never explicitly designed for. We got close to a contextual character output with Astro’s sound. When a specific piece of environmental context was fed in, the system adapted beautifully, and Astro felt completely alive. But every state like this was still a prediction we made by hand—a situation we had to imagine in advance and design a response for. A random home throws more situations at a robot than anyone can possibly predict, so there was always a longer tail of moments the system was never prepared for. The difference between a product people describe as “smart” and one they describe as “aware” often comes down to this. Smartness is capability. Awareness is context. Presence is character. And character is always in reaction to the people around it, to its environment, to its own evolving state. That’s what makes it feel like something is emotionally present with you. This is where AI changes the game for character design in ways that go well beyond what was possible with Astro. AI-driven adaptation doesn’t require the contextual predictions that we relied on. It learns the specific rhythms, preferences, and emotional context of the people it lives and works with. The character doesn’t just respond to context. It __grows__ into it. What Industry Is Missing The character and soul of the impending wave of embodied AI products appears to almost always be an afterthought. And character defined late is character defined by default. It becomes the sum of a thousand small decisions made by different people thinking about anything but character. People project character onto devices whether you plan for it or not, especially if those devices move—a robot that moves is __already__ a character. If nobody has designed this character, the result will be products that feel like nothing, or worse, feel confusing and not trustworthy. Technically impressive, but lifeless. We did not get this fully right with Astro. So many things were moving in parallel that character was rarely treated as a utility, and it made sense why. When you are building a first-of-its-kind product, the things that are the loudest are the ones that break, the deadlines, the costs, the features a customer can point to on a box. Character is quieter than all of that. It’s easy to assume it can come later. On a team as large as the Amazon Astro team, it’s lucky to get any idea onto the roadmap when it is competing with a hundred others that all feel more urgent in the moment. None of this came from people not caring. It came from character being the kind of thing that is hard to prioritize until you see what its absence costs you. My Asks to Product Leaders If you are building a product that will share physical or conversational space with people, three things are worth considering: • *Define character before you define interactions.** You need a defensible character with enough emotional logic to answer hard questions consistently. Find answers to character questions early, and have every discipline build from the same foundation. • *Build story and sound into the character pipeline, not the production pipeline.** Story and sound developed alongside character definition has the chance to inform motion, expression, and interaction logic. This requires a different kind of collaboration, and a different kind of hire. • *Design for adaptation, not just consistency.** A consistent character is necessary, but the products that will matter most in people’s lives are the ones that deepen through use. The infrastructure to support that is more and more accessible, but the design thinking to take advantage of it is still rare. __An unabridged version of this story can be read on Medium (https://medium.com/@mikeforstmusic/what-amazons-astro-taught-me-about-giving-ai-a-soul-989fcd9c45f4).__ From Your Site Articles • Years Later, Alphabet’s Everyday Robots Have Made Some Progress › (https://spectrum.ieee.org/alphabet-robots) • Where’s My Robot? › (https://spectrum.ieee.org/ai-robots?itm_source=parsely-api) • Amazon’s Astro Is a Mobile Alexa and Cup Holder that Costs $1k › (https://spectrum.ieee.org/amazon-astro-robot) Related Articles Around the Web • Meet the Amazon researchers helping robots understand noise—and listen to the results › (https://www.aboutamazon.com/news/devices/meet-the-amazon-researchers-helping-robots-understand-noise-and-listen-to-the-results) • Amazon Astro Review: It's Cute, Getting More Automated and Not Worth It Yet - CNET › (https://www.cnet.com/home/smart-home/amazon-astro-review/) • Amazon.com: Amazon Astro, Household robot for home monitoring, with Alexa, Includes a Ring Home Trial : Everything Else › (https://www.amazon.com/Introducing-Amazon-Astro/dp/B078NSDFSB) Keep Reading ↓ Get the latest technology news in your inbox Subscribe to IEEE Spectrum’s newsletters by selecting from the list. • Tech Alert • AI Alert • The Next Watt • Career Alert • Robotics News • The Future Lane • University Spotlight • Product Spotlight Email input • I agree to the IEEE Privacy Policy (https://www.ieee.org/security-privacy.html) Please select one or more newsletters, enter a valid email address and accept Privacy Policy Subscribe (https://spectrum.ieee.org/large-language-models-ieee-course) Close (javascript:) You have succesfully subscribed to the newsletters below: Thank your for your subscription. Computing (https://spectrum.ieee.org/topic/computing/)Magazine (https://spectrum.ieee.org/magazine/)Feature (https://spectrum.ieee.org/type/feature/) Why Orbital Data Centers Are Harder Than Silicon Valley Thinks (https://spectrum.ieee.org/orbital-data-centers-heat) Shedding heat will require ingenious new designs Andrew Cavalier (https://spectrum.ieee.org/u/andrew-cavalier1) 11 Jun 2026 10 min read javascript: 20 mailto:?subject=Why%20Orbital%20Data%20Centers%20Are%20Harder%20Than%20Silicon%20Valley%20Thinks&body=https://spectrum.ieee.org/orbital-data-centers-heat https://spectrum.ieee.org/orbital-data-centers-heat Edmon de Haro • *“Space computing, the final** frontier, has arrived,” Nvidia (https://spectrum.ieee.org/tag/nvidia) CEO Jensen Huang declared (https://nvidianews.nvidia.com/news/space-computing) at the Nvidia GTC (https://www.nvidia.com/gtc/) conference in March. Indeed, the idea of data centers in orbit has gone from science fiction to a serious spending category. Elon Musk’s SpaceX (https://www.spacex.com/) has acquired (https://x.ai/news/xai-joins-spacex)xAI (https://x.ai/) (also Musk’s) and is planning (https://spacenews.com/spacex-offers-details-on-orbital-data-center-satellites/) a constellation of space-based data centers. Google (https://research.google/), not to be outdone, announced Project Suncatcher (https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/) in partnership with Planet (https://www.planet.com/), planning to launch two satellites equipped with Google (https://spectrum.ieee.org/tag/google)Tensor Processing Unit (https://spectrum.ieee.org/tag/tensor-processing-unit) (TPU (https://spectrum.ieee.org/tag/tpu)) AI chips (https://spectrum.ieee.org/tag/ai-chips) by early 2027. Startup Starcloud (https://www.starcloud.com/) has already filed (https://www.pcmag.com/news/data-center-space-race-heats-up-as-starcloud-startup-requests-88000-satellites?test_uuid=04IpBmWGZleS0I0J3epvMrC&test_variant=B) a proposal with the Federal Communications Commission for an 88,000-satellite constellation for orbital data centers. As Starcloud’s filing suggests, these companies are all proposing fleets of satellites numbering in the thousands, each housing a rack or multiple racks of AI-grade GPUs (https://spectrum.ieee.org/tag/gpus), interconnected with each other through free-space optical links and communicating back to Earth via microwave links, either directly or through other satellites. Proponents tout (https://x.com/patrick_oshag/status/1998440819078898140) the many wonders of computing in space: abundant solar energy (https://spectrum.ieee.org/tag/solar-energy), free cooling, and freedom from Earth-based disturbances like earthquakes, floods, and protesters. But a sober look at the physics of space-based computing paints a much more nuanced picture. Free cooling is perhaps the biggest misconception. Space is cold, but it also has no atmosphere. That means the best heat-removal mechanisms, conduction and convection, are off the table. The only option is radiation. To prevent a chip from overheating in space, a large, costly surface area is required to dissipate the energy and then radiate it. Solar energy is abundant, but collecting it with functional solar panels (https://spectrum.ieee.org/tag/solar-panels) that maintain perfect alignment toward the sun is a complex task requiring extensive attitude control systems (https://spectrum.ieee.org/satellite-refueling-heats-up). On top of that, ionizing radiation in space from cosmic rays and other sources poses a unique challenge, degrading the solar panels, the radiative coolers, and the chips themselves. Because regular maintenance in space is difficult, redundancy has to be built in at launch, and cost estimates have to account for efficiency degradation over time. At ABI Research (https://www.abiresearch.com/), where I work as an aerospace analyst, we did a rough total-cost-of-ownership comparison between a data center on Earth and one in space. It showed that the cost to launch and run a GPU in space for a year is at least an order of magnitude higher than the same feat in a terrestrial data center. Our model was simple, assuming an Nvidia H100 server rack launched with the requisite-size solar panel and radiator on a spacecraft akin to Starcloud’s pilot launch (https://spectrum.ieee.org/nvidia-h100-space). We assumed SpaceX’s Starship was used at a highly optimistic launch cost per kilogram of US $44, and a terrestrial energy cost of $0.20 per kilowatt hour. This is a simple back-of-the-envelope calculation, but it does signal something real. From our perspective, the cost of delivery and space hardening of the payload makes general-purpose space-based data centers difficult to justify economically today, despite the fact that data-center builders in many regions are scrambling for electric power (https://spectrum.ieee.org/tag/electric-power). However, there are niche applications where the much higher costs of computing in space could be justified. Examples include preprocessing data from Earth-observation satellites, real-time detection and tracking of hypersonic missiles, and active collision avoidance (https://spectrum.ieee.org/tag/collision-avoidance) in the increasingly crowded low Earth orbit (https://spectrum.ieee.org/tag/low-earth-orbit). Even for these, though, contending with fundamental physics will still be a demanding challenge. And a technologically compelling one, too. The Cooling Challenge in Space Cooling is where physics separates the science from the fiction. The governing equation for radiative cooling (https://spectrum.ieee.org/tag/radiative-cooling), the only type of cooling available in space, is known as the Stefan-Boltzmann Law. It states that the amount of power you can radiate is proportional to the area of the radiator times its temperature to the fourth power. For a space systems architect, the implications of this law are brutal. In orbit, the only variable we can control is area. This restriction creates a geometric penalty, or a “physics tax,” for cooling in space: The more power you need to reject, the bigger the area of the radiator you need to bring along from Earth. The only cooling method available in space is radiation, and the radiator area required is derived using the Stephan-Boltzmann law. For a single chip drawing 700 watts, like Nvidia’s popular H100 GPU, the area required to keep it at 20 °C is just under 3 square meters, and it goes down to 1 square meter for an operating temperature of 85 °C. However, as the radiator surface is exposed to ionizing radiation, its emissivity decreases, and after 5 years in space the required area increases by about 40 percent. To understand how big this baseline area is in practice, I used the Stefan-Boltzmann law to model the heat-rejection area needed to keep a single chip that draws 700 watts of power—such as the H100 GPU chip, an AI stalwart—at a constant 60 °C, usually considered the sweet spot for GPU longevity and stability. I further assumed that the radiator is perfectly facing deep space (https://spectrum.ieee.org/tag/deep-space), at a chilly background temperature of 3 kelvins. By this calculation, a single chip would require 1.4 square meters of radiator surface. To put this into perspective, consider that a common AI rack can hold approximately 32 GPUs (four H100 server boards). With CPUs, memory, and networking equipment, this rack would draw around 40 kilowatts of power. This single rack includes 2.5 terabytes of memory—enough capacity to serve over 20,000 concurrent users or run 16 simultaneous instances of Llama (https://spectrum.ieee.org/tag/llama) 3, an open-source AI model. But to cool this thermal load in a vacuum, that single rack would require an 80-square-meter radiator, roughly the size of a pickleball court. For an aggregate 100-megawatt data center, you’d need at least 2,500 of those radiators. And that’s the best-case scenario. Additional problems are hidden in the low Earth orbit environment itself. Space exposes radiators and their coatings to a chemically hostile brew of ultraviolet light (https://spectrum.ieee.org/tag/ultraviolet-light) and atomic oxygen, quite the opposite of a clean-room environment. Over a LEO satellite’s typical 5-year lifespan, these elements degrade the radiator’s surface properties and lower its ability to shed heat. Including this degradation in the model reveals that as the radiator degrades from a “fresh” state to an “end-of-life” state, the physics demands a further penalty. To maintain that same 60 °C operating temperature for the GPU chips, the required surface area jumps from about 1.4 square meters per chip to nearly 2.0 square meters. In other words, the physics tax rises by 40 percent. Therefore, you must launch at least 40 percent more radiator mass, endure higher atmospheric drag, and sacrifice valuable launch volume just to survive the degradation of the thermal coating. This increase adds significantly to the launch cost and further erodes the economics of a space-based data center. The Silicon Challenge in Space Solving the heat problem is only part of the battle. The other significant challenge in low Earth orbit is ionizing radiation, which affects the computing hardware itself. Today’s satellites typically use radiation-hardened (https://spectrum.ieee.org/tag/radiation-hardened) processors, which are very reliable but also much more expensive, and they perform poorly compared to commercial off-the-shelf processors (https://ieeexplore.ieee.org/document/11068401). A standard rad-hard chip doesn’t have the processing power to run a modern large language model (LLM). As a result, satellite operators aspiring to launch a data center have no choice but to make a risky compromise: to use hardware meant for terrestrial use. In order to achieve the necessary compute density, orbital data centers must use the same Nvidia H100s or Google TPUs found in terrestrial server farms (https://spectrum.ieee.org/tag/server-farms). The problem is that these chips are “soft” targets in space. High-energy particles can flip bits in memory or cause “latch-ups” in logic that fry the circuit. Three big proposals for orbital data centers vary in satellite size, number, and cooling plans. One possible option is to shield the computers from radiation with thick, absorbent panels. However, the shielding would add significantly to the already heavy satellites. The other option is to compensate for the radiation damage with redundancy. Indeed, edge computing (https://spectrum.ieee.org/tag/edge-computing) architects are moving toward software-defined resilience, where instead of one perfectly hardened computer, operators fly a cluster of imperfect, commercial ones whose total cost could be as low as one-tenth to one-hundredth that of the rad-hard model. This redundant approach is used in many spacecraft, including Artemis II (https://cacm.acm.org/news/how-nasa-built-artemis-iis-fault-tolerant-computer/), which recently carried astronauts (https://spectrum.ieee.org/tag/astronauts) around the moon, as well as SpaceX’s flight computers and the Hewlett Packard Enterprise edge servers for the International Space Station. By running three (or more) instances of the same calculation on three different nodes and comparing the answers, the system can detect a corrupted processor. If a node fails, the “orchestrator” reboots it while the others continue the mission. While this ensures resiliency, it also means that some fraction of the compute capacity is dedicated to redundancy, further increasing the costs. The Energy Challenge in Space An often-touted advantage of space-based data centers is the seemingly unlimited supply of free, clean energy (https://spectrum.ieee.org/tag/clean-energy) from the sun. Solar energy in orbit is indeed abundant, at 1,361 watts per square meter. Of course, capturing that free energy is made possible only by the very costly launching of large solar panels into orbit. And those solar panels also degrade over time due to radiation exposure, typically losing 1 to 3 percent efficiency per year. Let’s say a solar array collects 1 MW of power to run an AI cluster. The laws of physics demand that the satellite must eventually radiate 1 MW of waste heat. Because the square area needed to generate the solar power—around 400 W/m2 (https://www.energydawnice.com/solar-panel-output-per-square-meter/)—and to reject the heat—around 450 W/m2—are nearly equivalent, every square meter of power generation now demands approximately another square meter of cooling. The radiator needs to be a structural equal, not merely a passive coating on a surface used for something else. As Elon Musk recently noted (https://www.youtube.com/watch?v=IgifEgm1-e0) in Davos, the most efficient radiator is one that never sees the sun. By orienting the spacecraft so the solar panels face the sun and the radiators face the deep vacuum of space, efficiency skyrockets for both. But there’s a catch: Maintaining this perfect three-way alignment—panels to sun, radiator to the void, antennas (https://spectrum.ieee.org/tag/antennas) to Earth—requires complex, high-torque attitude control systems. So this configuration means more payload and more computing power. Plus, these control systems are complex components with many failure modes, which is not optimal in a situation where maintenance is difficult. The Killer Apps for Computing in Space Given all these challenges of deploying massive radiators for satellites in the hostile environment of space, why build data centers in space at all? While training or inference on LLMs in space doesn’t seem economical today, there are other, very compelling applications for computing in space. Here are two: solving the downlink bottleneck from Earth-observation satellites and enabling collision-preventing maneuvers in the increasingly crowded low Earth orbit. The latest Earth-observation satellites, equipped with hyperspectral and synthetic aperture radar sensors, are used for a range of important reconnaissance missions, such as battlefield intelligence, tracking the global shadow fleet of ships carrying contraband, and assessing earthquakes or infrastructure failures down to the millimeter. These systems can generate hundreds of terabytes of raw data per day that must be transmitted to Earth. However, the radio-frequency “pipes” used to downlink the data are congested, and the ground infrastructure cannot absorb the sheer volume of raw data. Another immediate, mission-critical application for in-space computation is protecting the orbital environment. With over 17,000 satellites in orbit, the overwhelming majority of which are in low Earth orbit, avoiding collisions between these satellites is crucial. As NASA astrophysicist Donald Kessler (https://en.wikipedia.org/wiki/Donald_J._Kessler) pointed out back in 1978, a single space collision could cause a cascading effect that renders the entirety of LEO unusable. RELATED: Have We Reached a Space-Junk Tipping Point? (https://spectrum.ieee.org/kessler-syndrome-space-debris) According to SpaceX’s recent annual report, the Starlink constellation executes a collision avoidance maneuver every 2 minutes on average. Each maneuver already relies (https://spacexstock.com/25000-collision-avoidance-maneuvers-lessons-from-starlink/) on onboard AI systems but still requires most of the processing to happen on the ground. SpaceX’s Starlink system currently has over 10,000 satellites in low Earth orbit, each depicted here as a colored dot. Satellitemap.space As low Earth orbit gets increasingly populated, collision avoidance will have to break the traditional ground-loop model. In the megaconstellation era of space, the OODA (observe, orient, decide, act) loop must happen onboard, thereby reducing the analysis turnaround from minutes to milliseconds. The problem is that the flight computers standard on satellites are not built for this level of processing. The complex probability models required for maneuvering cannot currently be implemented by onboard computers in conjunction with their navigation systems. Clearly, more powerful computers are needed. This is the true economic justification for moving compute to space: to move insight generation there. By placing high-performance computing adjacent to the sensors, we can process terabytes of data in orbit and downlink only the relevant data in real time, and we can do the computations necessary to avoid satellite collisions in real time. The Future of Computing in Space So, assuming that some form of computing is inevitable in low Earth orbit in the foreseeable future, how will the heat be handled? The industry is currently experimenting with two main classes of solutions to cope with the Stefan-Boltzmann law. One creative option is to useorigami-inspired radiators, the kind used for the James Webb telescope. Companies are developing flexible, high-conductivity composite radiators that fold into a tight cube for launch and unfurl into enormous yet lightweight thermal wings in orbit. Another possibility is to useliquid-droplet radiators. This concept proposes removing the rigid radiator structure completely and instead spraying a stream of coolant oil directly into the vacuum of space. The fluid travels through an open loop, exposed to the near-absolute zero of the void, maximizing radiative surface area before being caught by a collector and pumped back into the ship. It sounds like science fiction, but as the heat loads climb into the megawatts, liquid-droplet cooling may be the only way to cheat the mass limits of this exponential reality. Options for Future Radiator Design There are at least two proposed radiator designs that may make cooling in space more economical. One is liquid-droplet radiators, which produce a constant stream of oil droplets directly into outer space, to be collected after traveling some distance. This way, the whole surface area of each droplet radiates away heat. Another is origami-inspired radiators, such as the ones used on the James Webb telescope. These bundle up into a tight cube for launch and unfurl into large, light radiators in space. Chris Philpot Our rough total-cost-of-ownership model uses optimistic versions of current numbers, such as launch cost, chip cost, and power use. A critic might point out that future technology will improve, both in efficiency, purpose-built designs, and costs. Sure, the technology is bound to improve. But the critical factor isn’t just launch cost; it’s the computing power per unit mass and electric-power economics. Radiators and solar arrays can consume 65 to 70 percent of total satellite mass, and space-grade photovoltaics run orders of magnitude more expensive than terrestrial equivalents. Chris Philpot Even as launch costs fall, the mass and cost burden of power generation and thermal management will remain a fundamental problem. Current space-grade solar panels rely on germanium substrates, whose supply is concentrated in China. It will be extremely difficult to scale up availability of these substrates. A transition to radiation-tolerant perovskite solar panels or a similar alternative could change the economics significantly, but that possibility is five years away or more. The technology will get cheaper, but the bottlenecks of power and thermal architecture will remain. Recognizing the thermal reality of cooling in space forces us to shift how we view satellite operations. We are moving away from the “launch and forget” era toward an era of “autonomous logistics.” As our thermal model demonstrated, the harsh environment of space steadily attacks the hardware. UV radiation degrades thermal coatings; cosmic rays degrade silicon. In a traditional satellite model, when the radiator degrades or the memory fails, the satellite becomes space junk. For a multimillion-dollar data center, that disposal model is potentially ruinous. To make the economics of orbital computation work, the infrastructure must be serviceable and the rockets to launch them reusable. The orbital domain will require automated servicing vehicles capable of swapping out degraded radiator panels and upgrading fried servers. In these ways, the future of the orbital data centers is dependent on the innovations of an emergent in-space economy. There’s a good argument to be made that the need for space-based computation is less of a hype cycle and more of an enabler for the new space economy. Look no further than SpaceX’s recent regulatory filings proposing a constellation of up to a million satellites in low Earth orbit. At such a scale, routing all raw data back to Earth is physically impossible; the network itself must become the data center. However, the winners in this sector will be determined by the systems architects who most cleverly accommodate the thermodynamics and the companies with sufficient vertical integration to take on the massive costs of operating data centers in orbit. Ultimately, the physics tax is universal. Whether managing heat rejection in the vacuum of low Earth orbit or managing power density in a hyperscale facility in Northern Virginia, the constraint is never the silicon. It’s the thermodynamics. From Your Site Articles • Nvidia Sends a Powerful GPU to Space › (https://spectrum.ieee.org/nvidia-h100-space) • How Stupid Would It Be to Put Data Centers in Space? › (https://spectrum.ieee.org/orbital-data-centers) Related Articles Around the Web • NVIDIA Launches Space Computing, Rocketing AI Into Orbit | NVIDIA Newsroom › (https://nvidianews.nvidia.com/news/space-computing) • Orbital data centers, part 1: There’s no way this is economically viable, right? - Ars Technica › (https://arstechnica.com/space/2026/03/orbital-data-centers-part-1-theres-no-way-this-is-economically-viable-right/) Keep Reading ↓ {"imageShortcodeIds":[]} Aerospace (https://spectrum.ieee.org/topic/aerospace/)Sponsored Article (https://spectrum.ieee.org/type/sponsored/) Meet NASA Low Outgassing Standards With Adhesives for Aerospace and Optical Systems (https://spectrum.ieee.org/adhesive-outgassing-nasa-standards) Learn how outgassing affects optical, semiconductor, and aerospace systems — and how to prevent it Master Bond (https://spectrum.ieee.org/u/master-bond) Master Bond (https://www.masterbond.com/) has been formulating and manufacturing Adhesives, Sealants & Coatings since 1975. Master Bond products are freshly made-to-order for maximum shelf life and shipped directly from our USA manufacturing facility. 26 May 2026 2 min read javascript: 2 https://spectrum.ieee.org/adhesive-outgassing-nasa-standards Generic adhesives allow volatile molecules to escape through a loosely bonded polymer network (left). NASA-compliant low outgassing adhesives use a highly cross-linked structure to keep contamination in check (right). Master Bond mailto:?subject=Meet%20NASA%20Low%20Outgassing%20Standards%20With%20Adhesives%20for%20Aerospace%20and%20Optical%20Systems&body=https://spectrum.ieee.org/adhesive-outgassing-nasa-standards This sponsored article is brought to you by Master Bond (https://www.masterbond.com/). Outgassing is the release of volatile substances from a cured adhesive over time. These released materials, which may include residual solvents, unreacted monomers, or other chemical species, can deposit on nearby surfaces, causing contamination that interferes with sensitive components. What Is Outgassing and How Is It Measured? The industry standard for measuring outgassing is ASTM E595, developed by NASA (https://www.masterbond.com/certifications/nasa-low-outgassing). This test exposes a cured sample to 125 °C at high vacuum (10⁻⁵ to 10⁻⁶ torr) for 24 hours, measuring Total Mass Loss (TML) and Collected Volatile Condensable Materials (CVCM). To meet NASA low outgassing requirements, materials must exhibit less than 1 percent TML and less than 0.1 percent CVCM. Optical assemblies need contamination-free bonding and prevention of fogging the optics to maintain clarity. High-vacuum scientific equipment, semiconductor manufacturing (https://spectrum.ieee.org/tag/semiconductor-manufacturing) tools, and aerospace electronics also demand low outgassing materials. Key Applications Low outgassing adhesives (https://www.masterbond.com/properties/low-outgassing-adhesives) are essential wherever contamination could compromise performance and this is particularly relevant for space and satellite systems. Optical assemblies (https://www.masterbond.com/industries/adhesives-sealants-and-coatings-optical-industry), including cameras, telescopes (https://spectrum.ieee.org/tag/telescopes), and laser systems, need contamination-free bonding and prevention of fogging the optics to maintain clarity. High-vacuum scientific equipment, semiconductor manufacturing tools, and aerospace electronics also demand low outgassing materials. Even terrestrial optical devices benefit from reduced outgassing to ensure long-term reliability. EP30-2 is a versatile system can be used in a variety of applications in aerospace, electronic, optical and specialty OEM industries, especially when optical clarity and low outgassing are important criteria.Master Bond (http://spectrum.ieee.org/engineering-resources/master-bond) Ensuring Low Outgassing Performance Through Proper Handling Achieving specified outgassing performance requires attention to storage, mixing, and curing. For two-part systems, use the correct mix ratio and mix thoroughly to ensure complete reaction. Follow recommended cure schedules — adding heat, even at modest temperatures of 150-200 °F, significantly improves cross-linking and reduces outgassing. For UV-curable adhesives, ensure complete cure by using the correct lamp wavelength (typically 365 nm), adequate intensity, and proper exposure time with no shadowed areas. Troubleshooting Outgassing Issues If contamination appears on optical surfaces or outgassing test results are higher than expected, an incomplete cure might be one of the root causes. The first step is to verify that the adhesive has fully hardened to its specified Shore hardness. The next step is to consider adding or extending heat cure to improve cross-linking. Master Bond Product Recommendations Master Bond offers a range of adhesives meeting NASA low outgassing requirements. EP30-2 (https://www.masterbond.com/tds/ep30-2) and EP21TCHT-1 (https://www.masterbond.com/tds/ep21tcht-1) are some examples of two-part epoxy systems that have been successfully deployed in demanding vacuum applications, including ultra-high vacuum environments. For applications requiring UV cure, Master Bond provides specialty UV formulations such as UV16 (https://www.masterbond.com/tds/uv16) meeting ASTM E595, as well as dual-cure systems (UV plus heat) such as UV22DC80-10F (https://www.masterbond.com/tds/uv22dc80-10f) for assemblies where shadows prevent complete UV exposure. These dual-cure products initiate with UV light and complete curing with heat as low as 180 °F (80 °C). Keep Reading ↓ Telecommunications (https://spectrum.ieee.org/topic/telecommunications/)Aerospace (https://spectrum.ieee.org/topic/aerospace/)Whitepaper (https://spectrum.ieee.org/type/whitepaper/) Direct-to-Cell Technology: Enabling Satellite Connectivity for Legacy Devices (https://content.knowledgehub.wiley.com/direct-to-cell-technology-enabling-satellite-connectivity-for-legacy-devices/) How DTC works as a spaceborne cell tower Rohde & Schwarz (https://spectrum.ieee.org/u/rohde-schwarz) Rohde & Schwarz (https://www.rohde-schwarz.com/) is one of the world’s leading manufacturers of test and measurement, secure communications, monitoring and network testing, and broadcasting equipment. 02 Jun 2026 1 min read javascript: mailto:?subject=Direct-to-Cell%20Technology%3A%20Enabling%20Satellite%20Connectivity%20for%20Legacy%20Devices&body=https://spectrum.ieee.org/direct-to-cell-technology-enabling-satellite-connectivity-for-legacy-devices Direct-to-cell technology uses LEO satellites (https://spectrum.ieee.org/tag/satellites) as spaceborne cell towers. It delivers LTE (https://spectrum.ieee.org/tag/lte) services to existing smartphones (https://spectrum.ieee.org/tag/smartphones) without hardware changes, bridging global coverage gaps. What Attendees will Learn 1. How DTC works as a spaceborne cell tower — LEO satellites carry LTE eNodeB payloads in regenerative mode. How they serve unmodified phones using quasi-earth-fixed multi-beam antennas (https://spectrum.ieee.org/tag/antennas). How the satellite compensates for Doppler shift and time delay on thenetwork side. 2. Why Doppler shift and round-trip time are critical challenges — A LEO satellite’s high velocity causes carrier frequency offsets in OFDMA systems. Pre-compensation at a reference point helps, but cell-edge users still face residual Doppler. 3. How spectrum sharing (https://spectrum.ieee.org/tag/spectrum-sharing) and regulation shape DTC deployment — DTC has no dedicated spectrum allocation. It relies on spectrum sharing between terrestrial and satellite operators or re-farmed MSS bands. How national regulations like the FCC (https://spectrum.ieee.org/tag/fcc) SCS framework govern access. 4. Where DTC fits in the evolution toward 5G (https://spectrum.ieee.org/tag/5g) NTN and 6G (https://spectrum.ieee.org/tag/6g) — DTC is an interim technology offering fast time-to-market satellite services. It bridges the gap until 3GPP NR-NTN matures. How NR-NTN will bring purpose-built NTN features and international spectrum frameworks. Download this free whitepaper now! (https://content.knowledgehub.wiley.com/direct-to-cell-technology-enabling-satellite-connectivity-for-legacy-devices/) Keep Reading ↓ Aerospace (https://spectrum.ieee.org/topic/aerospace/)News (https://spectrum.ieee.org/type/news/) Optical Metasurface Captures Solar Magnetic Fields (https://spectrum.ieee.org/optical-metasurface-solar-telescope) The telescope-integrated device reveals the signal in snapshots of polarized sunlight Rachel Berkowitz (https://spectrum.ieee.org/u/rachel-berkowitz) Rachel Berkowitz (http://www.rberkowitz.net/) is a freelance science writer and editor with a Ph.D. in geophysics from the University of Cambridge. She is a corresponding editor at the American Physical Society's Physics Magazine. Her work has appeared in Scientific American, New Scientist, Science News, Physics Today, and the newsrooms of several U.S. national laboratories. 18 Jun 2026 4 min read javascript: https://spectrum.ieee.org/optical-metasurface-solar-telescope Gianna Kim, a student at UC San Diego, created this illustration depicting polarized sunlight swirling in the sky as it enters the Dunn Solar Telescope. Gianna Kim mailto:?subject=Optical%20Metasurface%20Captures%20Solar%20Magnetic%20Fields&body=https://spectrum.ieee.org/optical-metasurface-solar-telescope optical metasurface (https://spectrum.ieee.org/tag/optical-metasurface)astronomy (https://spectrum.ieee.org/tag/astronomy)polarization (https://spectrum.ieee.org/tag/polarization)nanopillars (https://spectrum.ieee.org/tag/nanopillars)telescopes (https://spectrum.ieee.org/tag/telescopes) A new nanopatterned structure based on sub-wavelength physics (https://spectrum.ieee.org/metamaterials-camera-chip) has proved its value in a real-world setting, and it could qualify for use in outer space. The device, integrated with one of the world’s most advanced solar observatories, captured pictures of the Sun’s magnetic field in a new, advantageous way: via a single snapshot with no moving parts. The demonstration (https://www.science.org/doi/10.1126/sciadv.aee8035), published 10 June in __Science Advances,__ offers a promising tool for astronomy (https://spectrum.ieee.org/tag/astronomy), consumer electronics (https://www.wired.com/story/metalenz-polareyes-polarization-camera/), quantum optics (https://spectrum.ieee.org/tag/quantum-optics), and other applications (https://spectrum.ieee.org/metasurface-displays) that involve measuring polarized light. The structure is based on an optical metasurface (https://spectrum.ieee.org/tag/metasurface), which refers to a patterned array engineered at sub-wavelength scales (https://spectrum.ieee.org/optical-metamaterials-ai-data-centers) to manipulate the diffraction of the wavefront. That’s not unlike classical diffraction gratings, whose periodic etchings split light into different colors and directions. But metasurfaces (https://spectrum.ieee.org/tag/metasurfaces) offer another advance: They can split light into its polarized components. Researchers in the lab of Noah Rubin (https://nrubin.ucsd.edu/noahrubin/), professor of electrical and computer engineering (https://spectrum.ieee.org/tag/computer-engineering) at University of California, San Diego, created the device. “What’s special about metasurfaces is you can design an array of elements that respond in one way to this polarization (https://spectrum.ieee.org/tag/polarization) and in another way to that polarization,” says Rubin. “This is a capability that’s only fully emerged in recent years.” Now the metasurface is poised to “leave the lab and go into a serious piece of scientific instrumentation, possibly for one of the first times,” he says. Integrating the component with the Dunn Solar Telescope (https://nso.edu/telescopes/dunn-solar-telescope/dunn/) at the National Solar Observatory in Sunspot, New Mexico (https://spectrum.ieee.org/tag/mexico), showed that it could perform key measurements that today normally rely on complex rotating components—an advance that may one send it on a space mission. Polarized light reveals space weather (https://spectrum.ieee.org/tag/space-weather) Studying our Sun is fundamental for predicting space weather and the coronal mass ejections (https://spectrum.ieee.org/tag/coronal-mass-ejections) that influence life on Earth. The key to these dynamics is magnetism (https://spectrum.ieee.org/tag/magnetism), which we can detect from afar with polarized light. Although light emitted from the Sun starts out unpolarized, magnetic fields (https://spectrum.ieee.org/tag/magnetic-fields) on the Sun’s surface and in its atmosphere polarize some of the light. Measurements of the polarized light provide a set of parameters from which astronomers can deduce the magnetic field strength and direction. But the classic way of doing this involves an optical component that rotates within a camera, taking separate measurements of multiple parameters before reconstructing the full image. This demonstration integrated the device with a ground-based telescope (https://spectrum.ieee.org/tag/telescope), but it could also be used with space-based telescopes (https://spectrum.ieee.org/tag/telescopes) in the future. “If you’re thinking about a space mission, you don’t want a moving component. It’s a single point of failure,” says Rubin. Moreover, measuring subtle polarization signatures requires advanced hardware to compensate for jitter in the satellite itself. Simultaneously acquiring all of a scene’s polarization data using a passive component could greatly relax the engineering requirement of future satellite-borne imagers (https://spectrum.ieee.org/tag/imagers). That’s where a diffraction grating (https://spectrum.ieee.org/tag/diffraction-grating) with special polarization behavior can shine. The metasurface integrated with the Dunn Solar Telescope at the Sunspot Solar Observatory can snap images of the Sun’s magnetic field in a new advantageous way.Noah Rubin How nanopillars (https://spectrum.ieee.org/tag/nanopillars) parse polarization Rubin and his colleagues first proposed the design for this camera in a previous conceptual paper (https://www.science.org/doi/10.1126/science.aax1839). They added a polarization component to a widely-used optics (https://spectrum.ieee.org/tag/optics) theory that describes how light diffracts and informs the way engineers design patterns to manipulate that diffraction. With some mathematical wizardry, the researchers described a periodic surface whose elements capture discrete components of polarized wavefronts. The team then fabricated their surface on a glass substrate. Each unit of its repeating pattern contained 144 rectangular pillars arranged in an area less than five micrometers wide. When placed within a camera lens system, each diffracted component captured a different part of incoming polarized light—recording the entire polarization picture simultaneously. “It’s like a special beamsplitter that has an arbitrary number of channels, and we can control their sensitivity,” says Rubin. He and his colleagues built a high-performance version of their metasurface-enabled imaging device to fit with a 6-millimeter aperture lens. They teamed up with industry partner BAE Space & Mission Systems (https://www.baesystems.com/en-us/who-we-are/space-and-mission-systems) to run it through the vibration and thermal tests required to send equipment to space. Finally, they integrated their device to the light-feed of the Dunn Solar Telescope and recorded snapshots of the sun. Comparing their magnetic field images to ones taken from instruments in low-Earth orbit showed the correct order of magnitude and spatial patterns. With the researchers’ design, 70 percent of incoming light reached the metasurface and could then be received by the detector, and the device achieved near state-of-the-art contrast. That’s exciting for more than just solar astronomy. “These are a good target for any application where we want to condense a lot of polarization elements into one device,” says Rubin. Examples include facial recognition (https://spectrum.ieee.org/tag/facial-recognition) technology for cell phones, entangled photon preparation for quantum experiments, and coronagraphs that block light from distant stars for hunting orbiting planets. NASA (https://spectrum.ieee.org/tag/nasa) intends to launch a solar monitoring mission in the 2030s. The UCSD (https://spectrum.ieee.org/tag/ucsd) team and their industry partners are part of an initial study to examine new approaches for instrument design. That means competing with other teams to develop new technology that NASA may select for the mission. “Ours could be it,” says Rubin. From Your Site Articles • Optical Metasurfaces Shine a Light on Li-Fi, Lidar › (https://spectrum.ieee.org/lifi-lidar-metasurface-applications) • Here's How This Metasurface Lens Could Improve Imaging › (https://spectrum.ieee.org/metasurfaces-metamaterials-image-processing) Related Articles Around the Web • A Small Optical Component Could Change How Telescopes View the Sun › (https://today.ucsd.edu/photo-essays/a-small-metasurface-could-change-how-telescopes-view-the-sun) • Metasurface-enabled astronomical polarimetry | Science Advances › (https://www.science.org/doi/10.1126/sciadv.aee8035) Keep Reading ↓