OpenAI CEO Sam Altman is thinking about expanding into the final frontier for data centers, and his efforts to follow through on that thought reportedly turned into talks with Stoke Space, a rocket startup headquartered just south of Seattle.

Altman looked into putting together the funding to invest in Stoke Space, with an eye toward either forging a partnership or ending up with a controlling stake in the company, according to an account published by The Wall Street Journal. The discussions reportedly began this summer and picked up in the fall, but are said to be no longer active.

Such a move would open up a new front in Altman’s competition with SpaceX founder Elon Musk, who has talked about scaling up Starlink V3 satellites to serve as data centers for AI applications. “SpaceX will be doing this,” Musk wrote in a post to his X social-media platform.

Jeff Bezos, the founder of Amazon and the Blue Origin space venture, has voiced a similar interest in orbital data centers — as has Google CEO Sundar Pichai. Google is partnering with Planet Labs on a space-based data processing effort known as Project Suncatcher.

The tech world’s appetite for data processing and storage is being driven by the rapidly growing resource requirements of artificial intelligence applications. Altman addressed the subject on Theo Von’s “This Past Weekend” podcast in July.

“I do guess that a lot of the world gets covered in data centers over time,” Altman said. “But I don’t know, because maybe we put them in space. Like, maybe we build a big Dyson sphere on the solar system and say, ‘Hey, it actually makes no sense to put these on Earth.'”

Citing unidentified sources, the Journal said Altman has been exploring the idea of investing in space ventures to follow through on that thought. Kent, Wash.-based Stoke Space, which is working on a fully reusable rocket called Nova, reportedly became a focus of Altman’s interest.

Nova is expected to have its first launch in 2026. Just this week, Celestis announced that Stoke Space would use Nova to send cremated remains and DNA samples into deep space for Celestis’ “Infinite Flight” mission in late 2026.

Much has changed on the AI frontier in recent weeks. OpenAI is facing a strong challenge from Google and its Gemini chatbot — and this week, Altman ordered OpenAI to refocus urgently on upgrading ChatGPT, its flagship AI platform. Such down-to-earth market concerns may have been one of the factors putting Altman’s space aspirations on hold.

A spokesperson for Stoke Space said the company would not comment on the Journal’s report.

There’s another Seattle-area space venture that may well offer the kind of play that Altman is looking for: Redmond, Wash.-based Starcloud is developing its own platform for AI data centers in space. Like Stoke Space, Starcloud went through the startup accelerator program at Y Combinator, which Altman ran for a time before he became OpenAI’s CEO.

Last month, Starcloud had its first test satellite launched into space with an Nvidia data-processing chip on board. The startup is already partnering with a Colorado-based company called Crusoe to offer limited GPU processing capacity in space by early 2027.

Editor’s note: This series profiles six of the Seattle region’s “Uncommon Thinkers”: inventors, scientists, technologists and entrepreneurs transforming industries and driving positive change in the world. They will be recognized Dec. 11 at the GeekWire Gala. Uncommon Thinkers is presented in partnership with Greater Seattle Partners.

BOTHELL, Wash. — Before he became the CEO of Portal Space Systems, Jeff Thornburg worked for two of the world’s most innovative space-minded billionaires. Now he’s working on an idea those billionaires never thought to pursue: building a spacecraft powered by the heat of focused sunlight.

Thornburg and his teammates are aiming to make Bothell-based Portal the first commercial venture to capitalize on solar thermal propulsion, a technology studied decades ago by NASA and the U.S. Air Force. The concept involves sending a propellant through a heat exchanger, where the heat gathered up from sunlight causes it to expand and produce thrust, like steam whistling out of a teakettle.

The technology is more fuel-efficient than traditional chemical propulsion — and faster-acting than solar electric propulsion, which uses solar arrays to turn sunlight into electricity to power an ion drive. Solar thermal propulsion nicely fills a niche between those two methods to move a spacecraft between orbits. But neither NASA nor the Air Force followed up on the concept.

“They didn’t abandon it for technical reasons,” Thornburg said. At the time, it just didn’t make economic or strategic sense to take the concept any further.

What’s changed?

“Lower launch costs, coupled with additive manufacturing, are the major unlocks to bring the tech to life, and make it affordable and in line with commercial development,” Thornburg said.

Thornburg argues that it’s the right time for Portal’s spacecraft to fill a gap in America’s national security posture on the high frontier. “There was no imperative for rapid movement on orbit in the 1990s,” he said. “Only recently have the threats from our adversaries highlighted the weaknesses in current electric propulsion systems, in that they have so little thrust and can’t enable rapid mobility.”

Portal’s vision has attracted interest — and financial support — from investors and potential customers. Since its founding in 2021, the startup has raised more than $20 million in venture capital. In 2024, Portal won a commitment for $45 million in public-private funding from SpaceWERX, the innovation arm of the U.S. Space Force. And next year, Portal is due to demonstrate its hardware for the first time in orbit.

So, how did Thornburg hit upon the idea of turning a decades-old idea into reality?

The path to propulsion

Thornburg, who’s now 52 years old, has focused on making things fly for most of his career. It all started when he was a college student in Missouri in the early 1990s, earning his aerospace engineering degree with an ROTC scholarship from the Air Force. He recalled a conversation he had with an instructor who was an old F-4 fighter pilot.

“With my nearsightedness, I was out of the game from a pilot standpoint,” Thornburg said. “But he said, ‘Thornburg, if you can’t fly the planes, go be as close to them as you can.'”

Thornburg signed up for a program that fast-tracked him into an aircraft maintenance role. He traveled around the world with KC-135 cargo planes, supporting missions that included the NATO-led air campaign against Yugoslavia in 1999. During his time as a flight commander and aircraft maintenance officer at MacDill Air Force Base in Florida, “I had a couple of hundred enlisted people who worked hard to keep me out of trouble,” he said.

The Air Force is where he earned his master’s degree in aerospace engineering. “My adviser had a friend that worked at the Air Force Research Lab,” Thornburg recalled. “He called him and said, ‘The Air Force is about to send this guy to do something with airplanes, but I’m pretty sure he’s going to be disappointed if he can’t come out and work on rocket engines.'”

Sure enough, Thornburg was soon working on rocket propulsion development, including a project to create what’s known as a full-flow staged combustion cycle engine. “We made what people thought was not possible possible with that program,” Thornburg said.

In 2004, Thornburg left the Air Force to work on rocket propulsion systems at Exquadrum, Aerojet and NASA. Then, in 2011, he took a phone call from SpaceX’s billionaire founder, Elon Musk. “We talked for about an hour, hour and a half on the phone — and he said, ‘I’ve got a project I want to talk to you about,'” Thornburg said.

That project led to the development of SpaceX’s methane-fueled Raptor rocket engine, which leveraged the technology that Thornburg helped pioneer at the Air Force. “That was a wild ride, because that felt like about 15 or 20 years of experience in a five-year time period,” he recalled.

After five years at SpaceX, Thornburg needed to wind down. He decided to do some consulting at his home base in Huntsville, Alabama, also known as Rocket City. “About six months in, I’m like, I need a real job again,” he said. “And some friends of mine introduced me to, ultimately, Paul Allen. Paul called me and said, ‘Can you come out to my Seattle office?'”

The Microsoft co-founder and software billionaire enlisted Thornburg to become the head of rocket propulsion development for Stratolaunch, Allen’s space venture. Thornburg led the effort to create a liquid rocket engine known as the PGA — which stood for “Paul G. Allen.”

Unfortunately, Allen passed away in 2018, just one month after the engine was unveiled. Under new ownership, Stratolaunch pivoted to hypersonic testing, and the PGA project fell by the wayside. Once again, Thornburg and his family hunkered down in Huntsville.

Building a business

“I decided to start my first space company after Paul died,” Thornburg said. “I focused on hydrogen propulsion technology and solutions, kind of like what we were working on for Paul.”

That first company, Interstellar Technologies, started working on projects for NASA, Northrop Grumman and a couple of other customers. Then the pandemic hit. “The investors that were about to provide funding disappeared,” Thornburg said. “NASA went home, Northrop Grumman went home. And so I had to find my small team other jobs.”

Just as Thornburg was about to resign himself to riding out the pandemic in Alabama, Amazon’s recruiters called. They asked him to move to Seattle to run engineering and manufacturing for Project Kuiper, the satellite internet project that’s now known as Amazon Leo. “That’s ultimately what got us moved to Seattle,” Thornburg said.

His yearlong stint at Amazon was long enough to establish the process for building Project Kuiper’s two prototypes and the production-grade satellites that came after them. Then he took on engineering management roles at Agility Robotics and Commonwealth Fusion Systems.

That’s when Portal Space Systems took shape.

To be fair, the seeds for Portal were planted back in 2016, just weeks after Thornburg left SpaceX. “Lawrence Livermore Lab had called and said, ‘We’re doing a seminar on the future of propulsion. Would you like to come be a speaker?'” he recalled. “I said, ‘Yes, what do you want me to talk about?’ They said, ‘We want you to tell us what the future of propulsion looks like.’ Oh my gosh, no pressure on that!”

As he did the research for his talk, he came across the idea of putting a nuclear reactor on a spacecraft, and using the concentrated heat from that reactor to blast a propellant through a thruster. The concept, known as nuclear thermal propulsion, seemed like a stretch — but then Thornburg had an uncommon thought.

“Can you concentrate solar energy to heat a thrust chamber and do the same thing?” Thornburg said. “You can. It’s not quite as effective as a nuclear reactor, for obvious reasons, but it’s all the same pieces. … Now I don’t have to wait on a low-cost, low-weight, space-rated nuclear reactor that doesn’t exist yet.”

Thornburg mulled over the idea for years. “I was thinking about Portal, and I was starting the beginnings of Portal in 2021, but I still had to pay the bills,” he said. For a couple of years, he worked during the day at Agility Robotics and Commonwealth Fusion — and spent nights and weekends laying the groundwork for the startup.

“When Portal could really start to stand on its own, as we started to win over the Defense Department, that’s when I made the switch with all of my time focused on what was going on in Portal,” Thornburg said. In April 2024, the startup emerged from stealth and announced it had received more than $3 million in funding from the Defense Department and the Space Force.

The road ahead

Portal’s flagship vehicle is called Supernova. It’s a rapid-transorbital, multi-mission vehicle that should be capable of moving itself and its payloads from one orbit to another — even from low Earth orbit to geostationary Earth orbit, more than 20,000 miles higher up. And it should be able to do that within hours or a day, rather than the weeks or months that are typically required.

The spacecraft itself will be about the size of a restaurant refrigerator. To concentrate sunlight on its heat exchanger and thruster system, Supernova will use sheets of reflective material that can unfold to a width of roughly 55 feet. Ammonia will serve as the propellant. The 3D-printed heat exchanger thruster, dubbed Flare, was successfully tested earlier this year.

Next year’s orbital demonstration will involve putting an instrument package known as Mini-Nova, which is about the size of a tissue box, on a satellite platform that’s due for launch on a SpaceX rideshare mission. The demonstration is meant to validate Supernova’s system design.

In late 2026, Portal plans to send up a free-flying spacecraft called Starburst, which will be equipped with thrusters powered by an electrothermal heating system. Starburst won’t be as powerful as Supernova, but it will provide Portal’s customers with an early option for rapid maneuverability in orbit. If next year’s test goes well, Starburst is expected to start taking on customer missions in 2027.

2027 is also the year when Supernova is scheduled to make its debut. All of the development work for Supernova and Starburst will be taking place at Portal’s 8,000-square-foot lab and 50,000-square-foot manufacturing facility in Bothell.

Throughout Portal’s formative years, Thornburg has worked with fellow members of the “small team” he assembled at Interstellar Technologies. Both of Portal’s other co-founders — chief operating officer Ian Vorbach and engineering vice president Prashaanth Ravindran — crossed paths with Thornburg at Interstellar, and at Stratolaunch before that.

Vorbach, whose background includes startup experience as well as engineering experience, said Portal’s business model has been fine-tuned to make sure it addresses the needs of its target market. He and Thornburg identified the U.S. military’s need for tactical responsiveness in space as the top priority.

“What happens a lot in the space industry is that you have incredibly technical, talented people who have a technology that provides some very unique performance, and then they build it, and it turns out that performance isn’t needed,” Vorbach said. “There’s got to be a reason to bring that innovation to market.”

Vorbach is grateful for Thornburg’s leadership. “We work very long hours, but I think Jeff does a great job of making sure people know that they’re valued,” he said. “I appreciate that, and I think it’s why we, fortunately, are able to hire great talent from the places he’s come from, whether it’s SpaceX or Kuiper.”

Ravindran, who worked at Jeff Bezos’ Blue Origin space venture before taking a founder’s role at Portal, agreed with that assessment. “It’s always amazing to have someone like Jeff out there, because he’s come up the engineering road to realize our pain points as well, and he doesn’t try to hold us to unfair standards,” he said. “That way, we are not set up for failure.”

Stan Shull, a space industry analyst at Bellevue, Wash.-based Alliance Velocity, gives Portal high marks. “In space terms, a highly maneuverable satellite is said to have high delta-V,” he told GeekWire in an email. “Portal, as a company, feels high delta-V too.”

Thornburg’s experience and expertise are big factors behind Portal’s rapid progress, Shull said. “He’s very knowledgeable about national security issues and is a straight shooter about the growing threat environment in orbit,” he said. “It’s no surprise the Space Force is among the many customers interested in what the company is up to.”

What will Portal be up to next? Looking long-term, Thornburg is intrigued by the quantum frontier. “I think there are some very interesting things happening in our understanding of quantum physics that will have propulsion applications, that won’t look like propulsion as we know it right now,” he said. “If we could fold spacetime in clever ways … there’s been plenty of writing about that.”

But when he takes a more realistic look at what could happen in his lifetime, Thornburg can’t stop thinking about nuclear propulsion. “Our Supernova spacecraft will have a version that will leverage a nuclear reactor at some point. That was always the going-in position,” he said.

The way Thornburg sees it, the nuclear option will revolutionize spacecraft — and expand humanity’s reach on the final frontier while we figure out how to fold spacetime.

“Nuclear thermal will get us further into the solar system, and this Earth-moon-Mars becomes our backyard,” he said. “But, you know, for my 12-year-old version of myself, that’s not enough.”

Amazon Leo — the satellite internet service provider formerly known as Project Kuiper — says it has started shipping its top-of-the-line terminals to select customers for testing.

Today’s announcement serves as further evidence that Amazon is closing in on providing space-based, high-speed access to the internet to customers around the world after years of preparation. Amazon Leo is still far behind SpaceX’s Starlink satellite network, but the Seattle-based tech giant has lined up a wide array of partners to help get its network off the ground.

The top tier of Amazon Leo’s global broadband service, known as Leo Ultra, will offer download speeds of up to 1 gigabit per second and upload speeds of up to 400 megabits per second, Amazon said today in a blog post. That’s the first time Amazon has shared details about uplink performance.

During an enterprise preview, some of Amazon’s business customers will begin testing the network using production-grade hardware and software. Amazon said the preview will give its Leo teams “an opportunity to collect more customer feedback and tailor solutions for specific industries ahead of a broader rollout.”

“Amazon Leo represents a massive opportunity for businesses operating in challenging environments,” said Chris Weber, vice president of consumer and enterprise business for Amazon Leo. “From our satellite and network design to our portfolio of high-performance phased array antennas, we’ve designed Amazon Leo to meet the needs of some of the most complex business and government customers out there, and we’re excited to provide them with the tools they need to transform their operations, no matter where they are in the world.”

The 20-by-30-inch antennas for the Leo Ultra terminals are powered by a custom silicon chip that’s been optimized for applications including videoconferencing, real-time monitoring and cloud computing. The service can connect directly to Amazon Web Services as well as other cloud and on-premise networks, allowing customers to move data securely from remote assets to private networks without touching the public internet, Amazon said.

In addition to Leo Ultra, Amazon will offer two lower tiers of service: Leo Nano, which will use a compact 7-inch antenna to provide download speeds of up to 100 Mbps; and Leo Pro, which will use a standard 11-inch antenna supporting download speeds of up to 400 Mbps.

Amazon said it’s shipping Leo Ultra and Leo Pro units to select companies for the preview program. “We’ll expand the program to more customers as we add coverage and capacity to the network,” the company said. Pricing details have not yet been disclosed.

Among the companies listed as customers and partners in today’s announcement are JetBlue, Vanu Inc., Hunt Energy Network, Connected Farms and NBN Co, which operates Australia’s National Broadband Network. Amazon Leo’s other announced partners include Verizon, Vodafone and Vodacom, L3Harris, NTT and SKY Perfect JSAT in Japan, plus DIRECTV Latin America and Sky Brasil.

Photos released today by Amazon show installations of Leo hardware at Hunt Energy facilities, where the network will provide high-speed connectivity for Hunt’s infrastructure assets.

“Hunt Energy Company operates a wide range of energy assets across the globe, and this requires exceptional connectivity to be able to operate, maintain and deliver our products,” said Hunter Hunt, CEO of Hunt Energy Holdings and board chairman of Hunt Energy’s Skyward division. “The combination of Amazon Leo bandwidth capabilities and the secure private link is exactly what we needed.”

JetBlue intends to use Amazon Leo to boost the low-cost airline’s in-flight Wi-Fi service. “Having collaborated with Amazon before, we knew Amazon Leo would share our passion for customer-first innovation,” JetBlue President Marty St. George said. “Choosing Amazon Leo reflects our commitment to staying ahead of what customers want most when traveling, such as fast, reliable performance and flexibility in our free in-flight Wi-Fi.”

Amazon Leo plans to offer high-speed satellite internet service to millions of people around the world, as well as to commercial ventures and government entities. But it still has a long way to go to follow through on that plan.

Over the past year, 153 of Amazon’s production-grade satellites have been launched into low Earth orbit (also known as LEO, an acronym that inspired the newly announced name of the service). Amazon plans to fill out its first-generation constellation with more than 3,000 additional satellites. Under the terms of its license from the Federal Communications Commission, half of those satellites are supposed to be launched by mid-2026. It seems likely that Amazon will seek an extension of that deadline.

Meanwhile, SpaceX is continuing to expand its Starlink constellation and its subscriber base. There are more than 9,000 Starlink satellites in orbit, serving the needs of more than 8 million active customers around the world. Starlink satellites are built at SpaceX’s facility in Redmond, Wash., while Amazon Leo satellites are built nearby at a production facility in Kirkland, Wash.

Just a week after a successful launch of its heavy-lift New Glenn rocket, Jeff Bezos’ Blue Origin space venture says it will make New Glenn even heavier.

The new super-heavy-lift variant of Blue Origin’s most powerful rocket, known as New Glenn 9×4, will feature nine methane-fueled BE-4 engines on the first stage, up from seven; and four hydrogen-fueled BE-3U engines on the second stage, up from two. The 9×4 rocket will also have a bigger fairing, or nose-cone section, measuring 8.7 meters (28.5 feet) wide, as opposed to 7 meters (23 feet) for the fairing currently in use.

Kent, Wash.-based Blue Origin says it’s working to enhance the performance of the rocket engines on both the New Glenn 9×4 and the standard 7×2 model. Other upgrades will include a reusable fairing, a lower-cost tank design and a higher-performing thermal protection system.

The upgrades will be phased into upcoming New Glenn missions starting with the next launch, which is expected to occur early next year. “These enhancements will immediately benefit customers already manifested on New Glenn to fly to destinations including low Earth orbit, the moon and beyond,” the company said today in an online update.

Customers can choose between the variants for missions to low Earth orbit (including satellite launches for the Amazon Leo internet mega-constellation); to the moon and deep space (including next year’s Blue Moon Mark 1 uncrewed lunar landing); and for national security missions such as the proposed Golden Dome missile defense system.

Blue Origin said the 9×4 model will be capable of carrying more than 70 metric tons to low Earth orbit (vs. 45 tons for the 7×2), more than 14 tons to geosynchronous orbit, and more than 20 tons on a trip from Earth to the moon. That would make New Glenn 9×4 more powerful than SpaceX’s Falcon Heavy rocket (64 tons to LEO), but less powerful than SpaceX’s Starship (100 to 150 tons to LEO).

The enhancements appear likely to up the ante in Blue Origin’s competition with SpaceX and United Launch Alliance. New Glenn has been launched only twice, as opposed to hundreds of launches for SpaceX’s Falcon 9 and scores of launches for ULA’s Atlas 5. But last week’s successful launch of twin probes to Mars and the first-ever recovery of an orbital-class New Glenn booster have raised Blue Origin’s profile in the launch industry.

That booster, nicknamed “Never Tell Me the Odds,” flew itself back to a touchdown in the Atlantic Ocean atop Blue Origin’s recovery barge, which was named Jacklyn in honor of Jeff Bezos’ mother. This week the booster was brought back to port and transported to the company’s processing facility at Cape Canaveral in Florida, with Bezos looking on.

Yesterday, we welcomed Jacklyn and our fully reusable New Glenn first stage back to the Space Coast. Thank you, Team Blue. pic.twitter.com/WQyvFqn5Cd

— Blue Origin (@blueorigin) November 19, 2025

Creating a virtual brain may sound like a science-fiction nightmare, but for neuroscientists in Japan and at Seattle’s Allen Institute, it’s a big step toward a long-held dream.

They say their mouse-cortex simulation, run on one of the world’s fastest supercomputers, could eventually open the way to understanding the mechanisms behind maladies such as Alzheimer’s disease and epilepsy — and perhaps unraveling the mysteries of consciousness.

“This shows the door is open,” Allen Institute investigator Anton Arkhipov said today in a news release. “It’s a technical milestone giving us confidence that much larger models are not only possible, but achievable with precision and scale.”

Arkhipov and his colleagues describe the project in a research paper being presented this week in St. Louis during the SC25 conference on high-performance computing. The simulation models the activity of a whole mouse cortex, encompassing nearly 10 million neurons connected by 26 billion synapses.

To create the simulation, researchers fed data from the Allen Cell Types Database and the Allen Connectivity Atlas into Supercomputer Fugaku, a computing cluster developed by Fujitsu and Japan’s RIKEN Center for Computational Science. Fugaku is capable of executing more than 400 quadrillion operations per second, or 400 petaflops.

The massive data set was translated into a 3-D model using the Allen Institute’s Brain Modeling ToolKit. A simulation program called Neulite brought the data to life as virtual neurons that interact with each other like living brain cells.

Scientists ran the program in different scenarios, including an experiment that used the full-scale Fugaku configuration to model the entire mouse cortex.

“In our simulation, each neuron is modeled as a large tree of interacting compartments — hundreds of compartments per neuron,” Arkhipov said in comments emailed to GeekWire. “That is, we are capturing some sub-cellular structures and dynamics within each neuron.”

During the full-scale simulation, it took no more than 32 seconds to simulate one second of real-time activity in a living mouse brain. “This level of performance — 32 times slower than real time — is quite impressive for a system of this size and complexity,” Arkhipov said. “It is not uncommon to see a factor of thousands of times slower for such very detailed simulations (even much smaller than ours).”

The researchers acknowledge that much more work is needed to turn their simulation into a model capable of tracing the progress of a neurological disease. For example, the model doesn’t reflect brain plasticity — that is, the brain’s ability to rewire its own connections.

“If we want to mention something specific besides plasticity, then one aspect that is missing is the effects of neuromodulators, and the other is that we currently do not have a very detailed representation of sensory inputs in our whole-cortex simulations,” Arkhipov said. “For all of these, we need much more data than currently available to make much better models, although some approximations or hypotheses could be implemented and tested now that we have a working whole-cortex simulation.”

Arkhipov said the project’s long-term goal is to simulate an entire brain, not just the cortex. “There’s a distinction between whole-cortex and whole-brain,” he pointed out. “The mouse cortex (and our model of it) contains about 10 million neurons, whereas the whole mouse brain contains about 70 million neurons.”

A human-brain simulation would require an even greater leap. The human cortex alone contains not just 10 million neurons, but 21 billion.

The good news is that a sufficiently powerful supercomputer might be up to the task. “Our work shows that very detailed microscopic-level simulations of larger brains may be possible sooner than previously expected,” Arkhipov said. “The results suggest that a simulation of the whole monkey brain (such as that of a macaque monkey with 6 billion neurons) can fit on the full-scale Fugaku system.”

Arkhipov said it was important to point out that creating a brain model on a supercomputer “does not mean that such a model is complete or accurate.”

“Here we are talking about technical feasibility of simulations, and it looks like such simulations even at the scale of the monkey brain are now within reach,” he said. “But to make such simulations biologically realistic, much more experimental data production and model building work would need to happen.”

Rin Kuriyama and Kaaya Akira of the University of Electro-Communications in Tokyo are the principal authors of the paper presented at SC25, titled “Microscopic-Level Mouse Whole Cortex Simulation Composed of 9 Million Biophysical Neurons and 26 Billion Synapses on the Supercomputer Fugaku.” In addition to Arkhipov, authors from the Allen Institute include Laura Green, Beatriz Herrera and Kael Dai. The study’s other authors are Tadashi Yamazaki and Mari Iura of the University of Electro-Communications; Gilles Gouaillardet and Asako Terasawa of the Research Organization for Information Science and Technology in Hyogo, Japan; Taira Kobayashi of Yamaguchi University; and Jun Igarashi of the RIKEN Center for Computational Science.

Jeff Bezos’ Blue Origin space venture sent twin orbiters on the first leg of their journey to Mars today, marking a successful sequel to January’s first liftoff of the company’s heavy-lift New Glenn launch vehicle.

The trouble-free launch of NASA’s Escapade probes, plus today’s first-ever recovery of a New Glenn booster, bolstered Blue Origin’s status as a worthy competitor for Elon Musk’s SpaceX, which has come to dominate the space industry. SpaceX is the only other company to bring back an orbital-class booster successfully.

Even Musk recognized the achievement: “Congratulations @JeffBezos and the @BlueOrigin team!” he wrote in a posting to X / Twitter, the social-media platform he owns.

New Glenn — which is named after John Glenn, the first American to go into orbit — rose from its launch pad at Cape Canaveral Space Force Station in Florida at 3:55 p.m. ET (12:55 p.m. PT). Today’s liftoff followed attempts earlier this week that had to be scratched, initially due to cloudy weather on Earth, and then due to a solar storm in space.

Even on the day of launch, the countdown had to be held and recycled a couple of times for unspecified reasons. But in the end, liftoff was gloriously nominal.

Minutes after New Glenn rose into the sky, the mission plan called for the rocket’s first-stage booster to fly itself back to a touchdown on a floating platform in the Atlantic that was named Jacklyn after Bezos’ late mother. Blue Origin’s first attempt to recover a New Glenn booster failed in January — but this time, the maneuver was successful.

That achievement was greeted by wild cheers from Blue Origin team members watching the webcast, including Jeff Bezos at Mission Control and a crowd at the company’s headquarters in Kent, Wash. The uncertainty about recovering the booster was reflected in the nickname it was given: “Never Tell Me the Odds.”

“Congratulations, Team Blue — you guys did it!” launch commentator Ariane Cornell, vice president of New Glenn strategy and business operations, said during the webcast. “What an incredible day for Blue Origin, for the space industry.”

Cornell’s co-host for the webcast, Tabitha Lipkin, was similarly enthused. “I think I hurt my hand on the table banging too much,” she said.

Meanwhile, New Glenn’s second stage pressed onward to orbit. A little more than half an hour after launch, the second stage deployed two robotic spacecraft for NASA’s Escapade mission to Mars. (The name for the $78.5 million mission is an acronym for “ESCApe and Plasma Acceleration and Dynamics Explorers.”)

The twin probes will follow a loitering, looping trajectory that includes an Earth flyby a year from now. That slingshot maneuver should provide an extra boost to put the spacecraft into Martian orbit in 2027. Once the probes have settled into synchronized orbits, they’ll fly in formation to map the Red Planet’s magnetic field, upper atmosphere and ionosphere in stereo. The science mission is due to last until 2029.

Scientists say Escapade should help NASA prepare for future crewed missions to Mars.

“Understanding how the ionosphere varies will be a really important part of understanding how to correct the distortions in radio signals that we will need to communicate with each other and to navigate on Mars,” principal investigator Robert Lillis, a space physicist at the University of California at Berkeley, said in a news release. Findings from Escapade could also help scientists work out ways to deal with the radiation risks associated with missions on Mars.

On the space science side of things, Escapade could shed light on the process by which Mars lost much of its atmosphere over the course of billions of years. “To understand how the solar wind drives different kinds of atmospheric escape is a key piece of the puzzle of the climate evolution of Mars,” Lillis said.

NASA put UC-Berkeley in charge of operating the probes, which have been named Blue and Gold in honor of Berkeley’s school colors. Rocket Lab USA built the spacecraft, and Blue Origin won the launch order in 2023, two years before New Glenn ever flew.

Escapade was originally scheduled for liftoff a year ago, but NASA postponed the start of the mission, citing the potential costs of a launch delay that “could be caused by a number of factors” — presumably including a scenario in which Blue Origin’s rocket wasn’t yet ready for liftoff. Additional delays arose as Blue Origin followed up on lessons learned from January’s first New Glenn launch.

In addition to launching the Escapade probes, New Glenn carried demonstration hardware for ViaSat’s HaloNet telemetry relay service. HaloNet was tested as part of a program aimed at switching space communication channels from NASA’s Tracking and Data Relay Satellite system, or TDRS, to commercial satellites.

In a post-launch news release, Blue Origin CEO Dave Limp said “we achieved full mission success today, and I am so proud of the team.”

“It turns out Never Tell Me The Odds had perfect odds — never before in history has a booster this large nailed the landing on the second try,” Limp said. “This is just the beginning as we rapidly scale our flight cadence and continue delivering for our customers.” 

New Glenn is designed to send up to 45 metric tons of payload to low Earth orbit, and smaller payloads to destinations beyond Earth orbit. That makes the rocket more powerful than SpaceX’s workhorse Falcon 9 rocket (23 metric tons to LEO), but less powerful than the Falcon Heavy (64 metric tons) or Starship (100 to 150 metric tons). Starship is still in development; a modified version of that rocket is currently due to carry NASA astronauts on the lunar surface in the 2027-2028 time frame.

Jon Edwards, SpaceX’s vice president of Falcon launch vehicles, joined his boss in congratulating Blue Origin: “Recovering an orbital-class rocket is incredibly hard. Well done!” he wrote on X. “We as Americans should be very proud of what we are accomplishing in space.”

Acting NASA Administrator Sean Duffy added his congratulations.  “This heliophysics mission will help reveal how Mars became a desert planet, and how solar eruptions affect the Martian surface,” Duffy said in a written statement. “Every launch of New Glenn provides data that will be essential when we launch MK-1 through Artemis. All of this information will be critical to protect future NASA explorers and invaluable as we evaluate how to deliver on President Trump’s vision of planting the Stars and Stripes on Mars.” 

Sometime in the next few months, Blue Origin plans to use New Glenn to launch an uncrewed Blue Moon MK-1 lander to the moon’s south polar region. And thanks to today’s successful recovery at sea, there’s a chance that “Never Tell Me the Odds” could be reused as the first-stage booster for that launch.

Bezos and Limp both posted pictures and videos on social media with comments on the day’s achievements. Here are a few highlights:

Good overview of the landing. We nominally target a few hundred feet away from Jacklyn to avoid a severe impact if engines fail to start or start slowly. We’ll incrementally reduce that conservatism over time. We are all excited and grateful for yesterday. Amazing performance by… pic.twitter.com/DCEMsuSyPm

— Jeff Bezos (@JeffBezos) November 14, 2025

Some fun stills! Congrats Team Blue and NASA on an amazing day. Never tell me the odds! pic.twitter.com/jUr31RoOgH

— Dave Limp (@davill) November 13, 2025

Another view pic.twitter.com/MqTIfurI4R

— Dave Limp (@davill) November 14, 2025

This report has been updated with comments from Musk, Bezos, Limp and Duffy.

Bothell, Wash.-based Portal Space Systems has added another spacecraft to its product line: a rapid-maneuverability vehicle called Starburst, which takes advantage of technologies that are being developed for its more powerful Supernova satellite platform.

Starburst-1 is due to star in Portal’s first free-flying space mission with live payloads a year from now, starting with a launch on SpaceX’s Transporter-18 satellite rideshare mission. Portal says the mission will demonstrate rendezvous and proximity operations, rapid retasking and rapid orbital change for national security and commercial applications.

Starburst is designed to bring maneuverability to missions that rely on constellations of small satellites, an approach known as proliferated space architecture. Such an approach is already being used for commercial constellations including SpaceX’s Starlink and Amazon’s Project Kuiper, and the concept is also gaining traction for national security applications.

Portal says Starburst and the larger Supernova platform will share many manufacturing processes and core systems, including the thrusters being developed for Supernova’s reaction control system. Like Supernova, Starburst will use heated ammonia as a propellant.

“Our strategy is to deliver what customers need now and accelerate what they’ll need next,” Portal CEO Jeff Thornburg said today in a news release. “Starburst gives operators a maneuverable bus that supports proliferated architectures in the orbit that matters to them. Supernova brings the trans-orbital reach. Flying Starburst-1 in 2026 lets us field capability quickly and advance the shared systems that raise confidence for Supernova’s 2027 debut.”

Starburst-1 is to be deployed into a sun-synchronous orbit for a one-year primary mission. Portal’s target for on-orbit maneuverability is 1 kilometer per second of total delta-v, which translates to a change in velocity amounting to more than 2,200 mph.

The ESPA-class spacecraft will carry two hosted payloads: a stereo video monitoring system provided by California-based TRL11; and a superconducting magnetic actuator provided by New Zealand-based Zenno Astronautics. Zenno plans to demonstrate the magnet technology that it has developed for satellite positioning and precision interactions between satellites.

In an email, Thornburg told GeekWire that “the Starburst-1 mission is completely funded by Portal to reduce risk and prove capability for our customers ahead of future contracted missions.” Portal plans to offer Starburst for customer missions starting in 2027.

Seattle-based Radical says it has put a full-size prototype for a solar-powered drone through its first flight, marking one low-altitude step in the startup’s campaign to send robo-planes into the stratosphere for long-duration military and commercial missions.

“It’s a 120-foot-wingspan aircraft that only weighs 240 pounds,” Radical CEO James Thomas told GeekWire. “We’re talking about something that has a wingspan just a bit bigger than a Boeing 737, but it only weighs a little bit more than a person. So, it’s a pretty extreme piece of engineering, and we’re really proud of what our team has achieved so far.”

Last month’s flight test was conducted at the Tillamook UAS Test Range in Oregon, which is one of the sites designated by the Federal Aviation Administration for testing uncrewed aerial systems. Thomas declined to delve into the details about the flight’s duration or maximum altitude, other than to say that it was a low-altitude flight.

“We take off from the top of a car, and takeoff speeds are very low, so it flies just over 15 miles an hour on the ground or at low altitudes,” he said. (Thomas later added that the car was a Subaru, a choice he called “a Pacific Northwest move, I guess.”)

The prototype ran on battery power alone, but future flights will make use of solar arrays mounted on the plane’s wings to keep it in the air at altitudes as high as 65,000 feet for months at a time. For last month’s test, engineers added ballast to the prototype to match the weight of the solar panels and batteries required for stratospheric flight. Thomas said he expects high-altitude tests to begin next year.

Thomas and his fellow co-founder, chief technology officer Cyriel Notteboom, are veterans of Prime Air, Amazon’s effort to field a fleet of delivery drones. They left Amazon in mid-2022 to launch Radical and have since raised more than $4.5 million in funding. September’s test of a full-size drone follows up on the 24-hour-plus flight of a 13-pound subscale prototype in 2023.

The company’s manufacturing operation is based in Seattle’s Ballard neighborhood. There are currently six people on the team, plus a new hire, Thomas said. “We’re still lean,” he said. “To make this airplane work, it has to be really efficient, right? Really efficient electronics and aerodynamics. And you also need a really efficient team.”

Thomas said Radical has attracted interest from potential customers, but he shied away from discussing details. “We’re working with groups in the government and also commercially,” he said. “Obviously there are applications at the end of this that span everything from imagery through telecommunications and weather forecasting. There are a lot of people really interested in the technology, and the thing that stops us from serving those customers is not having a product up in the sky. So, that’s what we’re working through.”

Radical’s solar-powered airplane, known as Evenstar, is just one example of a class of aircraft known as high-altitude platform stations, or HAPS. Thomas and his teammates use a different term to refer to Evenstar. They call it a StratoSat, because it’s designed to take on many of the tasks typically assigned to satellites — but without the costs and the hassles associated with launching a spacecraft.

Potential applications include doing surveillance from a vantage point that’s difficult to attack, providing telecommunication links in areas where connectivity is constrained, monitoring weather patterns and conducting atmospheric research.

“We have customers who are really excited about the way that this can improve how we understand Earth’s weather systems and climate,” Thomas said. “That’s an application that we’re really excited to get into.”

Evenstar will carry payloads weighing up to about 33 pounds (15 kilograms). “That was based on analysis about major use cases,” Thomas explained. “That payload is enough to carry high-bandwidth, direct-to-device radio communications, or to carry ultra-high-resolution imaging equipment.”

Radical isn’t the only company working on solar-powered aircraft built for long-duration flights in the stratosphere. Other entrants in the market include AeroVironment, SoftBank, BAE Systems, Swift Engineering, Kea Aerospace, Korea Aerospace Industries and NewSpace Research & Technologies. Airbus’ solar-powered Zephyr set the record for long-duration stratospheric flight in 2022 with a 64-day test mission that ended in a crash.

Among those who tried but failed to field stratospheric solar drones are Alphabet, which closed down Titan Aerospace in 2016; and Facebook, which abandoned Project Aquila in 2018.

Thomas said the outlook for high-flying solar planes has brightened in the past decade.

“The key supporting technologies have matured enormously,” he said. “Commercial battery energy density has doubled in that 10-year time period. Solar cells are 10 times cheaper than they were just 10 years ago. And then you have advances in compute and AI, and all of these things feed into the situation we have now, where it’s actually possible to make the models close — whereas when we run the 10-year-old numbers, we can’t close the models.”

The way Thomas sees it, the concept behind Radical isn’t all that radical anymore.

“Not only do our models say this will work, but we have flight data that agrees with our models, and says this is a technology that can serve its purpose and unlock the potential of persistent infrastructure in the sky,” he said. “I can see why other people are pursuing it. It’s not a new idea. It’s one that people have wanted to crack for a long time, and we’re at this critical inflection point where it’s finally possible.”

How will the companies that have invested tens of billions of dollars in the infrastructure for artificial intelligence fare when the enshittification hits the fan? That question came in for a lot of attention — and snark — when tech pundits Cory Doctorow and Ed Zitron sat down in Seattle to muse about what’s happening in the world of AI.

Both men know a thing or two about enshittification, the process by which tech offerings gradually turn to crap due to the hunger for profits.

Doctorow’s Seattle stopover was part of a publicity tour for his newly published book on the subject, “Enshittification: Why Everything Suddenly Got Worse and What to Do About It.” For Wednesday night’s appearance at the Seattle Public Library, he was paired with Zitron, a public relations specialist, podcaster and writer who surveys the tech scene with a critical eye.

The way they see it, the bursting of the AI investment bubble is a given. And that’s not by any means a contrarian view. Even Microsoft CEO Satya Nadella and Amazon founder Jeff Bezos have acknowledged that the AI tech sector seems likely to go through some retrenchment, while insisting it will be followed by a resurgence that will bring huge benefits to society.

That’s where Doctorow and Zitron part ways with Nadella and Bezos.

“This cannot succeed,” Zitron said. “On top of the fact that everyone’s unprofitable, it’s not actually that popular, either. ChatGPT is very popular because a lot of people love being driven insane. … People will tell you AI’s coming, you must learn AI. The reason it’s not able to do your job is, it’s shit.”

Even if the bubble bursts for good, Doctorow holds out hope that there’d still be some use for the billions of dollars’ worth of AI-boosting chips that have already been purchased for power-gobbling data servers. “If you want to think about a post-AI world, imagine what you would do if GPUs were 10 cents on the dollar, if there were a lot of skilled statisticians looking for work, and if you had a bunch of open-source models that had barely been optimized and had a lot of room at the bottom,” he said.

Zitron is more pessimistic. “The thing that terrifies me about this bubble is, this is not useful infrastructure at all.”

Doctorow defended his vision of a post-AI world by citing a few examples — including AI-augmented search engines that can plow through hours upon hours of podcast audio to find a key quote and whip up a transcript, and AI models that can serve as a backstop when radiologists search through X-rays for signs of cancer.

He also pointed to a nonprofit organization called the Human Rights Data Analysis Group. “They did a project with Innocence Project New Orleans, where they used LLMs [large language models] to identify the linguistic correlates of arrest reports that produced exonerations,” he said. “They used that to analyze a lot more arrest reports than they could otherwise. And they put that on top of a funnel where lawyers and paralegals were able to accelerate their exoneration work.”

When do Doctorow and Zitron think the AI bubble will burst? “No later than Q3 2026,” Zitron said. Doctorow had a less definitive answer.

“I’m a firm believer that the market can remain irrational longer than you can remain solvent,” he said. “I would say that the number of foundation models that will be around after the crash very likely could be zero. I’m not saying that it must be zero. … You can’t kill an open-source model if people like it and contribute.”

So, what’s a techie to do? “Now is the time to unionize,” Doctorow said. He acknowledged that President Donald Trump’s moves against the National Labor Relations Board have made things more difficult for unions, but insisted that all was not lost for the labor movement.

“Trump thinks that we fired the referee, and so that means all the players have to leave the field,” Doctorow said. “He’s wrong. When you fire the referee, it means there are no more rules, right? And there’s a reason that fascists attack unions first. It’s because the opposite of fascism is solidarity.”

When asked what they would advise the next generation of techies to study, neither tech pundit mentioned computer science. “Finance,” Zitron said. “In all seriousness, finance is not as difficult or complex as it sounds, and indeed, the world runs on money.”

Doctorow passed along the advice he said he gave to his daughter. “If you don’t know what you want to do at university, don’t go to university. Go to college and become an electrician,” he said. “There’s so much work for electricians, and we are going to be solarizing for the next 40 years. … It’s like being a plumber, but you don’t have to touch poo.”

He noted that the benefits extended beyond the money. “If you want to learn more, and you like it, you can become an e-eng [electrical engineer],” Doctorow said. “And if you don’t, you can put yourself through college by being an electrician — and learn finance.”

Nine years ago, Clara Sekowski was part of the engineering team at Planetary Resources, a Seattle-area startup that planned to mine precious metals on asteroids. Six years ago, she joined other veterans from Planetary Resources at First Mode, another trailblazing startup that focused on clean energy for industrial applications. Now she’s the CEO of Special Teams, a consulting firm founded with fellow engineers from First Mode.

“Third time’s the charm, right?” she says.

Both of those earlier startups attracted high-profile backers for their ambitious plans, only to face setbacks as reality set in. Special Teams is starting smaller, but it’s gaining traction: The bootstrapped venture and its team of just over 10 engineers recently moved into a 7,400-square-foot office and workshop facility in Seattle’s SoDo neighborhood — and it’s already exceeding its revenue target.

“In our first year, we’re almost at $2 million, which is above and beyond a goal we had set for Year One,” she told GeekWire.

The Special Teams roster includes engineers with experience in aerospace, software development, and even the gaming industry.

“We use systems engineering to bridge the gap between innovation on paper and operational deployment,” Sekowski said. “We design and build prototypes and custom simulations to prove that concepts can work in real-world conditions.”

Special Teams’ to-do list includes helping BHP lay the groundwork for deep-mining automation; working on a confidential nuclear project; and advising Scuderia Cameron Glickenhaus, a high-performance car company, on its plan to create a hydrogen-powered pickup truck.

“I have worked for decades with top engineering and design companies around the world … on highly complex and challenging projects,” Jesse Glickenhaus, Scuderia Cameron Glickenhaus’ CEO, said in an emailed statement. “SpecTeams is by far the most incredible team I have worked with when doing something new and technologically challenging needs to be paired with significant safety and environmental concerns.”

Special Teams’ tangled origins

Sekowski and her teammates have had to deal with their share of technological challenges over the past decade. Redmond, Wash.-based Planetary Resources initially planned to build a fleet of spacecraft to seek out and mine asteroids, and then widened its focus to include Earth observation as well. But funding eventually ran out, and by the end of 2018, Planetary Resources was history.

Some of the engineers who left Planetary Resources started up First Mode. Initially, the engineering consulting firm played a supporting role in a variety of space projects, including NASA’s Artemis moon program and the Perseverance rover mission on Mars.

As time went on, First Mode tightened its focus to concentrate on clean-tech power systems. It established a proving ground for those systems at a former coal mine in Centralia, Wash. Then, in 2023, the Anglo American mining conglomerate took a majority interest in First Mode and accelerated its drive to develop hydrogen-fueled and hybrid powertrains for heavy trucks.

Last year, First Mode opened a 40,000-square-foot factory in the SoDo district — but the company also had to trim back its workforce to adjust to the market demand for clean tech. Months after the factory opened, Anglo American cut off funding for First Mode, setting the stage for a bankruptcy filing and an asset acquisition deal with Cummins, an Indiana-based power solutions company.

Molly Puga, First Mode’s general manager, said the company is continuing to pursue its decarbonization mission as part of the Cummins Power Systems global network.

“Since the acquisition of First Mode, Cummins has been hard at work bringing hybrid solutions for mining to reality,” Puga said in an emailed statement. “We have announced a partnership with Komatsu to scale our hybrid solution, received a grant from TransAlta to support operations at our proving grounds in Centralia, and have shipped product to South America to be commissioned at a mine in the next several months. First Mode continues to employ about 70 employees globally, with the majority of them based in Seattle.”

Special Teams has also been hard at work at its new SoDo facility.

“We are excited to be in the space that we have. It’s a great mix of space where we can build hardware and have a forklift and move around and make noise, and office space where we can get the rest of our work done,” Sekowski said. “SoDo is a great spot for that, because this is really the only area where you can have hybrid spaces. We are going to be here until our hardware doesn’t fit in the building anymore.”

Startups sparked, lessons learned

Special Teams isn’t the only startup founded by First Mode alumni: Last year, former CEO Chris Voorhees and former chief operating officer Rhae Adams established a think tank called Sol Zero Group to support new engineering ventures — including Special Teams.

Civic Forge, which advises businesses on government affairs and public policy strategies, is another venture in Sol Zero’s family of companies. It was founded last year by Conor Duggan, First Mode’s former director of government affairs; and Adam Day, who previously served as First Mode’s senior government affairs manager.

This year, Duggan took on a new role at a clean-tech startup called Vaulted Deep, and Day took charge as Civic Forge’s CEO.

“Civic Forge works alongside Special Teams when a challenge has both policy and technical dimensions,” Day told GeekWire via email. “Together, we pair engineering with policy, which helps clients de-risk designs, win public support and hit milestones faster. Our combined goal is simple: Help organizations move faster across dimensions so they can focus on building.”

Sekowski said that she gets “so much joy” from seeing former colleagues do well, at First Mode and at new ventures. “I think that’s part of the legacy,” she said. “It’s not just Special Teams, but it’s all of us, taking what we learned there and putting it into the next things that we do.”

She’s grateful for the experience she gained at First Mode. “Some of the things I learned at First Mode are important to us here as well,” she said. “Doing work in a broad domain, not just in space, but across energy and other industries, helps us do better work. And so that’s something that we’ll continue.”

Sekowski is even grateful for the setbacks she and her teammates encountered along the way. “I learned so much through the process of being a part of Planetary Resources, and then what was a strong correction to some of the funding issues there with First Mode, and now getting to put some of those lessons together to build this team,” she said. “You don’t get a lot of chances to grow in that way. … We’ve got the battle scars. We know where we want to go, what we don’t want to do, and how to get there.”