Blue Origin's First Outside Raise Is Oversubscribed, Says CEO Dave Limp

Limp did not confirm the size of the round or how much founder Jeff Bezos has put in, according to CNBC's account of his remarks at the Trump administration's "Hello, America" tech summit in Washington. "I'm so amazed the investor community embraces space this much," Limp said. "They know that space is kind of infinite and the possibilities of putting commercial things in space, and it is proof of the tailwind of raising this round."

That leaves a gap between the company's public position and last week's reporting. The Wall Street Journal reported that Blue Origin had already raised $10 billion at a $140 billion valuation, with Bezos contributing $2 billion in the current raise and $30 billion since he founded the company in 2000. Limp declined to confirm those numbers. The two accounts do not contradict each other, but only one is on the record, and it is the vaguer of the two.

Why the round matters beyond the number

For 26 years Blue Origin ran on Bezos's personal balance sheet. Bringing in outside investors is a structural change for a company that has spent most of its life as a privately funded research project with a rocket attached. Limp called the timing a "combination of things": New Glenn has flown multiple times, two large satellite constellations have been unveiled, and the company needs an external valuation to recruit and keep engineers.

"Attracting great talent ... is difficult in a competitive world, and obviously having an external valuation on the company that shows employees what the value is is very helpful," he said. He also called the raise, in his view, "the best investment Jeff has ever made."

The company's own plans are the other half of the pitch. TeraWave, unveiled in January, is a communications constellation aimed at enterprise, data center and government customers rather than consumers, with more than 5,400 satellites in low Earth orbit. Project Sunrise, disclosed in March, would deploy roughly 56,000 satellites for AI computing in orbit. Limp said many of the heavy data centers would sit in sun-synchronous orbit, where solar power is constant, and would launch on New Glenn. They would carry little communications gear of their own and hand results to TeraWave.

A rocket grounded since May

None of that flies until New Glenn does. The vehicle exploded on Blue Origin's Cape Canaveral pad in late May, causing what CNBC describes as extensive damage. Limp said the plan is still to return to flight before the end of this year, with 1,000 people on the pad around the clock, split between a 500-person day shift and a 500-person night shift.

New Glenn's record so far is mixed. Its first flight reached orbit early last year. A second mission landed the booster last autumn. A third in April landed the booster but deployed its payload, a satellite for AST SpaceMobile, into the wrong orbit, and the satellite was lost. The May explosion followed.

AMD buys World Labs for $8.2 billion

The week's other large AI transaction is a straight acquisition rather than a raise. AMD will acquire World Labs, the world models company founded in 2024 by computer vision scientist Fei-Fei Li with Justin Johnson, Christoph Lassner and Ben Mildenhall, for $8.2 billion, pending regulatory approval, Ars Technica reported on 29 September. The startup had raised $230 million, some of it from AMD itself.

World Labs' first public tool, Marble, generates small 3D spaces using Gaussian splats that can be exported as assets for film and game production. A later model, Atlas, has a different emphasis. AMD has built its own AI models, including the open source Micro-World, but has not fielded a stack that competes with Nvidia's Cosmos line in simulation and robotics.

Under the deal, Li becomes executive vice president and chief scientist at AMD. Johnson and Mildenhall will continue leading the World Labs team as it joins AMD. Ars Technica notes that AMD's announcement leans toward robotics and physical AI, while World Labs' early public successes were in content creation tooling. The acquisition arrives sooner than expected for a two-year-old company.

Nuclear and fusion money moves on public land and state support

Two energy bets tied to AI demand surfaced in the same 24 hours. Valar Atomics is proposing what it calls Project Beehive on more than 9,000 acres of Bureau of Land Management land near Price, Utah, according to documents reviewed exclusively by NPR and published on 30 September. The plan lists data centers, 456 small nuclear reactors, a nuclear fuel production facility and nuclear waste storage. Together the reactors would produce around 9.6 gigawatts of power, against Utah's average of roughly 4 gigawatts statewide.

The timeline calls for nonnuclear construction to start as soon as the end of this year, with the first reactors online sometime in 2028. Valar has also sought about 650 acres of state land for the same project, which the Salt Lake Tribune reported last week. The BLM's Utah office told NPR it had received the application and is reviewing it for completeness. Valar said no construction begins before permits and approvals.

In Japan, Helical Fusion expects to begin preliminary power-on tests at its demonstration reactor in 2027, it said on Tuesday, with construction of the first spiral coil at the National Institute for Fusion Science in Toki, Gifu Prefecture, now largely complete. The company completed a high-temperature superconducting coil test in 2025, which it says was a world first. Power-on tests of the completed integrated facility are targeted around 2030, and a larger 50 MW reactor called HELIX KANATA would follow, with full-scale investment in the 2030s. CEO Takaya Taguchi told reporters the target for a commercial plant is the 2040s, and that cost, safety and construction speed remain open problems. Japan's government has designated fusion as one of 17 strategic sectors and plans 3.1 trillion yen ($19.7 billion) in public and private funding by the 2040 financial year, according to The Asahi Shimbun.

The common thread is that AI's capital needs are no longer confined to software. Blue Origin wants an outside valuation to staff up for orbital compute. AMD is buying simulation and robotics research. Valar wants federal acreage for reactors. Helical Fusion wants a grid-scale answer by the 2040s. Only one of those has a closed round, and even that one is not closed yet.