The structural forces behind the AI compute shortage Larry Fink declared publicly — and why orbital compute is emerging as the most credible structural solution in 2026.
Big Tech is projected to spend approximately $655 billion on AI infrastructure in 2026 alone — an unprecedented build-out — and the shortage persists regardless, exactly as Larry Fink described at the Milken Institute Global Conference.
Data centers now account for 7% of US electricity demand, with utilities projecting continued double-digit growth through the end of the decade. Power, not chips, is the sharpest infrastructure bottleneck of the AI era.
SpaceX's $75 billion IPO — the largest in Wall Street history — closed its first trading day up 19.2%, at a $2.1 trillion market cap, with its orbital AI1 satellite design unveiled days earlier as a core part of the investment thesis.
Elon Musk projects SpaceX's orbital AI compute fleet could reach 100 gigawatts of capacity by 2030, scaling an order of magnitude annually from an initial 1 gigawatt target by late 2027 — entirely powered by continuous orbital solar generation.
More than $45 billion was invested in Low Earth Orbit infrastructure in 2025 — up sharply from just under $25 billion in 2024 — as orbital access becomes, in the words of one cybersecurity CEO, "a strategic asset much like ports, cables, or energy grids."
Larry Fink leads the world's largest asset manager, with over $11 trillion under management. When he tells the Milken Institute the US is short power, compute, chips, and memory simultaneously, capital allocators worldwide take note immediately.
| Company | Orbital Compute Play | Scale / Status | Relevance to OrbiCompute.com |
|---|---|---|---|
| SpaceX | AI1 satellite — 120kW avg, 150kW peak per unit | $75B IPO, $2.1T market cap (June 2026) | Primary category validator — orbital compute is now a public-market thesis |
| xAI (merged with SpaceX) | AI compute demand driving orbital buildout | $1.25T combined valuation | Demand-side validation — AI labs need more compute than Earth can supply |
| Axiom Space | Orbital Data Center nodes — first launched Jan 2026 | $350M+ raised | First operational orbital data center — proof of concept achieved |
| Starcloud (fka Lumen Orbit) | First H100 GPU in space, first LLM trained in orbit | $1.1B valuation (fastest YC unicorn) | Direct technical and commercial validator of orbital AI compute |
| Blue Origin | TeraWave — 5,400+ LEO satellites, 6 Tbps | Private (Bezos-funded) | Second major hyperscaler-adjacent orbital infrastructure bet |
| BlackRock | Data center & energy infrastructure investment, $40B Aligned Data Centers deal | $11T AUM | Institutional capital validator — Fink's own firm is deploying into this thesis |
| Sophia Space | Solar-powered compute tiles, NVIDIA Jetson/Blackwell compatible | $10M raised (Feb 2026) | Early-stage entrant — exactly the company that needs OrbiCompute.com |
| Cowboy Space (fka Aetherflux) | Integrated 1-megawatt solar AI satellite units | Early stage | Direct technical peer in the solar-powered orbital compute category |
Microsoft's Satya Nadella has stated publicly that his company has AI chips sitting idle simply because it cannot access enough electricity. Building new terrestrial power generation and grid interconnection takes 5–10 years. AI compute demand is growing on an 18-month cycle. The mismatch is structural, and orbit is the only environment that sidesteps grid interconnection entirely.
If raw compute becomes a tradable futures commodity — as Fink predicts, comparing it directly to oil and grain — the infrastructure layer producing the underlying supply needs a recognised brand. Orbital compute, with its near-infinite solar power ceiling, is positioned as the supply-side answer to a demand curve that institutional capital is already pricing.
SpaceX's AI1 satellite generates 250 watts per square metre of continuous solar power with no day-night cycle limitations at the right orbital altitude, no weather interruption, and no grid congestion. This is a structural power-generation advantage terrestrial solar and wind cannot match, directly addressing the power shortage Fink named first among the four.
As SpaceX has framed it: scaling a terrestrial data center requires years of construction; scaling an orbital compute fleet requires manufacturing more satellites — a factory problem, not a construction problem. SpaceX's 11-million-square-foot Gigasat factory in Bastrop, Texas illustrates this thesis at industrial scale, targeting 1 gigawatt of orbital AI compute by late 2027.
The FCC's 2025–2026 Notice of Proposed Rulemaking explicitly creates new modular license types — Variable Trajectory Space Systems and Multi-Orbit Satellite Systems — designed to accommodate exactly this category of infrastructure, described as "the friendliest regulatory environment in the world" for the space industry.
Nations increasingly view AI compute capacity as a strategic resource comparable to energy or defence infrastructure. Orbital compute offers a path to sovereign AI capacity that bypasses both domestic grid constraints and dependency on foreign chip supply chains — a national security dimension that amplifies institutional and government interest simultaneously.
Larry Fink's remarks at Milken did not happen in isolation — they came alongside BlackRock's own deepening involvement in AI infrastructure, including a roughly $40 billion acquisition of Aligned Data Centers and new partnerships with Microsoft, Nvidia, and MGX deploying tens of billions of dollars into data centers and energy companies.
Brookfield CEO Bruce Flatt, sharing the stage with Fink, said he believes the global economy will be reshaped around data centers and AI over the next decade. This is not retail enthusiasm. This is the largest pools of long-duration institutional capital in the world publicly committing to the AI infrastructure thesis — with power and compute scarcity as the central constraint they are racing to solve.
OrbiCompute.com sits precisely at the technical and narrative centre of where that capital is flowing.
This exact six-week window — Fink's Milken remarks to SpaceX's record IPO — is the highest-attention moment this category will ever have. Acquire the namespace before a competitor recognises the same opportunity.
Acquire OrbiCompute.com