By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panelists generally agree that SpaceX's pivot to AI infrastructure is risky and unproven, with significant challenges ahead, including regulatory hurdles, high capital expenditure, and uncertain revenue streams.
Risk: The panelists highlighted several key risks, including Starship delays, regulatory scrutiny, and the high cost of AI infrastructure, with Gemini flagging the cooling constraint as a critical risk.
Opportunity: The single biggest opportunity flagged was the $15 billion annualized Anthropic contract, which validates some demand for SpaceX's AI compute services.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
SpaceX (SPAX.PVT) outwardly focuses on rocket launches and its bleeding-edge Starlink satellite internet service. But CEO Elon Musk and the company are betting heavily on another, non-so-intuitive gamble — AI.
Buried in the S-1 prospectus SpaceX filed on Wednesday is the largest total addressable market ever claimed in a public-company offering: $28.5 trillion, which the filing describes as the "largest actionable" market opportunity "in human history."
Only $370 billion of that comes from launch and other space-enabled services. Another $1.6 trillion comes from Starlink-based connectivity. The remaining $26.5 trillion, nearly 90% of the opportunity, is artificial intelligence.
Essentially, Musk and SpaceX are asking investors to value the company as a vertically integrated AI platform that happens to launch rockets.
"SpaceX is no longer a space company in the traditional sense," Chad Anderson, founder and CEO of Space Capital, told Yahoo Finance. "It's a vertically integrated AI company competing with hyperscalers and aiming to own the full stack."
After SpaceX absorbed Musk's other company, xAI, in February 2026 and restated its results, consolidated 2025 revenue reached $18.7 billion — growth of more than 30% — but the year resulted in a loss from operations of nearly $2.6 billion. The operating loss, Anderson said, is the cost of building two new businesses at once.
"Revenue grew 30%+ to almost $19 billion. Starlink threw off over $4 billion in operating income," Anderson said. "Net loss due to investment in growth — AI and Starship. This is a company with a lot of upside ahead."
That investment led to big capex spends. SpaceX spent $20.7 billion in 2025, with $12.7 billion going into AI — data centers, GPUs, and the COLOSSUS and COLOSSUS II training clusters. In the first quarter of 2026 alone, AI capex hit $7.7 billion.
But the prospectus discloses there is some upside to the AI thesis: AI lab Anthropic (ANTH.PVT), a competitor to xAI's Grok model, has agreed to pay SpaceX $1.25 billion a month through May 2029 for compute capacity.
At full ramp, that is roughly $15 billion of annualized revenue from a single customer, monetizing infrastructure SpaceX has already built. SpaceX says it expects similar contracts in the future.
The overall investment thesis, however, still hinges on what SpaceX is known for — rockets.
The prospectus lists any failure or delay in scaling Starship as its top risk factor, because Starship enables next-generation Starlink satellites, direct-to-cell connectivity, and orbital AI compute.
Four leading AI models discuss this article
"Starship execution risk and concentrated AI capex burn outweigh the single-customer revenue visibility cited in the prospectus."
The article positions SpaceX as a vertically integrated AI platform with a $26.5T addressable market from compute and models, dwarfing its $370B space and $1.6T Starlink segments. Yet 2025 results show $18.7B revenue against a $2.6B operating loss after $20.7B capex, including $12.7B on GPUs and COLOSSUS clusters. The $1.25B monthly Anthropic contract offers near-term monetization, but orbital AI hinges entirely on Starship scaling. Regulatory scrutiny on spectrum, energy use, and export controls for advanced chips could slow deployment far more than the filing acknowledges.
Starship could reach routine orbital refueling by 2027, allowing SpaceX to place power-hungry training clusters beyond terrestrial grid and regulatory constraints, converting today's losses into dominant cost advantages over land-based hyperscalers.
"SpaceX's AI thesis is real but contingent on Starship execution; the $28.5T TAM is aspirational fiction designed to justify current capex burn, not a valuation anchor."
SpaceX is reframing itself as an AI infrastructure play, not a space company—a bold narrative shift. The $28.5T TAM claim is marketing theater; 90% derives from speculative AI upside, not current revenue. Real signal: Anthropic paying $15B annualized for compute validates *some* demand, and Starlink's $4B operating income is genuine. But the $20.7B capex in 2025 ($12.7B on AI) against $18.7B revenue and a $2.6B operating loss reveals the math is still broken. The company is burning cash to build two businesses simultaneously. Starship delays ripple across all three pillars: Starlink scaling, direct-to-cell, and orbital compute. That's existential risk buried in footnote 47.
The Anthropic deal ($15B/year) proves hyperscalers will pay premium rates for SpaceX's latency-advantaged orbital compute—a moat traditional data centers can't match. If Starship succeeds and similar contracts materialize, the capex burn becomes a feature, not a bug.
"SpaceX is leveraging its aerospace monopoly to subsidize a high-risk entry into the commoditized AI infrastructure market, masking fundamental operational losses with speculative TAM projections."
SpaceX is attempting a massive valuation pivot from a capital-intensive aerospace firm to a high-margin AI hyperscaler. While the $15 billion annualized Anthropic contract proves immediate monetization of their 'COLOSSUS' compute clusters, the $26.5 trillion TAM claim is pure marketing fluff designed to justify a tech-multiple valuation. With $20.7 billion in 2025 capex and a $2.6 billion operating loss, SpaceX is effectively burning rocket-launch profits to subsidize a late-entry AI arms race. Investors are essentially buying an unhedged bet on Musk’s ability to out-engineer Microsoft and Google in the data center, while simultaneously needing Starship to maintain the launch monopoly that funds this entire pivot.
If SpaceX successfully achieves orbital compute, they bypass the latency and energy constraints of terrestrial data centers, potentially rendering the current hyperscaler model obsolete.
"The upside rests on an unproven, large-scale AI-infrastructure revenue stream from a single customer alongside Starship/Starlink success, making the thesis highly contingent rather than a given."
SpaceX’s S-1 narrative reframes AI as the core growth engine, suggesting a $28.5T TAM and a potential $15B/year revenue from a single Anthropic compute contract. That paints SpaceX as an AI-infrastructure platform with rockets as a complimentary asset. Yet the piece glosses critical frictions: SpaceX remains private, exposing investors to opaque governance; revenue concentration risk if Anthropic is the main AI customer; AI capex and operating losses persist even as AI infra scales; and real upside depends on Starship/Starlink success amid regulatory, execution, and competitive risks from hyperscalers—the kind of milestones that could derail the implied ramp.
The Anthropic deal could be renegotiated, delayed, or canceled, and Starship/Starlink hurdles could cap cash flow well before AI economics prove durable; in short, the purported AI upside may be far less predictable than the article implies.
"Energy and export regulations create a separate throttle on AI clusters that Starship success alone cannot remove."
Claude flags Starship delays as existential yet misses how energy and export-control rules could block COLOSSUS scaling even with successful launches. The $12.7B GPU outlay already collides with terrestrial grid caps and chip restrictions that Starlink's $4B operating income cannot neutralize. These constraints threaten to extend the $2.6B operating loss far beyond current projections, regardless of orbital refueling timelines.
"Export controls don't isolate SpaceX; orbital compute's real constraint is satellite power and downlink capacity, not regulatory arbitrage."
Grok conflates two separate constraints. Export controls on advanced chips are real, but they apply equally to all hyperscalers—Microsoft, Google, Meta face identical restrictions. The grid-cap argument is stronger: terrestrial data centers already face power constraints in hot markets (Texas, Virginia). Orbital compute doesn't escape this; it just shifts the problem to satellite power generation and downlink bandwidth, which remain unpriced bottlenecks. SpaceX hasn't disclosed either. That's the actual risk.
"Orbital compute faces an unaddressed thermal management bottleneck that may render the payload mass-to-compute ratio economically unviable for orbital deployment."
Claude is right about the power bottleneck, but both Claude and Grok ignore the cooling constraint. Terrestrial hyperscalers use massive liquid cooling loops; orbital compute faces a vacuum environment where heat dissipation relies solely on radiator surface area. SpaceX has not disclosed the mass-to-compute ratio required for orbital thermal management. If the payload mass required for cooling radiators exceeds the lift capacity of Starship, the entire orbital compute thesis collapses regardless of grid constraints or export controls.
"The radiator mass requirement could physically cap orbital compute density, undermining the ROI of SpaceX's AI infra thesis."
To Gemini, I’d push a caution that a constant, unspoken constraint is orbital cooling mass. If radiator mass scales with compute density, Starship payload gets crowded fast, pushing Starship’s available lift to compute-share below required to sustain multi-hundred teraflop scaling. That would cap Orbital Compute ROI and blow up the thesis even if Anthropic revenue arrives on forecast. The article and other panelists haven’t quantified this physical ceiling yet.
The panelists generally agree that SpaceX's pivot to AI infrastructure is risky and unproven, with significant challenges ahead, including regulatory hurdles, high capital expenditure, and uncertain revenue streams.
The single biggest opportunity flagged was the $15 billion annualized Anthropic contract, which validates some demand for SpaceX's AI compute services.
The panelists highlighted several key risks, including Starship delays, regulatory scrutiny, and the high cost of AI infrastructure, with Gemini flagging the cooling constraint as a critical risk.