Why SpaceX Could Become the Most Important AI Company Investors Aren't Calling an AI Company

SpaceX says $26.5 trillion of its $28.5 trillion addressable market is in AI and expects orbital data centers to begin deployment as early as 2028. The article highlights recent compute-rental deals worth $1.2 billion per month with Anthropic and $920 million per month with Alphabet, while Goldman Sachs projects AI revenue could rise from $3.2 billion in 2025 to $322 billion by 2030. The outlook is highly ambitious but still speculative, with heavy capital spending and volatile losses tempering the near-term impact.

Analysis

The market is still pricing SpaceX as a launch/provider story, but the more important read-through is that AI inference and training are becoming a power-and-latency arbitrage game. If orbital compute is even marginally viable, the winners are not just the model developers; it shifts bargaining power toward firms that can secure low-cost, unconstrained energy and vertical integration across launch, satellite, and compute. That creates a second-order threat to terrestrial data-center landlords and utility-adjacent infrastructure names whose valuations assume decades of incremental grid buildout.

The near-term catalyst is not commercial orbital compute in 2028; it is the capex cycle and pre-build of the enabling stack over the next 12-24 months. That should support space/launch supply chain demand, but it also increases execution risk sharply: any schedule slip, regulatory delay, or launch reliability issue will matter more because the market is already discounting a non-linear AI monetization curve. The biggest risk is that investors extrapolate a decade-long optionality story into 2026 earnings, when the actual cash burden is still concentrated upfront.

The contrarian angle is that the AI narrative may be directionally right but economically overstated. If orbital compute works, it likely serves only the most power-constrained, latency-insensitive workloads first, which means the addressable market is much smaller in the early years than headline TAM implies. In that regime, the most attractive public-market expression may be the picks-and-shovels beneficiaries of AI capex reallocation, not the headline platform names, especially if the market starts rewarding firms with visible cash flow rather than moonshot optionality.

Key Decisions for Investors

• Long GOOGL on a 3-6 month horizon: benefit from outsourced AI compute demand and capex-light AI scaling; target a 10-15% rerating if the market revises forward margin risk lower.
• Pair trade: long GOOGL / short a basket of over-owned data-center REITs and utility proxies over the next 6-12 months; thesis is that orbital compute optionality compresses the long-duration scarcity premium in terrestrial power-constrained infrastructure.
• Hold NVDA only via call spreads into 2H26 rather than outright equity: upside persists if AI capex broadens, but crowding and valuation limit reward; use defined-risk structures to capture another leg of accelerator demand without paying full multiple expansion.
• Avoid chasing pre-commercial space optionality in the next 12 months; wait for evidence of launch cadence, regulatory approval, and customer commitments before paying for 2028 economics. Execution risk is too high relative to the timeline.
• If you want exposure to the broader theme, favor picks-and-shovels contractors to AI/space capex over platform narratives; the first monetization wave is likely in enabling hardware, not orbital revenue.