SpaceX Has Landed Data Center Deals With Anthropic, Google, and Reflection That Could Be Worth Over $76 Billion Through 2029. Is That Enough to Justify Its Valuation?

SpaceX reportedly signed compute leases with Anthropic, Google, and Reflection that could generate more than $76 billion in revenue through 2029, including $1.25 billion per month from Anthropic, $920 million per month from Google, and $150 million per month from Reflection. However, each contract can reportedly be terminated with 90 days' notice, capacity is finite at roughly 1 GW across Colossus and Colossus II, and the article argues the valuation still depends on unproven future data center expansion and orbital AI plans. The news is supportive of SpaceX's growth narrative but is tempered by execution and duration risk.

Analysis

The market is likely underpricing the quality of the customer mix, not just the headline revenue number. A lease book anchored by large, repeat-capex AI buyers creates a financing substrate for SpaceX-like infrastructure buildout, and that matters most for the GPU ecosystem: every incremental MW diverted into a third-party hosted environment is another proof point that demand is still outrunning in-house deployment. The second-order beneficiary is NVDA, because persistent scarcity keeps utilization high and weakens any near-term bear case that enterprise AI spend is normalizing.

The real fragility is duration. These are cancellable contracts, so the current valuation uplift is more like an option on tight compute markets than a durable annuity; if pricing softens into 2027, the implied terminal value falls sharply because the valuation is being anchored off 2029 cash flows that may never exist. The bigger risk to the bull case is not cancellation alone, but that SpaceX’s constrained power envelope forces a capital-intensive expansion cycle before the orbital-compute narrative is investable, turning “asset-light revenue” into “asset-heavy execution risk.”

For GOOGL, the signal is more nuanced: leasing external compute suggests management is still optimizing for speed and access rather than insisting on full vertical integration, which is mildly positive for AI product velocity but marginally negative for long-run capex efficiency. That said, if compute remains tight, hyperscalers and frontier-model builders will keep paying up for capacity, which supports a higher near-term floor for AI infrastructure valuations. The consensus is likely too focused on SpaceX equity optics and not enough on what this says about the elasticity of AI demand across the supply chain.

Key Decisions for Investors

• Maintain/ADD long NVDA on any 3-5% pullback over the next 1-3 weeks; the supply-tightness signal supports sustained pricing power, with upside to estimates if AI capex remains elevated into 2H26.
• Initiate a tactical long GOOGL vs. short QQQ pair for 1-3 months: external compute demand implies AI execution urgency, while Google’s balance sheet can absorb near-term infra intensity; stop if broad AI multiples expand sharply.
• Avoid expressing this through private-market exposure to SpaceX at current implied expectations; treat the revenue stream as a call option on 2027-2029 renewals, not base-case cash flow.
• For higher-risk accounts, buy NVDA Jan-2027 calls funded by a partial call overwrite if you expect compute scarcity to persist through the next capex cycle; reward is convex to any reacceleration in AI infrastructure spend.
• Set a calendar alert for early 2027 termination windows; if renewal language weakens, expect the market to de-rate all ‘infinite demand’ AI infrastructure narratives within days, not months.