Version 6

SpaceX has secured roughly $26 billion in annual AI compute revenue heading into its June 12, 2026 IPO: Anthropic pays $1.25 billion per month for the Colossus 1 supercluster [2][3], with Bloomberg reporting a total contract value near $45 billion implying approximately a three-year term [4]; Google pays $920 million per month for 110,000 Nvidia GPUs at xAI data centers from October 2026 through June 2029, totaling roughly $30 billion [5][6][7][8][9]. SpaceX self-values at approximately $1.75–2 trillion [13], the $75 billion IPO offering is reported roughly 2x oversubscribed with around $150 billion in investor demand [14], and the listing is planned for June 12 on Nasdaq [12].

The SpaceX IPO is a public market test of whether multi-year AI compute contracts can sustain a roughly $1.75–2 trillion valuation, setting a pricing benchmark for AI infrastructure broadly. The two deals also concentrate a large share of frontier AI compute at a single Musk-controlled supplier, a dependency neither Anthropic nor Google has publicly characterized as a risk in regulatory filings.

SpaceX acquired xAI in February 2026, gaining ownership of the Colossus 1 supercluster in Memphis and immediately positioning it as a third-party AI compute supplier [1]. Anthropic agreed to lease the full cluster for $1.25 billion per month [2][3]; Bloomberg subsequently reported the total contract value at approximately $45 billion [4], implying a roughly three-year term. In early June 2026, Google disclosed through SEC filings that it agreed to pay $920 million per month for 110,000 Nvidia GPUs at xAI data centers, running from October 2026 through June 2029 and totaling approximately $30 billion [5][6][7][8][9]. Together, the two contracts commit roughly $2.17 billion per month to a single Musk-controlled entity [10]. Google's role is notable on two fronts: it is both a SpaceX shareholder and its largest compute customer [11].

Both deals surfaced in the weeks surrounding SpaceX's planned June 12, 2026 Nasdaq IPO [12]. SpaceX self-values at approximately $1.75 trillion [13]; the $75 billion offering is reported to be roughly 2x oversubscribed with around $150 billion in investor demand [14]. SpaceX's own filings frame its AI business as targeting a $23 trillion market, and Morgan Stanley projected SpaceX revenue could reach $3.4 trillion by 2040, with xAI's compute rental business as the primary long-run growth driver [15]. The GPU count for Colossus 1 has not been definitively settled: earlier reports cited 220,000 GPUs [2][16], while an IPO-period account put the figure at approximately 325,000 [17].

The deals have drawn competing assessments. Bullish observers argue SpaceX has entered the AI infrastructure market with commodity-like leverage; the implied ~$8,400 per GPU per month rate in the Google deal reprices the IPO story [18][19]. A separate analysis notes both GPU deals were priced at the high end for their respective GPU types [20], which bulls read as pricing power and skeptics read as rates unlikely to hold at renewal. Critics including ProfG and Pierre Rochard argue the implied valuation does not hold against current revenue multiples, with space and telecom historically driving nearly all of SpaceX's actual revenue [21][22]. One commentator adds that SpaceX did not eliminate xAI's debt problem but appears to have converted expensive xAI/X-backed debt into cheaper SpaceX debt, shifting the obligation rather than retiring it [23].

A structural feature of the arrangement is the circular capital flow: Google is a major investor in Anthropic, Anthropic pays SpaceX roughly $15 billion per year for compute, and Google now also pays SpaceX $920 million per month directly [24][11]. Both Anthropic and SpaceX are separately preparing public offerings — Anthropic filed a confidential S-1 in early June 2026 [25] — requiring each to disclose the financial and competitive risks of their interdependence. Commentators have raised the concern that the two AI mega-IPOs may not carry independent valuations: if the companies' compute relationships make their financial fortunes mutually reinforcing, a correction in one could affect the other [26]. Governance critics argue that multi-year, multi-billion-dollar AI compute dependency on a Musk-controlled entity creates strategic and political exposure that neither company has yet publicly acknowledged [27][28].