• Anthropic has agreed to spend up to $120.5 billion on Nvidia (NVDA)-based computing capacity supplied through SpaceXAI arrangements extending into 2029, according to a breaking alert.
  • The figure is $36 billion higher than the previously reported $84.5 billion ceiling for similar agreements.
  • The multiyear spending ceiling is not an immediate payment or direct order to Nvidia; SpaceXAI acts as the infrastructure provider.

A Blockbuster Commitment

Anthropic has reportedly agreed to spend up to $120.5 billion on Nvidia-based computing capacity supplied through SpaceXAI arrangements extending into 2029. A First Squawk alert timestamped 9:10 a.m. EDT on October 26 reported the figure, just minutes before this request. The alert was located, but no accompanying public contract or primary filing confirming its detailed terms was available.

The key distinction is that this is a multiyear spending ceiling, not an immediate payment, acquisition, or direct $120.5 billion order to Nvidia. Earlier reporting, dated September 29, put Anthropic’s SpaceX (SPCX) compute agreements at up to $84.5 billion through 2029. The new headline is $36 billion higher, but the available disclosure does not establish whether that reflects added capacity, expanded options, or revised contractual terms.

Scaling for Claude

The deal builds on an operational partnership announced in May, when Anthropic secured access to all computing capacity at SpaceX’s Colossus 1 data center—more than 300 megawatts and over 220,000 Nvidia GPUs. Alongside its compute expansion, Anthropic doubled certain Claude Code usage limits and raised Claude Opus API limits.

Anthropic, the San Francisco-based frontier-AI developer behind Claude and Claude Code, has seen explosive growth. Reported third-quarter revenue reached $16.8 billion, up approximately 45% from the preceding quarter. First-half 2026 revenue was $16.3 billion, versus $1.1 billion a year earlier. But infrastructure spending is rising alongside it: first-half computing expenditure neared $9 billion, exceeding the $7.3 billion spent during all of 2025. The company reported a first-half net loss of $80.7 billion, though approximately $72 billion came from noncash valuation adjustments involving financing instruments.

SpaceXAI: Supplier and Competitor

SpaceX absorbed xAI in February, bringing the Grok AI business under SpaceX and creating an infrastructure operation that rents computing capacity to outside customers, including Anthropic and Google (GOOGL). This makes SpaceX both an AI competitor and a supplier to competitors. Nvidia supplies the underlying GPU technology, though contract spending paid to SpaceXAI should not be equated with Nvidia revenue dollar-for-dollar.

IDC describes the arrangement as evidence of Nvidia’s continuing importance in frontier AI, even as Anthropic also uses Amazon (AMZN) Trainium chips and Google TPUs. The deal secures a large, concentrated package of power, facilities, networking, operations, and GPUs—giving Anthropic infrastructure to serve growing Claude and enterprise demand. It also establishes Colossus as an external computing platform rather than solely an internal Grok development resource.

The arrangement illustrates two connected industry shifts: AI developers are diversifying beyond their principal cloud partners, and rival AI companies are becoming commercially interdependent. SpaceX can earn infrastructure revenue from Anthropic while competing with Claude through Grok.

Environmental and Regulatory Hurdles

The most concrete regulatory issue is local environmental permitting. On October 19, a federal judge allowed the NAACP’s lawsuit over turbines supplying SpaceXAI’s Colossus 2 operation to continue, rejecting the Justice Department’s attempt to end the case. The procedural decision did not establish that the company violated the law. Colossus 2 is distinct from the Colossus 1 facility identified in Anthropic’s original agreement.

Memphis approved a 135-day halt on new data-center construction on October 20 while examining development impacts, though the measure does not halt existing operations. Anthropic says it has committed to cover U.S. consumer electricity price increases caused by its data centers—a company commitment, not evidence that every local infrastructure or environmental cost has been resolved.

Financial Trade-off

The financial trade-off is substantial. Capacity can enable revenue growth and better service availability, but Anthropic’s rising compute spending and extensive future obligations increase the importance of utilization, pricing, and access to capital. Reported noncancelable commitments elsewhere should not automatically be assumed to have the same cancellation terms as this latest SpaceXAI arrangement.

Anthropic’s recent financial disclosures remain largely based on reporting about its confidential IPO documents, rather than a public prospectus. Reporting describes a potential valuation of up to $2 trillion; that remains a prospective valuation, not an established public-market value. Dario Amodei remains identified as CEO.

What’s Next

The immediate question is what the updated $120.5 billion ceiling actually buys. A primary disclosure is needed to establish capacity delivery dates, minimum purchase obligations, cancellation rights, and whether the change expands the earlier agreements. Until then, it is premature to translate the headline into annual spending, additional GPU counts, or guaranteed supplier revenue.

Anthropic and SpaceX have also expressed interest in multiple gigawatts of orbital AI computing. That remains an exploratory ambition, not a confirmed deployment or a demonstrated component of the newly reported $120.5 billion ceiling.

Correction: October 26, 2026: An earlier version of this article misstated the timing of the First Squawk alert. It was timestamped 9:10 a.m. EDT, not 9:10 p.m.