Analyst projects Anthropic $80–100B revenue

Aakash Gupta’s May 20 thread on X turned a disclosed infrastructure contract into a revenue argument. Gupta wrote that Anthropic’s agreement to pay SpaceX $1.25 billion a month for compute implies Anthropic could reach $80 billion to $100 billion in annual revenue later this decade if model inference economics hold. SpaceX’s filing-backed contract matters because it is unusually concrete for an AI market that is often discussed through fundraising decks and benchmark claims. Anthropic said on May 6 that it had agreed to a compute partnership with SpaceX that would “substantially increase” capacity, while CNBC reported the deal gives Anthropic all of the compute capacity at Colossus 1 in Memphis, Tennessee. ### What exactly is the contract Gupta is using? (letsdatascience.com) The disclosed agreement calls for Anthropic to pay $1.25 billion per month through May 2029 for access to Colossus and Colossus 2 capacity, according to reports citing SpaceX’s IPO filing. The same reports said May and June 2026 were priced at a reduced rate during the ramp and that either side can terminate with 90 days’ notice. (anthropic.com) More than $40 billion in potential contract value is the key starting point for Gupta’s math. If the deal runs near its full term, SpaceX would book revenue from a single AI customer on a scale that is rare even by large cloud-infrastructure standards, according to reports summarizing the filing. ### Why does that point to Anthropic revenue rather than just spending? (letsdatascience.com) Gupta’s argument rests on a ratio, not on the contract alone. If Anthropic is willing to commit to compute at that scale, the company would need a much larger revenue base to support it, especially if inference demand rather than one-time training runs is driving the spend; that is Gupta’s stated condition, not a disclosed company forecast. (letsdatascience.com) Anthropic’s recent operating figures give that claim a nearer-term anchor. CNBC reported on May 20 that Anthropic told investors it generated $4.8 billion in first-quarter revenue and could reach $10.9 billion in the June quarter, with an expected operating profit of $559 million. ### Why did Gupta compare it with SpaceX’s own revenue? SpaceX’s projected 2025 revenue of $18.67 billion is the comparison Gupta highlighted because it shows the size of the Anthropic commitment in familiar terms. (letsdatascience.com) Social posts and secondary reports citing the filing said SpaceX also projected a $4.94 billion net loss for 2025, underscoring how large the Anthropic contract is relative to the company’s current financial profile. (cnbc.com) That comparison does not mean Anthropic is paying all of that amount immediately or without contingencies. Reports on the filing said the contract ramps in 2026 and can be ended by either party with 90 days’ notice, which makes the headline value a ceiling tied to duration and utilization. ### What does the deal say about Anthropic’s immediate position? (letsdatascience.com) Anthropic described the SpaceX agreement as part of a broader push to lift usage limits for Claude Code and the Claude API. The company said on May 6 that the new compute capacity, along with other recent compute deals, allowed it to raise customer limits. Memphis is central to that expansion. CNBC reported Colossus 1 houses a large concentration of Nvidia processors, and NBC News said the arrangement gives SpaceX a marquee AI customer while helping Anthropic address capacity constraints tied to demand for products including Claude Code. (letsdatascience.com) ### What should readers watch next? May 2029 is the outer date attached to the contract as reported, but the nearer milestones are Anthropic’s quarterly revenue disclosures and any further details from SpaceX’s IPO filing. (anthropic.com) Investor materials cited by CNBC put Anthropic’s June-quarter revenue target at $10.9 billion, which is the next hard number against which Gupta’s late-decade projection can be judged. (cnbc.com 1) (cnbc.com 2)