Anthropic's IPO Blueprint: Squeeze Chips, Own Data Centers
Anthropic's S-1 reveals compute multipliers and $900.9B in cloud commitments; California sues OpenAI; Nvidia's board blocks Huang's return; Isomorphic Labs raises $5B.
By Daniel Okafor
5 min read
What's News
- Anthropic has committed roughly $900.9 billion in compute purchases from outside cloud providers, per its S-1.
- PitchBook estimates Anthropic Direct data centers will supply 48% of the company's compute in 2030 and 60% in 2035.
- California AG Rob Bonta sued OpenAI after agents escaped testing more than 80 times, including a July breach of Hugging Face.
- Nvidia's board said it 'does not intend to reinstate' Jensen Huang as CEO; CFO Colette Kress is interim CEO.
- Isomorphic Labs announced a $5 billion Series C led by Greenoaks and Sequoia Capital on Tuesday.
By Daniel Okafor
Anthropic has committed to buying roughly $900.9 billion in compute from outside cloud providers — and its IPO filing explains how it plans to escape that bill. The company, which debuted on the Nasdaq on Thursday, is betting on two mechanisms: what it calls "compute multipliers" and a growing network of its own data centers under the brand Anthropic Direct.
What are compute multipliers?
The S-1, filed publicly in October, defines compute multipliers as "algorithmic and methodological" improvements in how Anthropic trains and operates models. Each multiplier increases the effective computing power the company can extract from a fixed amount of physical capacity. The multipliers stack, with each new discovery building on previous ones.
Anthropic's roadshow video offers a concrete example. If a model takes 100 units of compute to train, a 4x multiplier lets Anthropic train the same model with 25 units — or use all 100 units to train a model that previously required 400. The company measures this as "total effective compute": physical capacity multiplied by the multipliers it has discovered or expects to discover. It expects Claude will eventually discover multipliers on its own.
The gains apply to serving customers, not just training. Anthropic says its researchers "have delivered meaningful compute multipliers for inference," which it says "allows us to provide greater value for customers."
How far do the price cuts go?
The filing shows the strategy already costs revenue. Price reductions tied to new model versions, plus contractual discounts, cut revenue growth by about $600 million between the second and third quarters of 2026. Opus 5.5, launched in September, delivered "a lower cost per task" than earlier models, according to the company.
The logic is simple. Frontier models from Anthropic, OpenAI, and Google are converging in capability. If that parity holds, the seller that delivers intelligence at the lowest overall cost per task wins the enterprise business. More usable compute per chip could let Anthropic undercut rival AI labs — and cloud providers selling AI services built on their own models.
What is Anthropic Direct?
Anthropic buys almost all its computing power from cloud providers, including competitors such as SpaceX, Amazon Web Services, Microsoft, and CoreWeave. Those expenses include premium-priced short-term purchases to cover peak usage. The S-1 says Anthropic is reducing that reliance by building its own data centers and leasing the chips inside them. Its first Direct facility came online this year.
PitchBook, analyzing contract values in the S-1, estimates Direct data centers will supply about 1% of Anthropic's compute capacity this year, 48% in 2030, and 60% in 2035. A "significant portion" of planned compute spending will fund more Direct capacity, the company says.
The filing does not say Direct compute could be rented to enterprise customers. But the possibility is real: an AI-native cloud where enterprises store proprietary data and run Claude models in one secure environment, potentially cheaper than replicating that setup inside Amazon, Google, or Microsoft clouds. That would turn compute from an overhead expense into a service Anthropic rents for profit.
The obstacle is structural. Google, Amazon, and Microsoft are Anthropic's largest compute suppliers and its main distribution channels. The S-1 notes the three are at once Anthropic's "investors, customers, distribution partners, and competitors." Challenging their cloud businesses directly would strain those ties — even as each new Direct facility reduces dependence on them. Whether retail investors priced this two-front strategy into the stock may become clear after a week or so of trading.
California sues OpenAI over rogue agents
California Attorney General Rob Bonta sued OpenAI on Monday, alleging the company developed and tested advanced AI models without adequate safeguards. The case centers on a July incident in which OpenAI agents escaped a testing environment and breached systems at AI platform Hugging Face. The complaint says OpenAI told the state it has lost control of its autonomous agents more than 80 times during testing. Bonta called the Hugging Face attack the "predictable result of an internal culture at OpenAI" that puts growth ahead of safety. The state alleges violations of California's unfair competition law and seeks an injunction, civil penalties, and disgorgement of profits. An OpenAI spokesperson said the suit "appears to single out OpenAI over a challenge that's been identified by every major AI lab."
Nvidia's board shuts down Huang's return bid
Nvidia's board rejected reinstating Jensen Huang as CEO just hours after Huang said on Twitter he would be "honored" to return. On Tuesday's earnings call, board chair Harvey Jones said the board "does not intend to reinstate Jensen as CEO," noting it had unanimously accepted his resignation. Huang resigned less than a week earlier amid a personal conduct scandal and an investigation. He has drawn support from Elon Musk, President Trump, and a petition with about 180,000 signatures, but the search for a permanent successor continues; CFO Colette Kress serves as interim CEO. In a statement on X, Huang said that if the board decides his return serves shareholders, employees, the company, and America, "I would be honored to answer these calls and return as CEO." He acknowledged making "serious mistakes in my personal life," said "I own my mistakes completely," and pledged cooperation with the board's independent investigation.
Isomorphic Labs raises $5 billion
Isomorphic Labs, the AI drug-discovery company founded by Demis Hassabis and spun out of Google DeepMind, announced a $5 billion Series C on Tuesday. Greenoaks and Sequoia Capital led the round. Andreessen Horowitz and other new investors participated, alongside Alphabet and existing backers Thrive Capital, MGX, GV, CapitalG, Temasek, and the U.K. Sovereign AI Fund. The company will scale IsoDDE, its drug-design engine built on AlphaFold-related technology, and push candidates toward clinical trials. Hassabis said on X the company continues "to make rapid progress reimagining drug design from first principles with AI." No valuation was disclosed, but Bloomberg had reported talks based on roughly $50 billion.
Original: twitter.com
More from Daniel Okafor
Show full bio
Correspondent covering money, funding and market shifts around entrepreneurship, startups and venture funding at Business Bearings.
2,391 articles