Anthropic’s IPO Pitch Faces Two Tests: Compute Growth and Enterprise Data Control

Anthropic is reportedly considering an IPO large enough to match or exceed the reported $75 billion SpaceX benchmark. The prospective deal would ask public investors to fund an AI business whose growth depends on computing capacity, while enterprise customers press a narrower question: where data generated through its models is kept.

The transaction’s central numbers are not set. Anthropic confidentially filed to go public in June and has held preliminary test-the-water meetings with bankers and investors in San Francisco, according to people familiar with the sessions. Those people projected a float at about $2 trillion, while the reported final valuation and offering size remain undetermined. The MK account says a public S-1 submission could come as early as the end of August.

The capital case depends on physical expansion

CNBC put Anthropic’s private-market value close to $1 trillion and said it had passed a $65 billion annual revenue run rate. In the investor meetings, CFO Krishna Rao has faced questions about competition, margin pressure from open-source models and the effect of a data-center construction slowdown. The questions frame a contrast at the center of the offering: rapid reported revenue growth does not eliminate dependence on further infrastructure buildout.

That infrastructure risk may become a formal investor disclosure, with people familiar with the planned prospectus telling CNBC that negative sentiment toward AI and data centers is expected to appear as a risk factor. A Gallup survey published in May found that seven in 10 Americans opposed AI data-center construction in their area, while close to half were strongly opposed; Pennsylvania Governor Josh Shapiro signed an executive order imposing stricter development standards in the state on August 18.

Close to $1 trillionReported private-market value

CNBC described Anthropic as valued at close to $1 trillion in private markets.

About $2 trillionProjected IPO valuation

People cited by CNBC projected this valuation; Anthropic has not publicly confirmed it.

Enterprise buyers are negotiating a different constraint

Anthropic is reportedly developing a system that would allow enterprise customers to store AI-model-generated data in their own cloud infrastructure, preserving a 30-day retention period while giving customers control over the storage location. Anthropic said in June that retained data from certain top-level and future state-of-the-art models would be used for cyberattack detection, not model training, according to the MK account. The reported design would make data custody more flexible while keeping the retention rule intact.

What remains unsettled for enterprise customers

• The storage system is reported to be under development, rather than an announced generally available product.
• Anthropic has reportedly discussed related data-storage measures with more than 100 customers in regulated industries.
• The reported approach changes storage location; it does not indicate that the 30-day retention period will be removed.

The competitive backdrop gives that design commercial weight. The MK account cited Ramp payment-data analysis showing Anthropic accounted for 44% of AI-service spending among 70,000 U.S. companies in July, compared with 40% for OpenAI. It is not a complete measure of enterprise AI adoption, but it depicts a market in which providers remain close enough for procurement requirements to matter.

One filing could clarify two separate bets

The prospective IPO and the enterprise storage work answer different demands. One is about financing capacity and pricing exposure to construction delays, public opposition and margin pressure. The other addresses a customer requirement that capital alone cannot solve. A public S-1, if filed, could make the first set of risks easier to assess; it would not by itself establish when the reported storage system will be available or which customers will adopt it.