Anthropic Soars to $11.5 Billion in Revenue, Forcing Consultancies to Renegotiate AI Contracts
Anthropic's revenue grows 14 times in the quarter, and OpenAI claims to have surpassed consumer revenue with $40 billion in ARR. The Big 4 and global banks are revising token contracts in light of this surge.
The Quarter That Changes the Corporate AI Landscape
Anthropic surpassed $11.5 billion in preliminary revenue in the second quarter of 2026, nearly fifteen times the $787 million from the same period in 2025, according to reporting by CNBC and Fortune published on Friday, August 15. The previous quarter had closed at $4.73 billion. The company also reported positive adjusted operating income, the first reported by a frontier lab at such a scale.
The 143% jump compared to the first quarter has no parallel in the previous corporate software cycle. ServiceNow, five years after its IPO, was still at $3.88 billion quarterly. Salesforce took nearly two decades to reach a similar volume. Anthropic has completed three years as a subscription business as of July.
Skepticism, In Its Own Name
Adjusted operating income is not accounting profit. Ed Zitron, author of the newsletter Where's Your Ed At, argued over the weekend that this figure excludes most of the real compute costs. The 20-year, $9.1 billion agreement that Anthropic struck in July with Riot Platforms for a 191 MW plant is, in effect, long-term capex financed by Wall Street. Combined with multi-year contracts with AWS and the TPU package with Google, the aggregate compute liability for the next five years exceeds tens of billions of dollars per market analysts' estimates.
The distinction is important before the corporate buyer signs a renewal. Anthropic sells to clients with existing IT budgets, which changes the economics compared to startups burning venture capital chasing free users. This is the same argument that Sarah Friar, CFO of OpenAI, used on Thursday, August 14, when telling investors that OpenAI's corporate revenue surpassed consumer revenue for the first time, with ARR reaching $40 billion and monthly growth of 20% in July, according to a CNBC report.
Global Implications for Consultancies and Banks
In the United States, JPMorgan operates the LLM Suite with over 200,000 employees, and Goldman Sachs has deployed the GS AI Assistant firmwide to about 46,000 staff. Both are purchasing tokens under multi-year contracts. The Big 4 (Deloitte, PwC, EY, and KPMG) resell this consumption embedded in process reengineering projects. With Anthropic increasing revenue fourteenfold, this arbitrage margin faces new downward pressure as the provider begins to offer direct volume discounts to clients above a certain ticket.
In Japan, the MUFG, Mizuho, and SMBC consortium gained preferential access to OpenAI models earlier this year, and MUFG announced a parallel investment in Sakana AI for domestic model production. None of the three are currently strategic clients of Anthropic, opening the door for commercial movement in the upcoming quarters. Japan has been charging a premium price from Anthropic for tokens consumed in regulated environments, making the market attractive for expansion.
In Europe, the curve is slower. The full enforcement of the EU AI Act on August 2, with the Commission's power to fine up to 3% of global revenue from general-purpose model providers, has added friction to contract renegotiations at banks such as Deutsche Bank and HSBC. Contracts lacking technical documentation clauses now need to be reopened. The short-term effect is a freeze on new rollouts. The medium-term effect is an advantage for suppliers that have already delivered complete model cards to the AI Office, a test that Anthropic, Google, and OpenAI have passed so far.
The Question That Follows
Anthropic's IPO has firmly entered the agenda. Private investors from the September 2025 round, valued at nearly $183 billion, are pressing for liquidity. Bankers interviewed by financial publications in recent weeks mention a listing window in the second half of 2027 at a conservative floor above $400 billion, a figure that now seems modest if the current curve persists.
What changes for the CIO outside the tech sector is not the price of the token. It is the architectural decision. Hiring Anthropic today means betting on a reasoning layer whose product design changes every semester and whose provider has yet to go public. The question is no longer whether generative AI pays for itself. It is how much critical infrastructure can be outsourced to a single address whose governance has yet to undergo open market scrutiny.