Two structural shifts in the enterprise AI market were confirmed on August 14 and 15, 2026, within hours of each other. OpenAI CFO Sarah Friar told investors that enterprise revenue has crossed consumer for the first time, with the company’s annualized run rate reaching $40 billion roughly two quarters ahead of its own forecast. Simultaneously, Anthropic disclosed Q2 2026 preliminary revenue of more than $11.5 billion, a 14-fold increase year over year and its first profitable quarter. Together, the two companies are approaching $100 billion in combined annualized revenue. The numbers signal that the enterprise AI market has crossed from early adoption into something that looks more like mature market dynamics.

The OpenAI Numbers

Friar’s August 14 investor meeting was scheduled before a turbulent week at OpenAI, which saw Chief Revenue Officer Denise Dresser leave after eight months and longtime executive Brad Lightcap end an eight-year tenure. Despite the C-suite disruption, the financial picture Friar described was clear: “We entered the year at 60-40, but enterprise has accelerated much faster than expected and those lines have now crossed. The majority of our revenue is now enterprise,” she told investors, according to a person who attended the meeting, as reported by CNBC.

The headline number is $40 billion in annualized recurring revenue. July delivered 20% month-over-month growth overall, with business customers growing 32% in the same period. Bloomberg first reported the $40 billion figure; CNBC confirmed it independently. ChatGPT’s advertising business, launched in February 2026, is approaching a $1 billion annual run rate on its own.

At the time of OpenAI’s March 2026 funding round at a $852 billion post-money valuation, the company stated enterprise was “on track to reach parity with consumer by the end of 2026.” The actual crossover arrived in August. That acceleration matters not just as a headline: it reflects two years of enterprise deployments reaching the scale where seat counts and token consumption compound.

The three products Friar and board chair Bret Taylor cited as primary growth drivers at a July 29 internal all-hands: GPT-5.6, ChatGPT Work (the enterprise agent product launched earlier in 2026), and Codex, OpenAI’s AI coding platform. Friar noted that the newest model is 54% more efficient on agentic coding tasks, which lowers the cost per measurable outcome, the exact framing enterprise buyers are now demanding.

The Buying Shift: From Tokenmaxxing to Cost Per Intelligence

Friar’s investor remarks included a specific diagnosis of how enterprise AI buying behavior has changed. The era of “tokenmaxxing” is over. That term, which circulated widely in enterprise AI discussions after the Linux Foundation tokenomics working group formalized the critique earlier this year, described a buying pattern where companies encouraged broad AI usage across employees without tracking what the spending actually produced.

“Enterprise customers have moved from tokenmaxxing to focusing on cost per unit of intelligence,” Friar said, according to the investor meeting account. The shift has a practical implication: buyers now evaluate AI vendors on the basis of measurable output per dollar, not on feature lists or benchmark scores. This favors models that can demonstrably complete agentic coding tasks, autonomous workflows, or customer-facing interactions at a traceable cost, rather than general-purpose tools priced by volume.

For enterprise AI buyers, this is the frame that matters: the question is no longer “do we use AI broadly?” but “what does each AI-completed task actually cost us, and what does it replace?”

MetricOpenAI (Aug 14 disclosure)
Annualized revenue run rate$40 billion
July revenue growth (MoM)20%
Business customer revenue growth (July MoM)32%
Enterprise vs. consumer splitEnterprise now majority
Original forecast for crossoverEnd of 2026
Actual crossoverAugust 2026 (roughly 2 quarters early)
ChatGPT advertising ARRApproaching $1 billion

Anthropic’s Numbers: First Profitable Quarter

While Friar briefed OpenAI investors on August 14, Anthropic was separately disclosing its own financials to prospective investors. Documents viewed by Bloomberg, as reported by CNBC on August 15, showed preliminary Q2 2026 revenue of more than $11.5 billion. That compares to $787 million in Q2 2025, a 14-fold year-over-year increase. Quarter over quarter, revenue nearly doubled from $4.73 billion in Q1 2026.

The structural milestone in Anthropic’s disclosure is not revenue size but profitability. Q2 2026 marked Anthropic’s first quarter of positive adjusted operating income. The company has been among the highest-spending AI labs relative to revenue, and reaching adjusted profitability while growing at this rate signals that the cost structure has scaled enough to support the revenue base.

Anthropic’s growth trajectory is faster from a smaller base. Earlier this year, Anthropic’s enterprise revenue was reported to be outpacing OpenAI in percentage growth terms, driven by Claude Cowork’s enterprise adoption and the Claude API’s strong uptake among development teams. The Q2 data confirms the trend held through the quarter.

Both companies are in various stages of IPO preparation, which provides additional context for why these numbers are being disclosed to investors now. OpenAI filed its confidential S-1 with the SEC in May 2026, confirmed publicly on June 8, with Goldman Sachs, Morgan Stanley, and JPMorgan as lead underwriters.

The Leadership Layer

The disclosures arrived during a complicated organizational week at OpenAI. Dresser joined as CRO in December 2025 to lead the enterprise push whose results Friar described months later. Her departure after eight months, alongside Lightcap’s exit, raises organizational continuity questions that will appear in OpenAI’s eventual public S-1 disclosure. Friar announced Dali Rajic as Dresser’s replacement at the investor meeting.

President Greg Brockman joined the meeting and told investors that cybersecurity is a critical piece of the business, brushing off concerns about open-source competition. The presence of both Friar and Brockman signals that the investor session was designed to manage the narrative around the departures while leading with the revenue story.

What This Means for Enterprise AI Strategy

Several practical conclusions follow from these disclosures for enterprise leaders evaluating AI strategy.

The market has consolidated around two dominant providers. OpenAI and Anthropic together represent nearly $100 billion in combined annualized revenue in a market that barely existed two years ago. That scale means both companies are investing aggressively in enterprise features, reliability, and support. The gap between these providers and the next tier is widening.

Agentic AI is the growth driver, not chat. Friar’s specific callouts, ChatGPT Work, Codex, and GPT-5.6, are all products oriented toward completing tasks and workflows autonomously, not toward interactive chat. Enterprise buyers are deploying AI to replace units of work, not to assist human workers in the same workflows. The revenue mix at OpenAI now reflects that.

The productivity measurement problem is unsolved. Friar’s “cost per unit of intelligence” framing acknowledges that enterprises are demanding measurement, but the tooling to measure AI output at the workflow level remains immature. Companies that build robust attribution frameworks now, connecting AI agent activity to business outcomes, will be positioned to expand AI budgets with confidence rather than facing board-level scrutiny when the next renewal cycle arrives.

Both companies are approaching liquidity events. For enterprise buyers with multi-year AI contracts, IPO filings from both OpenAI and Anthropic will add new considerations around vendor stability, pricing flexibility, and contractual protections. Procurement teams should begin reviewing AI vendor contract structures with the understanding that both companies will face public-market scrutiny of their enterprise contract terms.

If your organization is working through how to measure AI ROI or evaluate which enterprise AI platform aligns with your operational goals, the Enera team works directly with enterprise GTM and operations leaders on these decisions.

Sources: