Anthropic told its investors this week that it will more than double quarterly revenue to roughly $10.9 billion in the second quarter of 2026, and post its first operating profit, according to financial projections seen by the Wall Street Journal and confirmed by TechCrunch.

That number is a category-defining event. No major frontier AI lab has cleared the operating-profit line on its model business before. The companies that have tried have either burned billions building distribution (OpenAI) or stayed inside a parent that subsidizes the losses (Google, Microsoft). Anthropic just did the math without either crutch.

The first frontier AI lab to flip the burn-for-growth script changes the conversation. Pricing, talent, and competitive dynamics all reset around the model that pays for itself.

How Anthropic got here

Anthropic's path to profitability looks different from OpenAI's path to scale. OpenAI spent 2024 and 2025 chasing consumer subscriptions and then walking back its checkout ambitions when it handed checkout back to merchants. Anthropic took a different bet. The company focused on enterprise revenue from day one, and the API and Claude for Work product lines now drive the majority of its revenue.

Two structural moves backed the bet. The first is the Forward Deployed Engineer model, originally pioneered by Palantir, where senior engineers embed inside enterprise customers and build glue code. Anthropic closed a $1.5 billion joint venture with Blackstone and Goldman Sachs earlier this year, structured around that model. The economics swap one-time integration work for multi-year revenue commitments.

The second is the enterprise compute spine. We covered Anthropic's managed agents infrastructure deal earlier this year, where Anthropic landed long-term inference partnerships with cloud providers at favorable pricing. Combined with its own training runs on borrowed compute, gross margins on enterprise inference workloads are reportedly the highest in the category.

The talent signal

Andrej Karpathy, one of the most visible AI researchers of the past decade and a founding OpenAI member, announced on Tuesday that he is joining Anthropic to work on frontier large language models. In his own words, he believes the next few years at the frontier are "especially fundamental."

Talent moves of this size are not noise. Karpathy could have gone anywhere. He went to the lab that is about to be profitable. Researchers follow research opportunity, but they also follow runway, equity, and the implicit signal that the company they are joining will exist in five years.

The signal cascades. When the most talked-about researcher in the field picks Anthropic over his former home, every senior candidate at OpenAI re-evaluates their offer. Recruiting costs at OpenAI go up. Compensation at Anthropic, paradoxically, does not need to, because the option value of Anthropic equity just appreciated.

Pricing power changes hands

For two years the assumption inside enterprise AI procurement was that model prices would keep falling. The reverse is now happening. Google's Gemini 3.5 Flash, released this week, costs 5.5 times what its predecessor cost on the same benchmarks, with agent-task workloads running 75 percent more expensive than the supposedly pricier Gemini 3.1 Pro. Anthropic raised Claude API pricing earlier this year. OpenAI has signaled it will do the same with the next model generation.

This is what pricing power looks like in a market that has consolidated. Three serious frontier labs (OpenAI, Anthropic, Google) and a fading fourth (xAI) now define the curve. None of them have a meaningful incentive to compete on price now that the customer base is locked into specific model APIs and the switching cost is non-trivial.

The lab that hits profitability first gets to set the floor. Anthropic just earned that position.

What this means for competitors

OpenAI is responding by going wider, not deeper. The company stood up a $4 billion-plus Deployment Company built around the same Forward Deployed Engineer model, and we covered the $852 billion valuation that pushed it to the front of the AI capital markets conversation. The strategy is clear. Outspend everyone else on distribution and let the model business catch up later.

Google is consolidating. Gemini, AP2, and the agent layer are now bundled with Cloud and Workspace. The pricing of Gemini 3.5 Flash suggests Google is willing to trade developer love for enterprise margin. That works inside the Google customer base. It does not work for capturing share elsewhere.

xAI's recent deal with Anthropic was reported by Platformer as a quiet concession. The reading is that Elon Musk's lab does not see a path to a competitive frontier model on a standalone basis. We covered the larger agent-layer strategy earlier this year, where Anthropic and Google were positioning around different infrastructure plays. The gap has widened.

Microsoft sits in the most interesting position. Its OpenAI partnership remains the largest commercial AI relationship in technology, and yet the underlying model that Microsoft is reselling at premium prices comes from a company that is now spending heavily on its own distribution. The tension was always going to surface. Anthropic's profitability narrative accelerates it.

What this means for customers

Enterprise procurement teams should re-baseline their AI cost assumptions. The era of "models will be cheaper next year" is over for frontier capability. Mid-tier capability will continue to fall in price as open source catches up, but the leading-edge tier now has a pricing floor and a pricing operator.

Negotiation leverage shifts. Three-year commitments at fixed price made sense when the customer expected the price to fall. They look different now. Either get longer terms at today's pricing, or assume the renewal will be higher, not lower.

Model choice becomes more political. Companies standardized on a single frontier model in 2024 and 2025 because the alternatives were either materially worse or materially more expensive. Now that the survivors all have pricing power and the capability gaps are narrower, multi-model strategies become not just possible but probable. Enterprise architecture will route workloads to whichever lab offers the best price for each capability class.

What to watch

Q2 earnings reporting. Anthropic is not public, so the operating-profit claim will surface through investor updates and the financial press rather than a 10-Q. Expect more detail to land in June.

OpenAI's response to the pricing reset. If OpenAI matches Anthropic's pricing increases on the next model release, the market accepts the new floor. If OpenAI cuts price to defend share, the dynamics shift back to a race to the bottom.

The next round of Forward Deployed Engineer joint ventures. Anthropic's $1.5 billion deal with Blackstone and Goldman is the template. Watch for parallel announcements with JPMorgan Chase, BlackRock, or one of the major consulting firms.

Where Karpathy actually works. Researchers go where the interesting problem is. If Karpathy's first public output from Anthropic frames a new training direction, expect the rest of the OpenAI veteran network to follow.

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.

Sources

If Anthropic just earned the right to set the floor, who in the customer market has the leverage to keep the ceiling honest?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.