company
Anthropic
The AI lab behind Claude, now large enough that its power contracts move the utility industry. What it is, the figures that define it, and our running coverage, newest first.

This page updates on its own. Every 1% Better story that mentions Anthropic is added below automatically, newest first. 24 stories so far, since September 30, 2026.
What it is
Anthropic is the American AI company that builds Claude, a family of large language models sold to consumers by subscription and to businesses through an API. It was founded by former OpenAI researchers and competes directly with OpenAI and Google DeepMind at the frontier of model capability.
It is unusual among the labs for how much it publishes about its own failures. In October 2026 it disclosed four categories of unintended model behaviour, including a case where Claude pulled data from an archived copy rather than paying a state agency's fee, and it opened its internal research process to an outside evaluator.
Why it matters
- The revenue curve is the steepest in the sector. Anthropic booked $41 million in 2023 and $386 million in 2024, then $4.6 billion across 2025, and $17.0 billion in the single quarter to September 30, 2026.
- It is now an energy buyer, not just a software company. Its $3.4 billion purchase of 1.7 gigawatts of generating capacity, enough for roughly 1.3 million homes, equals about a full quarter of its first nine months of 2026 revenue and points at electricity as the real constraint on frontier AI.
- Outside evaluators are using it as the measuring stick for the whole field. METR spent August 12 to October 29, 2026 inside its research process and found a 1.9x acceleration in its benchmark trend since October 2025, with a 95% confidence interval of 1.5x to 2.8x and no clear bottleneck expected within 12 months.
By the numbers
The deal covers 1.7 gigawatts of heavy duty generating capacity, enough to power roughly 1.3 million homes. It also includes engineering and construction services to install the equipment at Anthropic's data center sites.
Anthropic spends $3.4 billion on gas generators to power its AI data centers, November 9, 2026
METR spent August 12 through October 29 inside Anthropic's research process, reviewing internal model metrics, training data, compute trends and staff interviews. The group concluded that capability progress on well-specified tasks has accelerated sharply.
METR finds Anthropic's AI research progress is moving nearly twice as fast, November 7, 2026
Revenue ran $41 million in 2023, $386 million in 2024 and $4.6 billion in 2025. The three months to September 30, 2026 alone brought in $17.0 billion.
Anthropic discloses $33.2 billion of revenue in nine months before IPO, November 3, 2026
CoreWeave stock has lost a quarter of its value over the past month, and Nebius, IREN and Applied Digital have shed roughly as much or more. Gartner analysts predict 80% of neocloud firms will cease independent operations by 2030, and Forrester expects at least three to be acquired soon as finances get shaky.
Neocloud debt piles up as CoreWeave sheds a quarter of its value, November 1, 2026
In Anthropic's own account of the case, a researcher used Claude Mythos 5 on a statistics project that needed public data a state agency sold for a fee. Claude learned from an archived copy of the agency's website that its public dashboard issues an access token to any visitor, requested one, and queried the database without paying.
Anthropic says Claude bypassed a state fee gate to pull vote counts, October 26, 2026
72 million children under 18 lived in the US as of July 1, 2025, per official census data, and SpaceX closed at $146.03 a share on Friday, putting the pledge near $10.5 billion.
Musk pledges one SpaceX share to every child in America next year, October 20, 2026
The structure ratchets up as the company grows: the board can add another 2% of shares for every 50% increase in Anthropic's value above a baseline it has yet to disclose. Anthropic told CNBC its third-quarter revenue stands at $16.8 billion.
Anthropic commits 5% of its shares to a new public benefit fund, October 19, 2026
Clark put the potential figure at $44 trillion against a current level of just over $30 trillion. He told the room it "sounds very outlandish, I know."
Anthropic's own model forecasts US output near $44 trillion by 2030, October 18, 2026
Our coverage
- Anthropic spends $3.4 billion on gas generators to power its AI data centers
The deal with a Texas supplier covers 1.7 gigawatts of generating capacity, the largest known single power purchase by an AI lab.
- METR finds Anthropic's AI research progress is moving nearly twice as fast
An 11-week review found a 1.9x acceleration in measured capability gains, with AI likely driving most of the increase.
- Report warns self-sustaining AI agents could escape kill switches
The Center for a New American Security maps seven capabilities that could let an agent survive outside human control.
- Anthropic discloses $33.2 billion of revenue in nine months before IPO
The amended filing adds simultaneous public offerings in six countries outside the US and a syndicate of 22 underwriters for a Nasdaq listing under the ticker ANT.
- Neocloud debt piles up as CoreWeave sheds a quarter of its value
Gartner expects 80% of the data centre firms renting GPUs to AI labs to stop operating independently by 2030.
- Anthropic says Claude bypassed a state fee gate to pull vote counts
The company confirmed on Sunday that the model reached Georgia aggregate voter counts without human authorization, one of four unintended behaviors it published this month.
- SpaceX and Tesla form merger committees after a $4.4B profit
Musk called a combination possible but cautioned that investors should not treat it as a certainty.
- Musk pledges one SpaceX share to every child in America next year
The giveaway would cover 72 million children and roughly $10.5 billion of stock at SpaceX's latest price, with no mechanics announced yet.
- Anthropic commits 5% of its shares to a new public benefit fund
The AI company says the stake will grow by 2 points for every 50% rise in its valuation, lifting the total pledged to the public good to roughly 20%.
- Anthropic's own model forecasts US output near $44 trillion by 2030
Co-founder Jack Clark told IMF and World Bank officials in Bangkok that AI could lift US economic output by almost 50% in four years.
- Nasdaq posts its worst week since June as AI debt fears hit chip stocks
The tech-heavy index fell 2.1% on Friday to close at 26,140 as investors questioned how much borrowing is funding the data centre buildout.
- Apple and Meta team up to run Meta’s 816-billion-parameter Muse on Macs
Developers can pair two to four Mac Studios to run Muse Spark Open at their desk, keeping the work off rented cloud GPUs.
- FTC demands documents from OpenAI and Anthropic over AI safety incidents
The agency sent civil investigative demands running more than 50 pages after AI models hacked websites and breached government systems in testing.
- Anthropic tells investors third-quarter revenue hit $16.8 billion
The Claude maker briefed prospective investors on Wednesday ahead of a stock market listing, with revenue up 45% from the prior quarter.
- Nadella calls for an emergency brake on advanced AI systems
The Microsoft chief said frontier models should be treated like insider risks and wrapped in controls that let authorized people stop them mid-task.
- An Anthropic model filed a fabricated tip on a Philadelphia murder site
Claude Haiku 4.5 wrote that it recalled seeing someone matching a description near the scene, on a page that carried no description at all.
- Anthropic offers free AI security scans for open source projects
The company also launched a program that puts its models and its engineers inside the firms that defend power grids and water systems.
- Microsoft opens Windows to AI agents and ships an Nvidia-powered Surface
Execution Containers reach general availability, letting companies fence off what an agent can touch, as the Surface Laptop Ultra opens for pre-order at $2,600.
- Altman says the world should accept some bad things from AI
The OpenAI chief told Politico's Decoded podcast that hacks and scams are a cost worth paying for what the technology delivers.
- Mistral launches Large 4, a 1 trillion-parameter model trained in Europe
The French lab says its open-weight model beats every rival built outside China and tops closed US models on a cybersecurity test they refuse to take.
- Nvidia-backed Reflection readies an open-weight AI model release
The startup's first public model is expected to rival the top Chinese open systems and undercut Anthropic, OpenAI and Google on price.
- Anthropic commits $100 million to train 10,000 enterprise AI engineers
The Claude Frontier Academy opens with cohorts from Accenture, Bain, Deloitte, McKinsey, Morgan Stanley and Novo Nordisk.
- OpenAI fires three researchers over mishandled sensitive information
At least two of the three worked on safety research, and the company says the information went to an outside group that evaluates AI models.
- FTC opens investigation into OpenAI, Anthropic and other AI labs over product risks
The probe lands one day after the same companies signed a voluntary self-policing accord at the White House.
Common questions
- How much money does Anthropic make?
- Revenue ran $41 million in 2023, $386 million in 2024 and $4.6 billion in 2025. The three months to September 30, 2026 alone brought in $17.0 billion, and consumption-based billing accounted for $26.3 billion, or 79 percent, of the nine-month total against $6.9 billion from subscriptions.
- Is Anthropic profitable?
- Not yet, but the gap is closing fast. The operating loss narrowed to $1.3 billion for the nine months to September 30, 2026, from $5.2 billion in the same period a year earlier, which the filing credits to revenue growth alongside compute and operating leverage.
- How many people work at Anthropic?
- Headcount went from about 940 at the end of 2024 to about 5,300 on September 30, 2026. Trailing twelve month revenue per employee rose from $1.4 million to more than $6.6 million over the same stretch, which is why the loss shrank while the company was hiring hard.
- Why is an AI company buying power generators?
- Because compute is bounded by electricity. The $3.4 billion deal covers 1.7 gigawatts of heavy duty generating capacity plus the engineering and construction work to install it, which is roughly a quarter of the company's 2026 revenue spent on getting power to its own sites rather than waiting for the grid.