The $100 Billion Quarter-Mile: Anthropic's Revenue Pace Doubles Down as Its IPO Slips to November

Three numbers now define Anthropic’s moment: $100 billion, November, and $2 trillion.

According to reporting from the New York Times, confirmed by Axios, Anthropic is now pacing to generate more than $100 billion in annualized revenue this year — a roughly 50% jump from the $65 billion figure the company had reached by July, and more than ten times the roughly $9 billion run rate it closed 2025 with. The same reporting says the company has pushed its initial public offering from October into November 2026, delaying the listing so it can fold third-quarter financials into the investor story.

If the deal prices near its target — a valuation of up to $2 trillion on a raise of as much as $100 billion — it would be the largest IPO in history, and the single biggest test to date of whether the generative AI business model can carry the capital expenditures that frontier-scale AI now demands.

The Numbers Behind the Pace

Anthropic’s revenue curve over the past two years reads like a case study in exponential scaling. Annualized revenue passed $1 billion in late 2025’s opening months, crossed $30 billion by April 2026, hit $47 billion by May, and topped $65 billion by the end of July. The new $100 billion pace implies the curve has not merely held — it has steepened — even as competitors slashed prices and open-weight models closed the capability gap.

The quarterly detail is just as stark. First-quarter 2026 revenue came in at roughly $4.8 billion. In the second quarter it nearly doubled to more than $11.5 billion — about 2.5x sequential growth, and more than 14 times the $787 million booked in the same quarter a year earlier. Computing infrastructure spending rose from $3.4 billion to $5.6 billion over the same period, a 65% jump, while adjusted operating margin improved from negative 13% in Q1 to positive single digits in Q2 (excluding stock-based compensation and other non-cash items).

The customer base is scaling in lockstep. The number of enterprises spending at least $100,000 on Anthropic software over a trailing twelve-month period grew from about 1,500 at the end of 2025 to roughly 6,000 by the end of Q2. More than 100 enterprises each spent over $10 million. Named customers in investor materials include AIG, Thomson Reuters, and Spotify, and the push is global: Japan’s NEC and Fujitsu have each signed strategic partnerships since April, with Fujitsu planning to put Claude in front of about 100,000 employees and NEC rolling it out to roughly 30,000 engineers.

Why November, Not October

The delay is not a wobble, advisers argue — it is arithmetic. A November listing lets Anthropic report a full third quarter that includes its response to OpenAI’s September launch of GPT-6 Astra, the rival model that has begun winning back enterprise customers. Some advisers have backed the later timeline precisely because the extra weeks let the company prove it can hold momentum against its largest competitor, rather than pitching a forecast built on second-quarter air.

The competitive data explains the caution. Corporate expense platform Ramp shows OpenAI’s Astra accounting for about 13% of enterprise AI spending it tracks, against roughly 8% for Anthropic’s Claude Fable. On OpenRouter, the widely watched developer routing platform, weekly customer spending on OpenAI models exceeded spending on Anthropic models for the first time in more than two and a half years — a symbolic inflection in developer mindshare that IPO investors will not miss.

Anthropic still holds a commanding revenue lead — its $65 billion July pace against OpenAI’s $40 billion-plus, and quarterly revenue over $11.5 billion against an undisclosed figure for its rival. And OpenAI CEO Sam Altman confirmed Saturday that his company will not go public in 2026, citing safety concerns, which removes the head-to-head listing race. But OpenAI is in early talks on a private round that could value it at $1.2 trillion, and some Anthropic investors worry that private fundraising at that scale could drain market appetite before the Claude maker lists.

The Fixed-Cost Anvil

Underneath the revenue story sits an infrastructure commitment that reads like a sovereign budget. A single agreement signed with SpaceX in May obligates Anthropic to pay about $1.25 billion per month through May 2029 for computing capacity that includes more than 220,000 Nvidia GPUs. That monthly commitment alone represents roughly two-thirds of the company’s total Q2 computing spend of about $1.9 billion per month. Add the April commitment of more than $100 billion to AWS over ten years, and the company’s fixed compute obligations stretch across multiple years at an unprecedented scale.

This is the trade the IPO must sell: $100 billion of annualized revenue growing at unprecedented speed, anchored by fixed obligations nearly as large. First-quarter net losses exceeded $15 billion — though about $12 billion of that stemmed from accounting adjustments tied to equity fair-value changes rather than cash burn — and the company has told banks it expects 2028 revenue of roughly $190–200 billion, a forecast that asks investors to underwrite years of continued hypergrowth.

Slowdown Sermons, Acceleration Incentives

The strangest tension in the story is the one Anthropic built itself. CEO Dario Amodei published a 3,800-word essay on September 12 calling on the industry to deliberately slow the pace of AI capability improvement — drawing rare public support from Sam Altman and Elon Musk. Days earlier, researcher Jacob Coxon resigned with the warning that founders and builders at AI companies genuinely believe there is a possibility of human extinction within the next several years.

Now the company is weighing a new model release ahead of the IPO to counter OpenAI’s momentum, per three sources — a deliberation complicated by its own public call for restraint. Investor meetings starting within days will press for details on how any slowdown in launches would affect financial forecasts. Anthropic’s advisers and existing investors maintain a development pause would not significantly dent revenue, and some project annual revenue exceeding $110 billion.

What to Watch

Three signals will matter between now and the November listing. First, whether the Q3 print confirms the $100 billion pace is real revenue rather than a run-rate artifact of usage spikes. Second, whether enterprise spend share stabilizes — the Ramp and OpenRouter datapoints are early, but they are directionally consistent. Third, whether Nvidia’s reported talks to anchor the listing with an investment of as much as $10 billion at IPO price hold together, which would put the most important supplier in the AI stack directly inside the deal.

The broader market context is less forgiving than private rounds were. Open-weight models from DeepSeek, Moonshot, and Meta have closed much of the performance gap while undercutting frontier pricing dramatically, and buyers like Ramp’s Eric Glyman now openly describe rotating models by task — cutting AI spend 40% in the process. “The AI market is not a cozy monopoly,” former a16z partner Mike Volpi told the Times. A $2 trillion valuation prices in years of extraordinary growth in exactly the market that is commoditizing fastest.

Anthropic’s IPO was always going to be a referendum on the AI business model. The November slip buys one more quarter of evidence. The $100 billion pace says the evidence is arriving faster than almost anyone modeled.