Anthropic Hits $65B Run Rate Ahead of Its Fall IPO

Anthropic’s revenue keeps outrunning the numbers it posted only weeks earlier. According to reporting from CNBC, Bloomberg, Fortune, and Axios published August 17–18, the maker of Claude told investors over the weekend that its annualized revenue run rate climbed to roughly $65 billion by the end of July. That is up from about $47 billion in May and represents a roughly sevenfold increase from a year earlier. The company shared the figure privately with backers, several of whom relayed it to reporters, and it lands as Anthropic prepares what could be the largest initial public offering in history.

The disclosure is the freshest data point in a growth story that has been redrawn almost month by month. In August, the blog covered Anthropic’s preliminary second-quarter revenue of more than $11.5 billion — a roughly 14-fold jump from the same quarter a year earlier. The $65 billion run-rate figure is the annualized reading on top of that quarter, and it now sits above the run rate most analysts attribute to OpenAI.

From $47 billion to $65 billion in two months

The headline is the slope of the curve. A run rate — a company’s most recent revenue pace projected across a full year — is a snapshot, not audited annual revenue, but the direction is unambiguous. Anthropic described roughly $47 billion annualized as recently as May. By the end of July, backers were told, that figure had reached about $65 billion, an increase of nearly 40 percent in two months. A year earlier, the company’s annualized pace was measured in single-digit billions.

The revenue is overwhelmingly enterprise-driven. Anthropic sells Claude through its API to businesses building software on top of the models, alongside subscription tiers for teams and individual users. Coding has been a particular engine: developer tools and agentic coding products that call Claude at high volume translate directly into token consumption, and token consumption is the meter. The company has also said it reached positive adjusted operating income for a recent period — a milestone it claims to have hit roughly two years ahead of its own projections, and one no other frontier lab has confirmed.

Investors expect the pace to hold. Half a dozen of Anthropic’s backers have told reporters they anticipate the company will finish 2026 with revenue somewhere between $100 billion and $120 billion — more than double the May run rate, and a figure that would have sounded absurd for an AI lab even eighteen months ago.

The IPO that would top SpaceX

The run-rate update matters most because of what comes next. Anthropic confidentially filed for an IPO with the SEC on June 1, 2026, and its bankers — reported to include Morgan Stanley, Goldman Sachs, and JPMorgan Chase — are preparing for a potential debut as soon as this fall, possibly October. Investors are reportedly targeting a valuation around $2 trillion.

If it prices there, Anthropic’s offering would be the largest IPO on record, eclipsing SpaceX, which went public in June at a $1.77 trillion valuation. It would also put Anthropic on the public markets before OpenAI, the rival it trailed for most of the generative-AI era. That reversal is the subplot Wall Street keeps returning to: a company that a year ago was the smaller of the two frontier labs is now first in line for the public markets and, on the run-rate math investors are circulating, first in revenue as well.

The $2 trillion target would also mark a steep step up from Anthropic’s reported $1.2 trillion valuation on the secondary market earlier in the year, itself a jump from the roughly $965 billion attached to its latest private round.

How it stacks up against OpenAI

The comparison to OpenAI is where the $65 billion figure does its heaviest work. Third-party trackers put OpenAI’s annualized revenue in the neighborhood of $40 billion as of mid-2026, up from roughly $20 billion at the end of 2025. On those estimates, Anthropic’s July run rate now sits comfortably ahead — a striking outcome given OpenAI’s larger consumer footprint through ChatGPT.

The gap, to the extent it holds, reflects mix. Anthropic’s revenue skews toward enterprise API and coding workloads that scale with usage, while a large share of OpenAI’s comes from consumer subscriptions. Enterprise spend tends to be stickier and higher-margin, which is part of why investors are willing to underwrite a trillion-dollar-plus valuation on a company that, by its own account, only recently turned an operating profit.

Caveats belong here. Run-rate figures are unaudited, self-reported, and annualize a single recent period — they flatter fast-growing businesses and can reverse quickly if a few large customers pull back. OpenAI and Anthropic also count revenue differently, so head-to-head “who’s bigger” claims should be read as directional, not precise.

The narrower point is that the two leaders have swapped growth profiles. For most of the ChatGPT era, OpenAI set the pace and Anthropic followed; the momentum in these figures runs the other way. Whether that reflects a durable structural edge — enterprise-heavy, coding-led demand — or simply the phase of the cycle each company is in is exactly what a public listing will test. Private run rates invite the flattering interpretation. Quarterly filings do not.

What the prospectus still has to show

None of this answers the questions a public listing will force into the open. When Anthropic’s full prospectus lands, these are the lines that matter:

• Gross margins. Revenue is only half the story. Serving Claude at scale means paying for inference on expensive GPUs and scarce high-bandwidth memory. How much it costs to earn each dollar determines whether the growth is a business or a subsidy.
• Customer concentration. How much of the $65 billion run rate rides on a handful of enterprise accounts or a single cloud partner? Concentration is a risk the run rate hides.
• Compute commitments. Anthropic has signed multi-year, multi-billion-dollar supply and data-center agreements. Those show up as long-term obligations that constrain flexibility if growth slows.
• Durability. A run rate is only as good as its retention. The prospectus will show whether customers who signed on last year are still spending — and spending more.

What it means

The $65 billion figure is less a new milestone than confirmation that Anthropic’s growth has not yet found its ceiling. In roughly a year, the company has gone from a challenger measured in hundreds of millions of quarterly revenue to a business investors are prepared to value near $2 trillion, with a run rate that now tops its larger rival’s.

The winners, if the trajectory holds, are Anthropic’s employees and early backers — including strategic investors and cloud and hardware partners whose balance sheets are increasingly tied to the company’s — and the bankers underwriting a record listing. The risk is that a public debut converts private optimism into quarterly accountability. Run rates can be annualized upward in a boom; they can also compress fast when enterprise budgets tighten, when open-weight models close the capability gap and pressure pricing, or when the memory and compute costs underpinning every token move the wrong way.

What to watch next: the timing and pricing of the IPO itself; whether the year-end revenue lands inside the $100–120 billion range investors expect; and, once the prospectus is public, the margin and concentration figures that a run rate can’t capture. Anthropic has made its case in numbers. The public markets will decide whether the numbers hold.