NextFin News - Anthropic's annual revenue run rate has climbed from about $9 billion at the end of 2025 to more than $65 billion by the end of July 2026, a sevenfold increase in seven months that has lifted the Claude maker's private valuation to $965 billion and set the stage for what could become a $2 trillion initial public offering. The question facing investors is no longer whether Anthropic is growing. It is whether growth this fast can survive a public-company reporting cycle — and whether the multiple it demands can survive the answer.
The Numbers That Define the Stakes
The top line is the story. Anthropic confidentially submitted a draft S-1 registration statement to the Securities and Exchange Commission on June 1, 2026, and has been feeding prospective investors a steady stream of financial updates. In the second quarter of 2026 alone, revenue exceeded $11.5 billion, up more than 14-fold from $787 million in the same period a year earlier and well above the $4.73 billion recorded in the first quarter. The annualized run rate walked up the ladder month by month: $14 billion in February, $19 billion in March, $30 billion in April, $47 billion in mid-May, and $65 billion at the end of July.
Behind the revenue sits a profitability narrative that is equally striking — and equally contested. The company told a small group of shareholders that its adjusted operating income will be positive for a second consecutive quarter, a measure that strips out costs including stock-based compensation. Its gross margins exceed 80 percent before accounting for revenue shared with distribution partners such as Amazon and the cost of training its models.
These figures arrive at a moment of deliberate caution from the company's own leadership. On September 12, 2026, co-founder and CEO Dario Amodei published an essay calling on the AI industry to slow the pace of capability gains.
We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.
The stance puts him at odds with a White House that has rejected such calls and raises an uncomfortable question for growth investors: is Anthropic about to step on the gas or the brake?
The combination of these facts defines what is at risk. Anthropic is racing OpenAI — valued at $852 billion after a $122 billion fundraising round in March, and itself a confidential filer since June — to become the first major frontier AI lab to list. If Anthropic prices near $2 trillion, it would be asking the public market to pay roughly 31 times its July run rate, or more than ten times its projected 2028 revenue of $190 billion to $200 billion. That is not a valuation that tolerates a stumble.
Which Margin Is the Real One
The first decision investors must make is which Anthropic they are buying. The version presented to shareholders shows gross margins above 80 percent and back-to-back quarters of positive adjusted operating income. The version that would appear under a stricter accounting lens looks materially different: one analyst estimates gross margins closer to 44 percent, with the company spending $0.71 on computing power for every dollar of revenue in the first quarter of 2026 and a projected $0.56 in the second.
The gap is not an accident. Adjusted operating income excludes stock-based compensation, and the headline margin figure excludes two of the largest costs in the frontier AI business: the revenue shared with the cloud partners that distribute the models, and the cost of training those models. For a company whose entire product is compute, excluding compute from the definition of profitability is a presentation choice that will face scrutiny once the prospectus becomes public. The SEC's guidance on non-GAAP measures warns that stripping out "normal, recurring, cash operating expenses necessary to operate a registrant's business" can be misleading — the same objection regulators raised against Groupon's pre-IPO metrics in 2011 and WeWork's contribution margin in 2019.
The mechanism is straightforward. Revenue growth in frontier AI is a function of inference demand, and inference demand is a direct claim on compute capacity. Every additional dollar of revenue carries a compute cost with it. If that cost is $0.56 per dollar, the business can scale profitably only if pricing power, model-efficiency gains, or utilization improvements outpace the raw cost of the chips. The 80 percent figure tells investors what the business looks like before the bill for the machinery arrives. The 44 percent figure tells them what it looks like after.
That distinction will do more than set one company's price. The first audited gross margin disclosed by a frontier AI lab will recalibrate the multiples investors assign to OpenAI's eventual offering, to the public AI infrastructure names, and to the entire hyperscale GPU buildout thesis. Anthropic's prospectus is not just a company document; it is the market's first real look at the unit economics of the industry.
Cyclical Pull-Forward or Structural Shift in Software
The second question is whether the revenue surge is a one-time step-change in the adoption curve or the beginning of a durable new revenue base. The evidence points to a structural shift in demand for AI coding and enterprise automation, overlaid with a cyclical wave of pre-deployment purchasing.
The structural leg is real and measurable. Claude Code, the company's agentic coding tool, became generally available in May 2025 and reached $1 billion in run-rate revenue in just six months, according to the company. By February 2026, that product alone was generating more than $2.5 billion in run-rate revenue, with enterprise customers accounting for a growing share. Developers and companies are embedding Claude into workflows, and usage-based contracts convert that embedding into recurring revenue. Enterprise deployment in regulated industries such as financial services and healthcare is the next test — whether Anthropic can turn relationships into large, repeatable deployments that survive budget cycles.
The cyclical leg is the pull-forward. Companies that know a public listing will bring more transparency — and possibly more scrutiny of AI spending — have an incentive to sign contracts now. Government agencies, meanwhile, continued using Anthropic's tools even while the Pentagon dispute played out. When the quarter closes and the IPO is done, some of that demand may prove to have been borrowed from the future rather than created.
The distinction matters because it determines the multiple. A structural shift in how software is written and how enterprises automate work can support a premium valuation for years. A cyclical pull-forward that mean-reverts once the listing is complete would leave the company growing into a multiple it can no longer justify. History offers a warning: the market has rewarded hypergrowth at IPO before, but it has punished companies whose growth decelerated faster than their multiple could absorb.
The Second-Order Risk: A Slowdown That Becomes a Slowdown
The most under-appreciated issue is not the margin accounting. It is the possibility that Anthropic's own safety stance collides with its growth narrative. Amodei's call to pace the frontier is a defensible strategic and ethical position, but it is difficult to reconcile with a valuation that prices uninterrupted hypergrowth. In his September 12 essay, he proposed three concrete steps: opening frontier models to third-party evaluators, establishing common safety standards among frontier companies, and coordinating democratic governments with authoritarian ones on capability limits. Each step, if implemented, would add friction to the release cadence that has powered the company's sevenfold revenue climb.
The transmission channel runs through the product roadmap. If Anthropic deliberately holds back model capability releases, integrates more aggressive safety controls, or declines certain government contracts on principle, its revenue growth rate will decelerate relative to a competitor that does not. That is already the trade-off the market must price. The second-order effect is that a safety-first posture, while it may reduce long-tail liability and improve public sentiment, also narrows the set of customers willing to pay premium prices for the most capable systems — precisely the customers, in defense and intelligence, who have the deepest pockets.
There is a partial offset. A reputation for restraint could become a moat in regulated industries where procurement committees are risk-averse and where a vendor's safety record is a qualification criterion rather than a footnote. Public sentiment is moving in that direction: more than half of Americans now say they are more concerned than excited about the growing use of AI in daily life, up from 37 percent in 2021. But that moat builds slowly, while the IPO multiple is being set now.
Some investors see the positioning differently. D.A. Davidson analyst Gil Luria noted that unless companies genuinely commit to using no more compute and training no more models, the slowdown rhetoric may not change the investment case. Others, including Gartner analyst Arun Chandrasekaran, argue that stricter standards could actually favor Anthropic and OpenAI if smaller competitors cannot afford the rigorous safety, evaluation, and security investments required for frontier-level models. Altimeter Capital CEO Brad Gerstner, whose firm invests in both Anthropic and OpenAI, wrote that "the market knows how to price risk," pointing to SpaceX as evidence of investor appetite for AI leaders.
The Pentagon Dispute: A Risk That Receded, Not Disappeared
The supply-chain risk designation imposed by the Pentagon earlier this year was the clearest illustration of the political overhang hanging over Anthropic's government business. The dispute began when the Defense Department pressed the company to remove safety guardrails from Claude so it could be used for "all lawful purposes," including mass surveillance and fully autonomous weapons. Anthropic objected to those two uses. The Pentagon responded by designating the company a supply-chain risk, and President Donald Trump directed federal agencies to sever ties.
A federal judge in San Francisco questioned the government's evidence at a July 30 hearing, before permanently blocking the designation on August 28.
I don't see additional evidence from the government really justifying what it did.
The White House and the company have gradually mended ties, and government agencies continued using Anthropic's tools throughout the ban.
The legal threat has receded. But the episode revealed the fragility of a business model that depends on government contracts while maintaining safety red lines that the government may not share. Investors buying the IPO are not just underwriting a revenue forecast; they are underwriting a set of ethical constraints and betting that those constraints will not be tested again.
The Counter-Thesis: Growth This Fast Needs No Defense
The strongest case against the skeptics is simple: Anthropic has already answered the sustainability question with its execution. Revenue grew more than 14-fold year over year in the second quarter. The company reached positive adjusted operating income two quarters in a row. Its largest backers — Amazon with roughly a 21 percent stake after committing about $33 billion, and Alphabet with about 15 percent — have both the incentive and the balance sheets to keep the compute pipeline full.
Matt Murphy, a partner at Menlo Ventures and an Anthropic investor, called the growth rate "off the charts."
Don't see why growth would slow or any other reason to wait.
From this vantage point, the margin debate is an accounting distraction: what matters is that demand for Claude is outstripping supply, that enterprise customers are locking in multi-year contracts, and that the company is converting usage into cash faster than competitors.
Lo Toney, managing partner at Plexo Capital and an Anthropic investor, framed the diligence question more precisely.
I would want to understand how the mix shifts between frontier training, post-training and inference as safety controls are integrated.
That is the metric that will separate durable revenue from one-time deployments.
The counter-thesis is formidable because it rests on realized numbers, not projections. But it also rests on an assumption: that the environment that produced a sevenfold revenue increase in seven months will persist unchanged into a public-company reporting cycle, with quarterly earnings calls, activist scrutiny, and a share price that reacts to every deceleration.
What to Watch and What Would Prove This Wrong
The verdict turns on a single variable: the durability of the revenue base once the IPO window closes. In the base case, enterprise and developer adoption of Claude represents a structural shift in software, coding agents become a permanent line item in IT budgets, and margins improve as model-efficiency gains outpace compute costs. In that world, a $2 trillion valuation is a starting point, not a ceiling.
In the downside case, a meaningful share of 2026 demand was pulled forward, compute costs keep gross margins in the 40 to 50 percent range rather than the 80 percent presented to shareholders, and a safety-first posture caps growth just as the company needs to grow into its multiple. At 31 times its July run rate, the stock would not need a collapse in revenue to disappoint — only a return to ordinary growth.
Three signals will tell the story. First, the gross margin disclosed in the public prospectus, and whether compute and revenue-sharing costs are included. Second, the revenue mix between frontier training, post-training, and inference — the metric Toney flagged. Third, the pace of new contract signings in the quarters after the IPO, which will reveal whether 2026 demand was durable or borrowed.
The falsifying signal is specific: if Anthropic's revenue run rate fails to grow by at least 50 percent from the July 2026 level of $65 billion within twelve months of listing — that is, if it does not exceed roughly $100 billion by late 2027 — the sustainability thesis is wrong, and the multiple will compress regardless of absolute revenue growth.
Short term, the stock will trade on sentiment and on the IPO dynamics that have surrounded the largest tech listings of 2026, including SpaceX's record June debut after its $1.25 trillion merger with xAI. Medium term, it will be judged on whether margins hold and whether the enterprise pipeline converts. Long term, the question is whether Anthropic has built a structural position in how the economy writes software — or whether it simply arrived first at a gold rush whose pace it has now vowed to slow.
The market is being asked to pay for a decade of growth in a company that is five years old. Anthropic's numbers justify the excitement. Whether they justify the price depends on a bet that the future will arrive as fast as the past did — and that is a bet the past has never had to make.
Explore more exclusive insights at nextfin.ai.
