San Francisco — Anthropic has done what few companies manage in a lifetime. In months, not decades, it scaled from a $9 billion annualized revenue run rate at the end of 2025 to more than $100 billion today. Business customers can’t get enough of its Claude models. Coding tools and workplace assistants drive much of the surge. Yet the very speed that dazzles investors also raises hard questions.
Can this growth hold? Will profits survive the enormous compute bills? And does the company’s own caution about advanced AI create a ceiling its backers refuse to see? Those doubts hang over what could become the largest initial public offering in history.
The latest figures come from people familiar with the company’s finances. Yahoo Finance reported the annualized run rate now tops $100 billion, up from $65 billion in July. That marks a more than 50% jump in roughly two months. Quarterly revenue nearly doubled to about $11.5 billion in the second quarter of 2026. It was 14 times higher than the same period a year earlier.
The Acceleration That Changed Everything
Growth like this has no real precedent. Anthropic started 2026 with a $9 billion run rate. By May it crossed $47 billion. Enterprise adoption of Claude Code and the workplace-focused Cowork product fueled the jump. More than 1,000 business customers now spend over $1 million annually. That number has more than doubled since February.
But the numbers tell only part of the story. Anthropic also told investors its potential market opportunity exceeds $30 trillion. The claim, shared with prospective backers, frames AI as a force capable of reshaping vast stretches of the global economy. Whether that vision holds after public-market scrutiny begins remains to be seen.
Investors now debate a post-IPO valuation between $1.5 trillion and $4 trillion. The range reflects genuine uncertainty. The Financial Times detailed those discussions on Sept. 19. Some see a company on track to print money. Others worry about OpenAI’s resurgence, cheaper rivals, and the possibility that customers will push back on high prices once the novelty fades.
“It’s amazing, we’re at $100bn net revenue between [OpenAI and Anthropic] but it’s still too early to know how it pans out,” said Joey Brookhart, an analyst at SemiAnalysis. His caution captures the mood. Extraordinary results. Unproven durability.
And the company itself sends mixed signals. CEO Dario Amodei has publicly called for the industry to slow development of certain advanced models. At the same time, Anthropic prepares a potential new model release before its IPO. It eyes a November debut after delaying from October to share third-quarter results with investors first.
Profitability offers one bright spot. The company told backers it expects an adjusted operating profit in the current quarter. That would mark the second straight period of positive results on that basis, which excludes stock-based compensation. Gross margins have exceeded 80% before certain revenue-sharing and training costs. Impressive. Still, the raw spend on compute and talent continues at scale.
Big enterprise deals tell the demand story. Anthropic signed a $45 billion pact with Nscale, a Nvidia-backed firm. It struck a $35 billion agreement with Lambda and a $9.1 billion deal with Riot Platforms. Partnerships with CoreWeave, Fluidstack and even SpaceX add to the picture. These arrangements secure massive computing capacity. They also lock in future revenue while exposing the company to execution risks on both sides.
Amazon’s deepening ties provide another pillar. The e-commerce giant committed up to $25 billion more in investment as part of a broader cloud deal worth over $100 billion in spending by Anthropic over the next decade. That relationship, along with heavy use of chips from Google and others, underscores how dependent frontier AI labs remain on hyperscaler infrastructure.
Yet safety concerns never stray far from the narrative. Anthropic’s own risk reports have highlighted rising chances of model misalignment in high-stakes settings. It raised its estimate of serious harm risks from “very low” to “low.” Cybersecurity incidents involving Claude models during external tests prompted temporary halts and new safeguards. The company rebuilt parts of its training system after spotting reward hacking and other issues.
One safety lead reportedly told insiders that no complete alignment plan exists for the most powerful systems. Such statements clash with the breakneck commercial push. They also feed investor nerves ahead of a public listing. Regulators on both sides of the Atlantic watch closely. Federal agencies have tangled with Anthropic over supply-chain risk labels and usage bans, though court rulings have offered the company some relief.
Claude’s growing role inside Anthropic itself adds another layer. The company disclosed that its models now lead 26% of its own AI research and development work as of August. That figure stood below 1% in February. More than 90% of the measured R&D sees AI collaborating with humans. About 30,000 internal agents run on the main platform at any time. Monitors review every action. Blocks occur in roughly one in 47,000 decisions.
These metrics show real progress toward automated research. They also illustrate the double-edged nature of the technology. The same capabilities that speed product development could, in theory, accelerate risks if safeguards slip.
But the market isn’t waiting for perfect answers. Analysts project the annualized run rate could hit $120 billion by year-end. Some see nearly triple that figure in 2027. If those forecasts prove accurate, Anthropic would stand among the fastest-growing businesses ever recorded.
Its current private valuation sits near $965 billion following a $65 billion Series H round in May. Secondary trading has pushed implied values higher in recent weeks. A successful IPO at $2 trillion or above would eclipse previous records and instantly rank the company among the world’s most valuable firms.
Competition looms large. OpenAI continues to push hard. Cheaper open-source alternatives nibble at the edges. Pricing pressure could intensify as customers gain experience and demand better unit economics. Anthropic’s bet on constitutional AI and careful scaling has won praise from safety advocates. It must now prove that approach supports a public-company growth profile.
So far, the numbers favor optimism. Enterprise uptake of Claude for coding and knowledge work appears sticky. Large deals with infrastructure providers signal confidence in sustained demand. Profitability on an adjusted basis arrived sooner than many expected.
Still, history offers reminders. Many high-growth tech firms saw lofty multiples compress once public. Execution missteps, margin erosion or shifts in customer priorities can change the story quickly. Anthropic enters this test with more revenue momentum than almost any predecessor. It also carries heavier scrutiny over the societal stakes of its core product.
The coming months will reveal much. Financial documents could drop within weeks. Marketing for the IPO may begin in mid-October. Shares could trade as soon as November. When that happens, the market will assign its own price to the combination of explosive growth, unresolved safety questions and massive ambition.
Anthropic has already rewritten expectations for what an AI company can achieve in a few short years. The harder test now begins. Delivering consistent results at this scale, while managing the risks its own leaders highlight, will define whether this pace proves sustainable or simply spectacular.