Anthropic's Revenue Machine Hits $65 Billion Run Rate as IPO Speculation Hits Fever Pitch

Anthropic is no longer just the “safety-focused alternative” in the frontier-model race — it is now one of the fastest-scaling software businesses in history. CNBC confirmed on Monday that the company’s annualized revenue run rate hit $65 billion at the end of July, roughly a sevenfold increase from a year ago. The figure was shared with investors ahead of an anticipated initial public offering, and it lands on top of an already-stunning quarterly report: preliminary Q2 revenue of more than $11.5 billion, up from just $787 million in the same quarter of 2025.

The numbers, in context

To understand how abnormal this growth curve is, stack the quarters side by side. Anthropic generated roughly $4.8 billion in revenue in Q1 2026 — a figure that itself looked explosive at the time. Q2 came in at more than $11.5 billion, a sequential jump of about 140% and a 14x year-over-year increase. By mid-May the company had disclosed a $47 billion run rate to investors; by the end of July that number had climbed another 38% to $65 billion. For comparison, Anthropic needed about 16 months to go from $1 billion in annualized revenue to $30 billion. It then added $35 billion of run rate in roughly three more months.

The more consequential detail buried in the investor documents: Q2 2026 was Anthropic’s first quarter of positive adjusted operating income. Every frontier lab has historically bought growth with billions in compute subsidies — effectively selling tokens below the cost of serving them. Crossing into positive adjusted operating territory signals that Anthropic’s mix has shifted decisively toward high-margin enterprise and API workloads, and that Claude’s pricing power is finally outrunning its inference bill.

The IPO math: $190–200 billion by 2028

The run-rate disclosures are not happening in a vacuum. Reuters reported on August 15 that Anthropic’s IPO valuation hinges on a projected $190–200 billion in revenue for 2028, according to sources familiar with the company’s financials. Backers, according to multiple reports, are floating a valuation that could brush $2 trillion — a number that would have seemed absurd for a company founded in 2021, and which would place Anthropic among the most valuable companies on Earth at debut.

The aggressive framing has drawn exactly the skepticism you would expect. Critics note that bankers are effectively pricing the offering off a 2028 forecast, “skipping three years of actual results entirely.” Even bulls concede the gap: $65 billion of run rate needs to roughly triple to approach $200 billion of annual revenue, and the AI demand environment would need to keep compounding without a single air pocket — through enterprise budget cycles, price pressure from open-weight rivals like Meta’s Muse Glimmer, and any macro tightening that hits discretionary AI spend first.

Where the money comes from

Anthropic’s revenue engine has three visible cylinders. First, API traffic from developers building on the Claude family, which has become a default choice in coding and agentic workloads. Second, enterprise deployments — Claude for Enterprise and cloud partnerships through Google Cloud and Amazon Web Services, which bundle Claude into managed offerings and carry multi-year commitments. Third, the consumer subscription base, smaller than OpenAI’s but growing as Claude expands its role as an everyday assistant.

The competitive stakes are sharpening on all fronts. OpenAI remains larger on consumer scale, Meta keeps pressure on price with open-weight releases, and Google bundles Gemini across its workspace empire. Anthropic’s counter has been depth over breadth: dominate the segments — coding agents, enterprise automation, long-horizon tasks — where Claude’s reliability commands premium pricing rather than competing for casual chat traffic.

Why it matters

The $65 billion run rate matters for three different audiences.

For AI bulls, it is the strongest evidence yet that frontier-model revenue is not merely round-tripped vendor financing — real enterprises are paying real money at escalating scale, and at least one lab is approaching self-funding economics.

For skeptics, the IPO’s reliance on a 2028 forecast makes it the cleanest test yet of whether public markets will underwrite AI growth narratives at face value. A wobbly debut or a post-IPO miss against the $190–200 billion path would reprice the entire sector’s private marks — including OpenAI’s — overnight.

For the industry itself, positive adjusted operating income changes the strategic game. A lab that can fund a meaningful share of its own training runs is less dependent on the circular financing structures (chip vendor invests in lab, lab buys chips, cloud partner commits to both) that have drawn increasing regulatory and investor scrutiny. It also gives Anthropic optionality: it can price aggressively to defend enterprise share without asking shareholders to absorb unlimited losses.

What to watch

Three signposts will tell us whether the $200 billion path is real. First, the Q3 print — anything above ~$16 billion in quarterly revenue keeps the run-rate compounding on schedule; a sequential deceleration would immediately complicate the IPO narrative. Second, the S-1 filing itself, whenever it arrives, which will force disclosure of customer concentration, gross margins on inference, and how much revenue flows through related-party cloud partnerships. Third, compute commitments — the pace at which Anthropic locks in data-center capacity (its own and through partners) is the leading indicator of how seriously management believes its own 2028 forecast.

One thing is already clear: the era of treating Anthropic as the scrappy second mover is over. The question heading into the IPO is no longer whether Claude can compete — it is whether the most aggressive revenue forecast in software history can survive contact with public markets.