The first AI IPO writes the comp sheet, and Anthropic is holding the pen

Anthropic's annualized revenue run rate reportedly topped $65 billion by late July, up from roughly $9 billion at the end of 2025, and bankers are now treating its expected IPO as the valuation benchmark for the wider AI sector ^[1][2]. Investment banks have told both Anthropic and OpenAI that whichever lists first will "create a model for the entire industry," which is the part that reaches your book: the first large AI listing sets the comparable that every private AI mark, including the ones you already own, gets argued against ^[3].

The mechanism is the interesting bit. Anthropic told participants in the IPO process that it expects 2028 revenue in a range the article prints as "$190 million to $200 million," previously undisclosed ^[6]. Banks are applying enterprise-value-to-revenue multiples to those 2028 projections rather than to actual current revenue, an approach normally reserved for high-growth private software ^[7]. The May Series H valuation of $965 billion works out to about five times the 2028 target ^[8], and 965 divided by 190 is 5.1, so the 2028 figure is plainly billions and the "million" is a typo ^[2].

Cryptopolitan sets that 5x against Palantir at roughly 53 times anticipated revenue and SpaceX and Cloudflare at about 41.6 times ^[9]. Those comparables use 2026 estimates while Anthropic's uses 2028, two extra years of assumed compounding in the denominator, so the gap is smaller than it looks ^[7]. Anthropic is reportedly seeking more than $2 trillion, which would be the largest IPO ever ^[4]. On the 2028 number that ask is about 10.5 times revenue ^[3]; on the current run rate it is about 31 times ^[4].

Treat the growth line with care. Run rate is an estimate of forward annual revenue extrapolated from recent history, not booked revenue ^[18]. Anthropic said in May that "our run-rate revenue crossed $47 billion earlier this month" ^[10], and the sequence runs $9 billion, then $47 billion, then $65 billion ^[11]. The same article calls that a tripling in seven months ^[19]; 65 over 9 is 7.2 times, so its own headline understates its own data ^[1]. Investors expect $100 billion to $120 billion of 2026 revenue, per the Financial Times ^[12], or 1.5 to 1.8 times the current run rate ^[5]. OpenAI has doubled to $40 billion from $20 billion at the end of 2025 ^[13], and the two companies may not measure revenue the same way ^[14].

Demand evidence is thinner than the multiple implies. Ramp's July AI Index puts Anthropic ahead of OpenAI on US business adoption, 43.5 percent of companies paying for subscriptions or tokens versus 39.7 percent ^[15], a lead of 3.8 points ^[6], and Ramp's own data also shows limits on how much firms will commit to frontier AI ^[21]. The 2028 forecast assumes revenue keeps outrunning the cost of chips, training and talent ^[17]. One more reason to wait for filed numbers: the same article says Anthropic "raised $65 billion in May" from Altimeter, Dragoneer, Greenoaks and Sequoia, a figure identical to the run rate it reports elsewhere ^[16].

Watch three things. Whether Anthropic actually lists before OpenAI, possibly this autumn ^[5]. Whether the 2028 range survives into a prospectus with segment detail behind it ^[6]. And whether public buyers accept a forward-year revenue base at all, because the article's own conclusion is that acceptance makes it easier for other AI companies to defend their multiples ^[20]. Rejection prices your portfolio off trailing revenue instead, and that arithmetic is much less friendly.