Anthropic's IPO case rests on a 2028 number, and your roadmap sits inside it

Anthropic's initial public offering hinges on an internal forecast that it will book between $190bn and $200bn of revenue in 2028, a previously unreported figure that bankers and investors are already using to justify a price tag approaching $2tn, according to a Reuters exclusive reported by TNW ^[1]. For anyone whose product depends on Anthropic's model pricing, rate limits and deprecation schedule, that forecast is now the plan they are attached to, whether or not they signed up for it.

Start with the arithmetic, because it is the whole story. Anthropic's revenue run rate sat at roughly $47bn as of May 2026, up from about $9bn at the end of 2025 ^[2], which is a jump of roughly five times in about five months ^[0]. The 2028 target implies revenue more than quadrupling again in a little over two years ^[3]; taking the midpoint, $195bn is about 4.1 times the current run rate ^[1]. Second-quarter 2026 revenue is projected to reach at least $10.9bn, more than double the previous quarter, alongside a forecast first quarterly operating profit of around $559m ^[4]. That profit is about a 5 percent operating margin on the quarter ^[2]. To land the midpoint in 2028, quarterly revenue has to average roughly $49bn, about 4.5 times that $10.9bn ^[3].

The valuation method matters less than it sounds. Bankers are reportedly applying enterprise-value-to-revenue multiples to the 2028 forecasts rather than current earnings, a two-years-forward approach that is rare in public markets but was used before the Cerebras and SpaceX flotations ^[5]. Palantir trades at roughly 53 times expected 2026 revenue, while SpaceX and Cloudflare sit around 41.6 times ^[6]. Against the $195bn midpoint, a $2tn valuation is only about 10 times revenue ^[4]; against today's $47bn run rate it is about 43 times, roughly the SpaceX and Cloudflare multiple ^[5]. The aggression, in other words, is in the forecast, not the multiple. TNW's framing is that the valuation rests almost entirely on a number nobody can yet check ^[7].

The cost base is where builders should pay attention. Heavy spending on GPUs, compute and model training is pressing on current margins, and the bull case assumes those costs fall as a share of revenue as the business scales ^[8]. The reporting does not say price increases are planned. It does say cheaper rivals are already threatening the economics that underpin these valuations ^[9]. A company underwriting a near-$2tn number on 2028 revenue has a structural interest in protecting realised revenue per customer, and customers are the surface that interest lands on.

David Merkel of Aleph Investments told Reuters, per TNW: "Could they get a $2 trillion valuation, yeah they could and I just wonder if it would stay there over time" ^[10]. The ladder has been climbing regardless: investor offers at an $800bn valuation, reports of something nearer $900bn in a mooted $50bn round, and IPO chatter now at $2tn ^[11].

Three things to watch. Whether the $559m operating profit actually prints, since the entire margin-improvement story hangs off that inflection ^[4]. Whether the $190bn to $200bn range survives into a prospectus, where it becomes a disclosed forecast rather than an internal one ^[1]. And whether the cheaper rivals TNW flags start showing up in your own build-versus-buy spreadsheet before 2028 does ^[9].