A single sell-side number, reported without its workings, is a weak foundation for any firm conclusion.
The Financial Times reports that Goldman Sachs expects Anthropic's revenue to exceed $600bn by 2030. That is the whole of what is known: a headline, a bank's name, a company's name, a number and a date. The article itself was not available, and no summary accompanied the report. Readers should treat the figure as a projection attributed to Goldman Sachs, not as a disclosed result.
The scale is what makes the claim notable. A revenue figure above $600bn would place a single AI developer alongside the annual output of large national economies, and far beyond the revenues of the software companies that dominated the previous technology cycle. If the projection is taken seriously by investors, it implies expectations of sustained enterprise adoption, pricing power and infrastructure spending on a scale that has not yet been demonstrated in public accounts.
What the report does not establish matters as much as the number. It is not known whether the figure refers to gross revenue or some adjusted measure, what assumptions about model pricing and compute costs sit behind it, or how Goldman Sachs frames the risks. Nor is it clear how the projection compares with Anthropic's own guidance or with estimates from other banks. A single sell-side number, reported without its workings, is a weak foundation for any firm conclusion.
The wider implication is that AI valuations are increasingly being anchored to long-range revenue forecasts rather than current earnings. That shifts the burden of proof onto the forecasters. If the projections hold, the sector's capital spending looks justified; if they slip, the correction would be felt well beyond one company. For now, the claim stands as a marker of how far expectations have travelled, and as a reminder of how little of the supporting analysis has been made public.
