What Anthropic is worth today depends on how AI develops to 2030. Set how likely you think each path is, and see the price change.
The small line under each value is how much that scenario counts in the headline. Click a scenario to see its path. Jobs under "AGI, plus a funding crash" are an estimate, from the crash effect on the improving path.
The model. This page runs ai/compute-demand-2030's scenario model v4.1 (model/scenarios_v41.py) in your browser and reproduces it cell by cell. It combines five ways the AI market could split its money with five paths for how fast AI develops, plus how fast US data centres can be connected to the grid. That makes 125 combinations, weighted by your sliders. The defaults are the project's adopted weights (statement 20261112-0105) and its view of a crash (ask 05: AI training is cut for everyone and stops where the cash cushion is thin; statement 20261112-1010).
Anthropic's value is the average of two revenue paths for the frontier labs, equally weighted, each counted at zero where the equity is wiped out. The low path ties AI's work revenue to the jobs it replaces and adds consumer subscriptions: about $438bn of lab revenue in 2030 where AI keeps improving and labs sell by the use. The high path follows the labs' own plans: about $850bn. The AGI scenarios use the labs'-plans path only, because tying revenue to jobs replaced understates a world where AGI arrives (statement 20261113-0900). It discounts at 10.75% fading to 9.13%, expenses share-based pay, and dilutes shareholders where a crash leaves a cash shortfall (mkt/anthropic-valuation engine v7, F-20261113-engine-v7-kristjan-basis-36-today). Where labs sell finished products, the model assumes they charge 1.6 times as much per unit of work; that adds about $4.70 a share, and the valuation analysis contests it, holding about $31-32 as the most defensible figure (R-20261113-c4-outcome-solvency-critic). A funding crash keeps Anthropic's compute commitments in full: they are take-or-pay, so a crash cannot make the shares worth more. The $83 IPO price is out of reach within a 35-55% share. The page assumes Anthropic takes a 35%, 45% or 55% share of frontier-lab revenue, interpolated in between. The per-scenario values come from ai/compute-demand-2030 (values_equal_weight). US jobs come from ai/jobs' v4.1 grid (F-20261112-jobs-grid-compute-v4-1-2030); for "AGI, plus a funding crash" they are an estimate, the AGI path scaled by the crash effect on the improving path. Neither responds to the model-assumption sliders.
Research, not a recommendation to buy, sell or subscribe. Values rest on scenario-model figures that are not yet verified. The best AGI cases are upside cases, not forecasts. The model was built in a Claude session, and Anthropic makes Claude.