high cagr

Agreed that Anthropic's S1 is not going to save the AI trade, and IMO, it does not have to. I may sound like a lunatic but I do not care at all, and I mean at all, what the labs are generating in revenue exiting this year. Widespread adoption has not even taken place yet. What, only 2-3% of households are paying for a subscription? I do think Anthropic IPO at $1.5T is a lot more reasonable than $2T. IMO, pricing at $2T risks setting up a disappointment and putting unnecessary pressure on the trade as a whole once it's public.

For a second, just remember Google was processing 9.7 trillion tokens/month in May 2024 and by May of this year 3.2 quadrillion. And yes, that is the old workload. A single-shot inference call is the 1x unit whereas reasoning models run 100x that and agentic systems near 1000x that. We’re still very early in the shift from chat to agents, so per-task compute is about to re-base on a user base that is still growing. We have yet to hit the point where the average request is a job rather than a simple question.

My Intel ER bingo card: > Revenue at/above $17.4B > Positive GAAP EPS > Non-GAAP EPS at/above $0.46 > Foundry revenue above $6.0B > Foundry operating loss below $2.0B > Foundry loss below $1.8B > External foundry revenue broken out as its own line > Any packaging or EMIB-T capacity disclosure > 18A yield # > Tesla 14A agreement volume, value, or timing > Capex raised > Q4 revenue guide midpoint $17.5B > Q4 guide above consensus > Client guided down QoQ > Memory cost inflation named as client headwind > Q4 GM guided above 43% > Second 14A customer named > Clearwater Forest timing reaffirmed with specifics > “Capacity constrained” / “sold out” on any node > DCAI growth above 50% > An analyst asks about Nova Lake tile sourcing