They are reserving future HW productions to meet their hypothetical usage as well. Which is why others (like Apple) can’t reserve it for their future products.
Yet the AI labs are speculating on usage, and spending money from investments without clear revenue path.
Sorry, I should have said "profit path", good catch!
They have revenue, but their cost scales with revenue and they're losing more than they are making.
Their costs do not scale linearly with revenue. Inference is expensive, but it's a variable cost. Anthropic's overall costs include massive fixed costs in training, which are the same regardless of usage.
It's easy to falsify the claim with a simple experiment: imagine they had no customer at all, $0 in revenue. Their costs would still be massive. If the claim were true, $0 revenue should mean $0 costs, right?
That's convenient accounting. The reality is that they can't stop training since they risk losing customers if they do so. So they shouldn't factor it out of profitability analysis.
This is not sustainable forever unless their hypothetical usage is realized, and eventually the bill will come due.
Meanwhile, component makers will surely be spinning up more capacity, some of them in a foolhardy manner, and if the bubble does burst, 3-6 months later we'll be seeing fire sales on components and component makers going bankrupt (or getting bailouts, if considered of national importance)
except if as cost here only the inference cost is considered, and not the capital investment, and maintenance costs (not to mention r&d, marketing, and others).
to put it another way, if they just had the corporate API subs today, would they be profitable?
Yet the AI labs are speculating on usage, and spending money from investments without clear revenue path.