The weekend headlines were ablaze with news of Anthropic’s quarterly revenue soaring by more than 14 times year on year. According to Bloomberg, we are talking about a jump from $787m to $11.5bn.
Whether these numbers are entirely accurate isn’t the point. In this wild world that we live in, it’s plausible that Anthropic achieved that level of growth and generated that magnitude of revenue.
For context, poor Adobe could only manage revenue of $6.6bn in the latest quarter, despite practically begging every computer user in the world to get a PDF summary of whatever they are reading on the screen.
This is the kind of fuel that investment bankers just love adding to an IPO fire as they look to maximise the valuation when the stock lists. The market is expecting the Anthropic IPO before the end of the year. It’s all but guaranteed that the accompanying presentations to investors will be built using Claude.
Anthropic’s valuation is anyone’s guess, but it’s going to be an eye-watering price-to-revenue multiple either way. Institutional investors and the broader market will no doubt keep an eye on SpaceX’s performance as a barometer for post-IPO behaviour in this market. The stock fell to below $105 after peaking at $225 in the aftermath of the recent IPO. Now at $140, it seems to be finding a trading range.
Meanwhile, at the top of the AI value chain, Taiwan Semiconductor Manufacturing Company (TSMC) has reported revenue growth of 44.7% in July. To give some perspective on how big this company is, the monthly revenue figure of $14.5bn is more than Anthropic made in an entire quarter. Here’s another way to think about it: TSMC generates more revenue in two weeks than Adobe does in three months.
Anthropic’s profitability remains to be seen, but we know for sure that TSMC makes plenty of money. In the quarter ended June 2026, it achieved a gross profit margin of 67.7%, up from 58.6% a year prior. This was driven by the manufacturing efficiencies that the company can achieve off the back of 36% revenue growth. July represented a sharp acceleration in revenue growth, so the market is practically licking its lips at the thought of the third-quarter margins.
This is why the TSMC share price is up 40% year to date in dollar terms, well ahead of Nvidia’s rally of 21%. Intel is worth mentioning for its 178% jump in 2026, a perfect example of what happens when the leaders of the free world decide they like your stock.
Another recent data point is CoreWeave, an AI-focused cloud infrastructure company. As a perfect example of how innovation can create new business models, CoreWeave has inserted itself into the value chain by warehousing and leasing expensive AI computing power. Customers range from AI model developers to enterprises building AI applications.
If you need graphics processing units (GPUs) on a reasonably flexible basis, then CoreWeave is only too happy to be your tech landlord.
You would expect this to be a lucrative space given the GPU shortage we keep hearing about. The share price is up 33% year to date, but it’s been on a wild ride with a 52-week high of $153.20 and a 52-week low of $60.55. The market responds sharply to news in this sector, with one of the major corrections in recent months being driven by Mark Zuckerberg’s announcement that Meta is willing to sell excess compute. When the word “excess” starts being used by a hyperscaler, it’s hard to use the words “scarcity” and “supply constrained” elsewhere in the value chain.
The subsequent recovery in CoreWeave’s share price suggests that the market is taking Zuck’s comments with a pinch of salt. It also helps that CoreWeave just grew revenue by a whopping 112% year on year. It claims to have a revenue backlog of $104bn — an astonishing number for a company that has guided revenue of about $3.5bn in the next quarter.
With revenue growth like this, you would expect profits to improve rapidly. Instead, the net loss more than doubled, from $290m to $626m.
But this is “Whose AI is it anyway?”, where everything is made up and the profits don’t matter. Adobe, the darling of the software-as-a-service era, is a $105bn company on a revenue multiple of 4.2. CoreWeave, a wildly loss-making company whose valuation seems heavily dependent on the persistence of today’s AI bottlenecks, is a $58bn company on a revenue multiple of 7.7.
At least one of these valuations is probably wrong. Fortunes will be made and lost while the market figures out which one that is.
