英伟达CFO指引FY2028收入增长70%远超华尔街预期44%，盘后涨5%；Salesforce指引略超预期盘后涨13%

AI revenue reporting: slop
Good morning. It’s an expectations game, innit? Last night, Nvidia’s earnings report included the CFO saying she expected revenues to grow 70 per cent in fiscal 2028. Wall Street has been expecting 44 per cent growth — implying an additional $100bn of unexpected sales. The market responded to this staggering disclosure by pushing the stock up by a nice-but-not-staggering 5 per cent in late trading. Meanwhile Salesforce, a software group whose shares have been taken to the woodshed by fears of displacement by AI, expects revenues for the rest of the year to come in a shade above analyst estimates, and its shares rose 13 per cent after the bell. Email us: [email protected].
AI revenue reporting
Anthropic, maker of Claude, is reportedly planning to list in October. It has mastered the dark arts of leaking financial information during what is supposed to be a quiet period ahead of the initial public offering. The steady drip of positive news stories about revenue growth, potential $2tn valuations and AI models too powerful to release to the public is doing the work of the IPO underwriters before the roadshow has even started. In less than eight months, Anthropic’s annualised revenue is said to have gone from $9bn to over $65bn.
Do the numbers stack up? The funny/scary thing about private companies is that there is no agreed standard for financial disclosures. We’re not saying anyone is cooking the books; internally, they have to follow the same accounting rules as everyone else. But privates can choose which financial metrics to disclose, and can throw in a few heroic assumptions for good measure.
To see how this works, just look at Silicon Valley’s preferred metric: annual recurring revenue (ARR). It takes a recent month’s revenue and multiplies it by 12 to set an expectation for the year to come.
There are many problems with this. To start, there are also significant differences in how companies calculate the figure. Here is Ed Zitron, perhaps the pre-eminent sceptic of AI business models:
[ARR] can mean everything from “[actual month] x 12” to “[30 day period of revenue] x 12” and in most cases it’s a number that doesn’t factor in churn. Some companies even move around the start dates for contracts as a means of gaming this number.
ARR, also, doesn’t factor seasonality of revenue into the calculations. For example, you’d expect ChatGPT to have peaks and troughs that correspond with the academic year, with students cancelling their subscriptions during the summer break. If you use ARR, you’re essentially taking one month and treating it as representative of the entire calendar year, when it isn’t.
Investors don’t seem to mind, at least while a company is private. If you’re a venture capitalist or large institutional investor, you’ll probably get a look under the hood anyway. It’s just the rest of us who are expected to take the number on faith.
It wasn’t ludicrous for early-stage software-as-service companies to report in this way. Their business model was built on locking in customers on recurring contracts. If, for example, Salesforce said in July that it had five customers paying $10mn a year, you could be reasonably certain it would book at least $50mn in revenue over the next year.
But for AI companies that have just switched from flat fees to usage-based pricing, this is nuts. The figures don’t represent contractually guaranteed revenue. Anthropic has been taking one month of consumption data and assuming that trend will hold for the next 12 months. And they don’t stop there.
They add up usage-based and subscription revenue to report a higher number. It’s very likely the company is seeing a surge in enterprise demand for its agentic tools, Claude Code and Cowork, but subscription revenues aren’t enough to justify a more than 600 per cent increase in annualised revenue in less than a year. That assumes that non-subscription revenue will stay high even as the industry enters a price war and is struggling to get customers to pay for the most advanced models. Chart courtesy of our news colleagues:
We haven’t even got to everyone’s favourite topic; circular financing, or gross margins (which are low). ARR fuels the sense of AI companies’ unstoppable momentum long before they have found a profitable business model.
(MacFadden)
Climate revisited
Last week, we wrote a bit about the effects of extreme heat on asset prices, mainly discussing the results of this paper. The authors — Viral Acharya of NYU Stern, Suresh Sundaresan of Columbia, Timothy Johnson at the University of Illinois and Tuomas Tomunen of Boston College, found that localities and companies exposed to heat stress paid wider spreads on their debts and a higher risk premium on their equity.
Luca Taschini of University of Edinburgh Business School wrote Unhedged to argue that what matters for equity returns isn’t heat levels, but abnormal temperature variability. Even without record-breaking heatwaves, he thinks, unpredictable temperature swings hurt exposed companies. In months with high temperature variability, in fact, Taschini showed in a recent paper (written with Leonardo Bortolan at the University of Bologna and Atreya Dey at the University of Cambridge) that highly exposed stocks underperformed their less exposed peers by 4.4 per cent, annualised.
Here’s Taschini:
Firms operating in areas experiencing elevated high temperature variability also see lower revenues and profits . . . Yet turning that weather info into an earnings view appears difficult: analyst forecasts diverge more [in the wake of temperature variability], and the eventual hit to earnings tends to be underestimated. Hence why speaking about expected impact on equity is harder.
In short, analysts are still struggling to price in the weather effects at the firm level.
There are caveats. For one, Taschini and his co-authors calculate the return spread between stocks affected by unstable weather and their unaffected peers that tend to disappear within two months. With greater frequency of weather irregularities, how these passing shocks change the picture over the long run is the crucial question. For now, the research is limited by lack of data: the persistent pattern on extreme weather events is a relatively recent phenomenon. Over time, though, the return impact of climate change should become clearer, and may be substantial.
(Kim)
One good read
“Peculiar forces of fate”
FT Unhedged podcast
Can’t get enough of Unhedged? Keep following Rob Armstrong and Katie Martin’s incisive and unfiltered analysis on the Unhedged podcast, a twice-weekly, 15-minute lively dive into the latest markets news and financial headlines. New episodes every Tuesday and Thursday.