Stay informed with free updates
Simply sign up to the Technology sector myFT Digest — delivered directly to your inbox.
Goldman Sachs last month made free-to-read a note on AI’s elusive productivity benefits and returns. It was written up in lots of places as a signal that the smart money was bailing out. Here’s tech blogger Ed Zitron:
This report is so significant because Goldman Sachs, like any investment bank, does not care about anyone’s feelings unless doing so is profitable. It will gladly hype anything if it thinks it’ll make a buck. [ . . . ] For Goldman to suddenly turn on the AI movement suggests that it’s extremely anxious about the future of generative AI, with almost everybody agreeing on one core point: that the longer this tech takes to make people money, the more money it’s going to need to make.
But like god and Rory Bremner, Goldman speaks in many voices. The intended audience for a publicly available global macro research note isn’t the same as for its investment advice. So while the general public is served non-specific caution, the brokerage clients get something more measured.
Goldman’s portfolio strategy team this week updated the bank’s view in light of a market pullback since February. It notes no sign yet of capitulation among companies selling shovels to the gold rush, like Microsoft, AMD and Super Micro. It’s those companies expected to turn generative AI into sellable products and services — software makers, mostly — that have been having a tough time:
While Nvidia, the AI OG, remains in a class of its own . . .
What’s notable however is that Nvidia’s recent rally has been on fumes. Earnings forecast upgrades underpinned Nvidia’s nearly 200 per cent gain last year, but most of this year’s advance is explained by multiple expansion.
Having started the year at 25 times forward earnings, Nvidia now trades at 42 turns. A richer valuation contributed about 70 per cent of its year-to-date gain, Goldman calculates.
Heightened optimism comes in spite of the consensus expecting Nvidia’s revenue growth to decelerate from 265 per cent in the fourth quarter 2023 to just 25 per cent for the last quarter of 2025. A sizeable gap has opened up between perceptions of the size of Nvidia’s addressable market, that of its customers, and that of their customers.
Nvidia is due to post second-quarter numbers in late August, towards the end of reporting season, by which time the mood might already have darkened. A few cautious updates from downstream companies would heighten concerns about whether the $357bn spent on capex and R&D last year by Amazon, Meta, Microsoft and Alphabet (the so-called hyperscalers) can generate a decent return.
Goldman offers the caveat that, relative to history and earnings power, $357bn a year isn’t that much to spend:
At the height of the Tech Bubble, TMT stocks were spending more than 100% of cash flows from operations (CFO) on capex and R&D as many telecom stocks “overinvested.” In contrast, today’s leading TMT stocks are extremely profitable. While capex and R&D as a share of sales has increased, capex and R&D as a share of CFO equals a more contained 72% currently, and has actually been declining modestly as earnings of the group have recovered. This compares with a 40-year median of 67%. This dynamic is borne out in the economy-wide data as well.
And while capex and R&D inflation will keep outpacing sales growth, investors have at least set earnings growth at appropriate levels for the investment to wash its face, Goldman says:
Today’s hyperscalers have converted 31% of trailing 3-year capex and R&D into earnings on average during the past 5 years. They’ve spent $1.1 trillion on capex and R&D during the past 3 years (2022, 2023, 2024E). While the mapping of AI investment to earnings is not one-for-one, this aggregation implies these firms need to generate $335 billion of earnings in 2025 to achieve an ROI similar to recent history. That level of earnings would represent 16% growth versus 2024E, compared with current consensus expectations of 16%. In the late 1990s, required earnings growth for the Telecom firms equaled 22%, higher than today but in line with the consensus expectation of 24%. But ultimately, earnings collapsed by -120%.
So in the end, everything hangs on whether hyperscaler revenue accelerates before the capex depreciation charges take effect. On that point, Goldman can only offer its clients a trillion-dollar question mark:
Read the full article here