Nvidia’s (NVDA) earnings call on August 28, covered a lot of ground.
The next-generation Blackwell chip is ramping up production in Q4, and the firm expects it to generate “several billion dollars” of revenue from it in the quarter.
The company beat on both the top and bottom lines (revenue of $30 billion exceeded expectations of $28.86 billion; EPS of $0.68 beat expectations of $0.64).
And forward-looking guidance was good, although it fell short of expectations.
But one important question came up among the analysts in the Q&A portion at the end of the call: What’s the return on investment for Nvidia’s customers.
In other words, what incentive do they have to keep spending veritable fortunes on chips?
Two years into AI fever, it’s a valid question, although asking the chip company how much money their customers earn from AI is a bit like asking Microsoft how much money their customers earn using Microsoft Office or asking McDonald’s how much their potato farmers make.
But the reason it’s worth discussing is that the return on artificial intelligence is arguably the biggest question on Wall Street.
Mega-cap tech companies have been quite vocal about their spending plans on AI data center infrastructure, with Meta (META) indicating it could spend as much as $40 billion this year and Microsoft (MSFT) spending $69 billion on CapEx (capital expenditures) in its latest full fiscal year.
The question prompted a research note in June from Goldman Sachs called “Gen AI: Too Much Spend, Too Little Benefit?” in which researchers question whether what could be a trillion dollars of spending in the coming years will pay dividends.
On the call, Nvidia CEO Jensen Huang highlighted the significant cost savings from “accelerated computing” as providing an immediate return on investment; he also alluded to potential revenue generation from any number of AI-focused start-ups who would presumably rent computing power from the data centers that are built on Nvidia chips.
What’s Behind the Question?
Data centers and high-end chips are expensive, and while big tech companies have been shoveling cash towards building them out, at some point their shareholders will need to see a concrete justification for those expenditures.
Sixty-nine billion dollars is not an insignificant amount of money, even for a company like Microsoft, and a return on investment that not only recoups those expenses but provides further room for growth is the finish line.
As it stands, Microsoft lumps all of its cloud revenues together, so there’s no definitive breakdown that shows revenue growth from AI initiatives versus, say, revenue growth from Azure web services.
For the last two years, big tech has been spending money on the promise of what AI can deliver for shareholders, but thus far their spending has really only benefited companies like Nvidia and OpenAI (which is privately held and likely not generating any revenue for itself at this stage due to the nature of its agreement with Microsoft and it’s status as a nonprofit organization with a for-profit business).
If AI does not successfully monetize, there’s little to differentiate it from other much-hyped money-burning tech products like VR (virtual reality) or the Metaverse.
So, how can we tell when AI is making money for the companies deploying it?
Big tech’s earnings calls. If the mega-cap tech companies begin highlighting the additional revenue they’re generating from spending on AI, that’s the ultimate green light. That means they’re now monetizing their initial investments. After that, the pertinent question becomes whether they spent too much rather than should they have been spending at all.
Startups. If we start seeing headlines about record fundraising for AI companies (aside from OpenAI), that investment is likely to flow through to the big tech companies. It’s not quite the same green light, but it’s significant progress.
Earnings further down the food chain. “AI” isn’t the magic word on earnings calls the way it used to be, but if companies further down the food chain are referencing new sales wins, significant cost savings, or new products powered by AI, we can presume that those investments support further spending by big tech.
What about red flags?
Most companies won’t come out and admit that they overinvested in some new growth arena (see the Metaverse), but what they will do is go quiet.
You’ll hear less emphasis on large language models (LLMs), generative AI, and AI agents in calls.
We got a taste of that in Nvidia’s earnings when management was quick to highlight their gains in accelerated computing (managing/accessing data faster) but was a little quieter on AI growth prospects.
The most important takeaway is that investment banks are already asking about the returns on AI, and now buy-side analysts have started too. Without meaningful progress in the next quarter or two, the big risk is that shareholders (especially large institutional investors) will begin to doubt the promise of AI and start slowing their investments into even the picks and shovels companies like Nvidia.
