Market Data 
Everyone on Wall Street wants to know whether massive AI investments will really pay off.
Nvidia CEO Jensen Huang tried his best during Wednesday’s closely watched earnings call, but he couldn’t expunge these doubts entirely.
The company’s third-quarter revenue forecast didn’t help. It missed so-called whisper numbers that represent the most bullish expectations. That left Nvidia stock down 7% in after-hours trading.
Analysts peppered the CEO with questions about the return on AI spending. His response was several versions of: “the people who are investing in Nvidia infrastructure are getting returns on it right away.”
Though Nvidia’s largest customers have yet to boast significant return on billions in AI investing, Huang drilled down on several places he sees profits coming from.
GPUs speed everything up
“Accelerated computing, of course, speeds up applications. It also enables you to do computing at a much larger scale, for example, scientific simulations or database processing,” said Huang.
“Accelerated computing” is Huang’s term for the kind that Nvidia graphics processing units enable. It’s also called parallel computing where the chips do many tasks simultaneously instead of in a sequence. That’s the foundation of generative AI and Huang contended that almost all existing computing jobs are headed in that direction for one major reason.
“It’s not unusual to see someone save 90% of their computing cost,” when they convert to accelerated computing, Huang said. The reason, he continued, is “you just sped up an application 50x. You would expect the computing cost to decline quite significantly.”
Consumer targeting
Recommendation engines, like the ones that tell you what to stream next, and digital ad targeting, were two modern data-processing tasks that are converting fast to accelerated computing, Huang said.
Cheaper, more accurate recommendations and better-targeted or ads could generate more revenue for companies that adopt those technologies. Meta, for instance, has turbo-charged profit in recent years by using AI to improve content recommendations and ad targeting.
Sophia Velastegui, a venture capitalist and former Microsoft AI executive, told Insider that identifying the return on AI investments may not be easy as it gets baked into these everyday functions of the internet.
“You may not be able to be like, ‘Hey, this growth is specifically from generative AI,” Velastegui said. She added that companies may not be keen to reveal the details of their AI gains for competitive reasons.
The AI cloud wave
Huang’s third pillar of AI ROI, at least for cloud providers, is the startup development frenzy across generative AI applications.
The largest cloud-computing providers, including Amazon and Microsoft, are major customers of Nvidia. When they buy GPUs, they put them in data centers and get paid relatively quickly for renting out this new AI computing capacity.
“Everything you stand up, you are going to get rented because so many companies are being founded to create generative AI and so your capacity gets rented right away, and the return on investment of that is really good,” Huang said.
Indeed, Nvidia has propelled a slew of new or reimagined cloud providers that purchase chips almost exclusively from Nvidia and specialize in the latest and greatest in AI computing.
It is going to be difficult to assuage investors concerns in this area because ultimately, Huang isn’t the executive they need to hear the message from. And what they’ve heard from other top tech execs, and top Nvidia customers, is calls for patience.
Meta spent $8.5 billion in the second quarter on computing infrastructure for AI and the metaverse. It plans to spend between $37 billion and $40 billion this year, though CEO Mark Zuckerberg told investors in July not to expect immediate returns.
