Everyone on Wall Street wants to know whether big AI investments will actually pay off.
Nvidia CEO Jensen Huang did his best during a highly-anticipated earnings call on Wednesday, but he couldn't completely dispel those doubts.
The company's third-quarter revenue guidance didn't help either, falling below the whisper number that represents the most bullish expectations, sending Nvidia shares down 7% in after-hours trading.
Analysts peppered the CEO with questions about the profitability of AI investments, and his response was that “those who are investing in Nvidia's infrastructure are seeing immediate returns.”
While Nvidia's biggest customers can't yet boast big, billion-dollar gains from their AI investments, Huang took a closer look at some areas where he sees benefits coming.
GPUs speed everything up
“Of course, accelerated computing will speed up applications, but it will also enable us to do much larger-scale computing, such as scientific simulations and database processing,” Huang said.
“Accelerated computing” is Huang's term for the computing enabled by Nvidia's graphics processing units. It's also known as parallel computing, in which the chip performs many tasks simultaneously rather than sequentially. This is the basis of generative AI, and Huang argued that nearly all existing computing jobs are heading in that direction for one big reason.
“When you switch to accelerated computing, it's not uncommon to see a 90 percent reduction in computing costs,” Huang says. That's because “applications become 50 times faster. You can expect your computing costs to drop dramatically.”
Consumer Targeting
Recommendation engines that tell you what to stream next and digital ad targeting are two modern data-processing tasks that are moving quickly to accelerated computing, Huang said.
Cheaper, more accurate recommendations and better targeted advertising could bring additional revenue to companies that deploy these technologies: Meta, for example, has seen its profits soar in recent years by using AI to improve content recommendations and ad targeting.
Sofia Verastegui, a venture capitalist and former Microsoft AI executive, told Insider that pinpointing the return on AI investments may not be easy as the technology becomes integrated into the everyday functioning of the internet.
“You may not be able to say, 'This growth is specifically due to generative AI,'” Verastegui said, adding that companies may be reluctant to disclose details of their AI benefits for competitive reasons.
The wave of AI cloud
The third pillar of AI ROI that Huang sees, at least for cloud providers, is the frenzy of startup development across generative AI applications.
Large cloud computing providers such as Amazon and Microsoft are major NVIDIA customers: they buy GPUs, install them in their data centers and then rent out this new AI computing power to generate revenue relatively quickly.
“Anything you launch will be rented out, because a lot of companies are being set up to develop generative AI, your capabilities will be rented out quickly and the return on investment is very good,” Huang said.
In fact, Nvidia has driven a slew of new or redesigned cloud providers that buy chips almost exclusively from Nvidia and specialize in the latest and greatest AI computing.
Assuaging investor concerns in this area will be tough, because after all, Huang isn't the executive they need to hear the message from. What they've heard from other top tech executives and from Nvidia's top customers is a plea for patience.
Meta spent $8.5 billion on AI and metaverse computing infrastructure in the second quarter and plans to spend $37 billion to $40 billion this year, though CEO Mark Zuckerberg told investors in July not to expect any immediate profits.
