For three years now, Nvidia’squarterly performance has become a milestone for the artificial-intelligence trade. In that time, the company’s sales have grown by 1,000% while net income has risen 2,750%. The stock has surged 363%.
Investors get their latest look at Nvidia earnings Wednesday afternoon when the AI-chip leader reports its second-quarter results.
Wall Street analysts expect second-quarter adjusted earnings per share of $2.09, doubling over the past year, on sales of $92.3 billion, also twice what the company posted a year ago. Like Alphabet, which reported last month, Nvidia may see a big boost to earnings from unrealized gains in its stake in SpaceX, which went public during the second quarter.
But Nvidia has been a continuing victim of high expectations. In each of the last eight quarters, the company has beaten the Wall Street consensus on key metrics; on six of those occasions, the stock has declined the following day.
Shares have underperformed the S&P 500 over the past year, and they’re down 1% in the last three months. Until Tuesday, the stock had fallen in seven straight trading sessions. Its forward price/earnings ratio sits lower than that of the S&P 500 index, despite the company’s projected 89% EPS growth this year.
Investors see a few weaknesses in the bull story. The first bear narrative is that the AI data center boom will begin to fade next year. Nvidia’s best customers—“hyperscalers” like Amazon.com, Microsoft, Alphabet and Meta Platforms, Inc.—have reached the limits of their own cash flow and are dipping into debt and now even equity financing. This year, the four companies will probably spend over $700 billion on new AI data centers. How much higher can that go?
About half of Nvidia’s data center sales come from non-hyperscale customers, including smaller cloud companies and enterprises. Debt is funding much of their chip purchases, as well. We are in the midst of a historic upswing in demand, but ultimately chip making is a cyclical business and a down cycle will come some day. Some investors are betting that day will be sooner rather than later.
Nvidia is countering this narrative by acting as the “buyer of last resort,” guaranteeing portions of large data center deals like the half-trillion dollar site SoftBank is building in Ohio for OpenAI. This effectively lends Nvidia’s credit rating to the deal, getting SoftBank a better rate than it could have gotten on its own. But it also further concentrates Nvidia’s investments in the AI boom. If AI demand goes south, it could ultimately cost the company over $100 billion.
Another threat to the AI boom looms larger every month, and that’s political opposition to new data centers, especially the large ones that are planned. In December, an Emerson College poll found that 42% of likely 2026 midterm election voters opposed nearby data-center construction. By July that opposition had risen to 63%. Even Texas—the state with the most data center construction and a preference for free-market policies—has slowed building down with new administrative requirements.
Data-center developers “had not been working in collaboration with local governments, and so they basically dug their own grave,” Texas Gov. Greg Abbott told ABC News this past Sunday. “That’s why they got the backlash they deserve.”
Opposition to data centers has become good politics for both U.S. parties, and the coming midterm elections is shaping up to be a key test of the buildout.
Lastly, Nvidia is seeing increased competition from traditional competitors like Advanced Micro Devices and other newer chip makers. Customers want to diversify their supply chains without having to pay the chip maker’s 300% price markup. The biggest competitive threat may come from Nvidia’s own customers: All the hyperscalers now make competing AI computing chips for their own use. Alphabet has been working on its chips for over a decade now.
Nvidia’s answer here is a relentless pace of innovation. Chips are released on a “tick-tock” cadence, with major architecture revisions every two years coupled with incremental improvements in the intervening year. Alongside the hardware improvements, the company continues to improve the software ecosystem around its chips, a key part of its protective moat.
Meanwhile, the data suggest that Nvidia’s products are as hot as ever. Rental prices for its servers remain strong and have gone up even as supply comes online in new large data centers. The only thing holding Nvidia back may be its ability to deliver new chips to customers.
