Nvidia’s most impressive feat yet? How the chip maker overcame a China ban.

On Thursday, a post-earnings rally briefly made Nvidia the most valuable company in the world. By the close it was No. 2, just barely behind Microsoft, with a market value of $3.4 trillion. This came a day after the company reported solid fiscal first-quarter results and gave a better-than-feared revenue outlook for the current quarter.

But the numbers don’t do the quarter justice. Nvidia handled the loss of a massive business line and still generated impressive growth.

Revenue for the April quarter was up 69% year-over-year to $44.1 billion, ahead of expectations. Nvidia’s data-center business, primarily driven by AI chip demand, grew even faster, up 73% from last year to $39.1 billion.

To be sure, Nvidia’s guidance was more mixed. For the current quarter ending in July, Nvidia provided a revenue forecast range with a midpoint of $45 billion, which was below analysts’ consensus of $45.9 billion. It’s the first time since Nvidia started its AI supercycle two years ago that the chip maker disappointed with its outlook. But in recent weeks, some on Wall Street had come to expect much worse.

The cause of the miss is largely outside of Nvidia’s control. It stems from President Donald Trump’s decision in mid-April to effectively ban sales of the company’s H20 chips to China.

Nvidia wisely quantified the revenue impact at $10.5 billion in total across the April and July quarters. That helped investors quickly realize that Nvidia would have significantly raised guidance versus consensus if not for the H20 effect. It suggests there’s better than expected strength in the non-China business.

Sure enough, after an initial earnings bump, Nvidia’s gains accelerated in after-hours trading. On Thursday, the stock finished up 3.3%, making it one of the S&P 500’s best performing stocks on the day.

The good news for Nvidia investors is the market is forward-looking. The China headwinds have been largely de-risked, with an outlook that’s reset for potential upside going forward.

Even with all the growth to date, Nvidia CEO Jensen Huang said demand for the company’s AI infrastructure products is still improving. “AI is this incredible technology that’s going to transform every industry," he said on the earnings call with investors. “We’re really at the very beginning of it, because the adoption of this technology is in its early, early stages."

Demand from the largest technology companies continues to rise. Heading into the latest earnings season, some on Wall Street were concerned that the industry would lower its full-year capex budgets to spend hundreds of billions on AI infrastructure amid macro uncertainty. Instead, the opposite has happened. Meta Platforms revealed it planned to spend more on AI and Microsoft said it was more supply-constrained for AI capacity than they anticipated.

Of course, robust demand doesn’t mean much if you can’t serve it. Nvidia’s management put to rest several supply side concerns this week too. Chief Financial Officer Colette Kress said the company’s Blackwell GPU production is ramping up, with major cloud computing companies now deploying nearly 1,000 top-of-the-line GB200 NVL72 server racks on a weekly basis. Each rack costs several million dollars. Kress said Microsoft alone is expected to purchase hundreds of thousands of GB200s to serve its customers.

More important, Nvidia’s next AI server, the Blackwell Ultra, is on track. The company sent test shipments of the Blackwell Ultra GB300 NVL72 AI server to some customers in May, Kress said, and expects to start production shipments later this quarter.

As a result, Wall Street is getting more optimistic about the prospects for Nvidia’s second half. The report showed an “acceleration of the business other than the China headwinds," Morgan Stanley analyst Joe Moore wrote. “Everything should get better from here."

Huang pointed to several growth drivers behind the rising demand for AI; so-called reasoning, agents, and sovereign AI.

The latest AI models have a reasoning feature that allows them to reflect more thoroughly on a query by performing numerous thought computations to arrive at a higher-quality response. Huang says reasoning consumes more than 100 times more compute resources compared with prior AI models. “Reasoning models are driving a step-function surge in inference demand," he said. Inference is the process of generating answers from AI models.

Then there are the AI agents, which according to Huang, are starting to become a reality. Agents can take simple directions and complete multistep computer actions often used to automate tedious business and personal tasks. “Agentic AI is game changing," he said, adding the technology is finally working and being successfully implemented by enterprises.

Lastly, there’s sovereign AI. Nvidia recently announced large deals with Saudi Arabia and the United Arab Emirates. The first phase of Saudi Arabia’s AI factory buildout features an order for 18,000 Nvidia AI servers with potential for “several hundred thousand" GPUs over the next five years. The U.A.E. agreement expects its first AI cluster to go live next year. Huang said on the conference call that there are “a whole bunch" of other deals that haven’t been announced, noting that he would be traveling to Europe next week to announce some of them.

Nvidia stock is now up nearly 40% from the trade war lows in April. It’s up 4% on the year. Even after the latest rally, the stock isn’t expensive, trading at 29 times forward earnings estimates, while analysts expect 45% sales growth over the next 12 months. Compare those numbers to Apple, which trades at 28 times with 4% sales growth. Nvidia is far more attractive on a valuation-to-growth basis.

With every earnings report, Nvidia looks more like a once-in-a-generation growth story akin to the early years of the PC revolution and the iPhone; those are trends that both lasted more than a decade. For Nvidia, we’re only two years in.

Write to Tae Kim at tae.kim@barrons.com