Nvidia Defends AI Story as Market Value Slips Below $5 Trillion
SANTA CLARA, Calif.—Nvidia Corp. held its annual shareholder meeting on Wednesday with a familiar message and a restless audience. Chief Executive Jensen Huang repeated his case that artificial intelligence constitutes a new industrial revolution and declared that “AGI is here,” drawing applause from shareholders in the room and skepticism from the market outside it.
The stock’s reaction was the story of the day. Nvidia shares continued to slide, extending a selloff that began on June 23, a session traders have dubbed “Black Tuesday,” and pushing the company’s market capitalization below the $5 trillion mark for the first time in months. The meeting gave Huang a stage, but it did not give investors a reason to buy.
Inside the auditorium, Huang sketched a future in which Nvidia’s chips underwrite everything from autonomous vehicles to drug discovery, and he pointed to demand that he said still outruns supply. Outside it, a different narrative prevailed. Analysts have grown louder in recent weeks about valuation, profit-taking and the arithmetic of a company that must sell ever larger clusters of accelerators to sustain growth.
The meeting produced one new commitment: Nvidia plans to invest roughly $2 billion in photonics-based interconnect technology, a bet aimed at the data-center bottleneck that has emerged as AI clusters grow too large for traditional copper wiring to move data efficiently between chips. People familiar with the plans said the investment spans internal development and stakes in startups working on optical interconnect, with the first products expected in data centers by late 2027.
Huang framed the photonics push as the natural extension of Nvidia’s ambition to sell entire AI factories rather than individual chips. As clusters scale past hundreds of thousands of processors, the connections between them become as important as the processors themselves, he said. The company has said its next-generation systems will pair its accelerators with optical links at the rack level.
The meeting also gave Huang a chance to address power, the other constraint on the AI buildout. Nvidia said its partnership with IREN Ltd., the data-center developer formerly known as Iris Energy, has locked in 5 gigawatts of electricity to power Nvidia-branded AI facilities, with the first sites under construction in Texas. The arrangement, first announced last year, is part of a broader push by Nvidia to guarantee the energy supply that hyperscale customers demand before they commit to clusters.
The announcements did little to calm shareholders worried about the balance sheet. Nvidia has been spending at record levels—prepaying for fabrication capacity, buying land for data centers, and now investing in photonics and power—while its cash position draws down. Several analysts have flagged the company’s rising capital intensity, and at least two have issued what they described as cautionary ratings on the stock, citing the risk that demand eventually normalizes faster than supply.
“There is no question about the technology. The question is the price embedded in the stock,” one portfolio manager who attended the meeting said. “The company is executing, and the multiple is still enormous.”
The selloff has been sharp. From its high in late May, Nvidia has lost more than 10% of its value, erasing hundreds of billions of dollars in market capitalization. The decline has dragged the broader chip sector with it, compounding worries that the AI trade, the engine of the past two years’ bull market, is entering a choppier phase.
Huang acknowledged the volatility only indirectly, telling shareholders that the company’s fortunes will ultimately track the pace of AI adoption, not daily quotes. He cited growth in enterprise software, robotics and sovereign AI programs as evidence that demand is broadening beyond a handful of hyperscalers.
Analysts said the meeting was notable for what it did not contain: no new product dates beyond the announced roadmap, no update on the company’s relationship with the largest cloud providers, and no direct response to the short-selling reports that have circulated in recent weeks. Instead, Huang leaned on the industrial-revolution argument he has used since ChatGPT upended the industry.
Whether that argument still moves the stock is the open question. Nvidia’s forward price-to-earnings ratio has fallen from its peak, but remains well above the market average. The company’s own guidance calls for continued triple-digit growth in its data-center segment through the year, and its customers have publicly committed to record capital spending.
Investors left the meeting with the photonics figure, the power agreement and the same conviction speech they have heard before. The market left with another day of losses. The gap between those two outcomes is now the largest risk in the AI trade, and it is a gap that no shareholder meeting, however polished, has yet closed.
