"AI investment demand will explode further"… Jensen Huang dismisses bubble talk

Nvidia posts blowout Q4 results

Q4 revenue hits $68.13 billion… up 73%

An inflection point driven by agentic AI

AI investment translates into revenue

Next-gen 'Vera Rubin' to launch in H2

"Many companies are generating real revenue through investments in artificial intelligence (AI). As a result, computing demand is growing exponentially." (Jensen Huang, CEO of Nvidia)

Nvidia reported fiscal 2026 fourth-quarter (November 2025–January 2026) results that far exceeded consensus estimates (the average market forecast). The company also projected that its fiscal 2027 first-quarter (February–April 2026) results would come in more than 7% above consensus. As the spread of “agentic AI” (AI that plans and executes complex tasks on its own to achieve a given goal) accelerates the race to invest in AI, demand for Nvidia’s AI accelerators also appears to be continuing to climb.

Gross margin at 75% despite a surge in costs

Nvidia said on the 25th (local time) that it posted fiscal 2026 fourth-quarter revenue of $68.13 billion and earnings per share (EPS) of $1.62. Both revenue and EPS topped consensus (revenue of $65.91 billion and EPS of $1.53). Year on year, revenue and EPS rose 73% and 82%, respectively. Gross margin reached 75.2%. The market assessed that “Nvidia has shaken off concerns that margins would fall due to a sharp rise in memory chip prices.”

Guidance for fiscal 2027 first quarter also beat expectations. Revenue guidance came in at $78.0 billion, above consensus ($72.78 billion). Gross margin was guided at 74.9–75.0%.

Agentic AI creates an inflection point

On the post-earnings conference call, CEO Huang focused largely on the explosive increase in computing demand and the next-generation AI accelerator, Vera Rubin, that can meet it. Asked about the likelihood that big tech companies such as Amazon and Google would continue capex worth hundreds of billions of dollars each year, Huang said, “Our customers’ cash flows will continue to improve.” He cited agentic AI as the reason.

Unlike traditional AI that merely answers questions, agentic AI is characterized by proactively carrying out problem-solving. Because it requires multiple stages of computation, demand for tokens (AI’s unit of computation) surges. “We are at an inflection point for agentic AI,” Huang said, adding, “Without computing power you can’t generate tokens, and without tokens you can’t generate revenue.” In other words, AI service providers have little choice but to invest in Nvidia’s AI infrastructure and platforms if they want to turn a profit.

Nvidia plans to respond with its current flagship Blackwell and the Vera Rubin series slated for launch in the second half of the year. Huang stressed, “Rubin samples shipped this week,” adding, “Agentic AI is making waves now, but what comes next will be physical AI such as robotics.”

Market relieved… but bubble concerns persist

The market reaction was closer to “relief” than “cheers.” In after-hours trading following the earnings release, Nvidia was up more than 4% at one point, but turned slightly lower toward the end of the session. The Wall Street Journal (WSJ) said, “Nvidia’s earnings have eased AI bubble fears, but the pressure is mounting to beat Wall Street expectations every quarter.” Calls are also growing that customers must use AI to generate not revenue but “profits.”

The rise of China’s AI industry and a shortage of memory chips are also headwinds for Nvidia. Nvidia said, “We have received approval to sell a small quantity of H200 in China, but no revenue has been generated yet,” adding, “Chinese AI chip designers are closing in fast.”

Reporter Jeong-su Hwang hjs@hankyung.com