Nvidia CEO Says AI Is Already Productive, Not Just Promising

What matters now, he told analysts, is that “AI is now doing productive and useful work” and “generating profitable tokens,” not whether it has cleared some abstract intelligence threshold. He waved off the industry’s fixation on artificial general intelligence entirely, calling AGI benchmarks “kind of senseless at this point.”

Nvidia’s own numbers backed the claim. The company reported record second quarter revenue of $96.2 billion, more than double the prior year period, as demand for AI computing kept outpacing what Nvidia can supply.

Huang pointed to customers across industries already extracting economic value from AI in production. Quantitative trading firms are running it. Chipmakers are running it. Drug developers are running it.

Where AI Is Already Doing the Work

Hudson River Trading and Jane Street are using Nvidia-powered AI factories to accelerate quantitative trading. Samsung Electronics is applying the technology to computational lithography, achieving up to 20 times greater performance than prior methods. Bristol-Myers Squibb is investing in Nvidia’s Vera Rubin AI factory platform to compress drug development timelines from years to months, a fast follow to similar buildouts at Roche and Eli Lilly.

Cybersecurity is another area Huang cited directly. He said a wave of new security companies has emerged that couldn’t exist without frontier AI models, building distributed, continuously running, autonomous defense systems that operate without constant human oversight.

Agentic AI, systems that reason through multiple steps and call outside tools before completing a task, is driving a structural change in how much compute each unit of work requires. Huang told analysts these agentic workloads consume 15 to 100 times more compute than a single human query, depending on the complexity of the problem, a gap he said explains why demand keeps outpacing supply. He said every company will eventually run its own fleet of AI agents working continuously in the background, comparing Nvidia’s roughly 40,000 human employees today to a future where companies operate agents in the hundreds of thousands or millions.

AI Adoption Stretches From Startups to Sovereigns

We’d love to be your preferred source for news.

Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!

Huang said the customer base for AI has broadened well past the handful of labs that drove early demand, describing a golden age of new AI labs, startups and a thriving open source ecosystem. Every country and every startup needs to build its own proprietary intelligence, he said, arguing that open models reaching frontier-level performance has made that possible.

Asked whether rising open-source adoption threatens Nvidia’s growth, since much of current demand comes from closed frontier labs, Huang said the distinction doesn’t matter to Nvidia’s business. Nearly all open models run on Nvidia hardware, he said, because of the reach of its Cuda software ecosystem, which spans PCs and edge devices to robots, workstations and data centers. Closed and open models are both growing simultaneously rather than competing for the same demand.

Huang was also asked how Nvidia can keep investing across the AI ecosystem while major customers, including OpenAI, build their own custom chips. He said Nvidia is building something categorically different: a full platform spanning the entire AI life cycle that runs in any cloud, and he expects frontier labs to remain Nvidia customers for years regardless of their internal chip efforts.

What Else Stood Out

• Enterprise on-premises deployment is growing across sectors, with automotive customers generating $8 billion in trailing 12-month revenue and financial services, manufacturing and healthcare combined contributing another $7 billion.

• Nvidia expanded its partnership with Amazon Web Services, which will deploy an additional 2 million Nvidia GPUs through fiscal 2029 and adopt Nvidia’s physical AI stack, Omniverse, Cosmos, Isaac and Jetson, for its warehouse robotics fleet.
• The company introduced a revenue-sharing structure with neocloud partners, providing a minimum revenue guarantee in exchange for a share of rental revenue above that floor.
• Nvidia disclosed partnerships with six infrastructure capital providers, Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR, to raise more than $500 billion in third-party financing for AI infrastructure buildouts.

Topline Results and Future Outlook

Nvidia reported second quarter revenue of $96.2 billion, up 106% year over year and 18% sequentially. Data center revenue reached $89 billion, up 117% year over year.

The company returned $26 billion to shareholders during the quarter, including $20 billion in share repurchases and $6 billion in dividends, bringing year-to-date capital returns to 60% of free cash flow against a stated target of 50% or more. Balance sheet inventory rose to $32 billion ahead of the Vera Rubin launch, and days sales outstanding increased to 60 days on extended payment terms for large customer orders.

For the third quarter, Nvidia guided revenue to $108 billion, plus or minus 2%, with gross margins expected at 74%, plus or minus 50 basis points. The company said gross margins will bottom in the fourth quarter in a range of 71% to 72% before recovering to 72% to 73% in fiscal 2028. Nvidia’s preliminary outlook calls for fiscal 2028 revenue growth of approximately 70% year over year, which the company described as a supply constrained figure rather than a reflection of total demand.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.