By Jeffrey Dastin

March 18 (Reuters) - ChatGPT defined the technology world in 2023. AI that reasons through problems dominated in 2024, and autonomous agents that code took the mantle in 2025.

For 2026, Nvidia CEO Jensen Huang suggested the next in line this week at the AI chip company’s GTC conference in San Jose: agents that run your computer, design beer for you, DJ for you, and do almost anything else. In a word, it’s OpenClaw, the agent-controlling tech that researcher Peter Steinberger openly released to the world.

Like a rock star in a black leather jacket and shoes to match, chip titan Huang took the crowd home in the SAP Center on Monday with a paean to OpenClaw, nearly two hours into his keynote.

In weeks, OpenClaw achieved the popularity that Linux, an open-source computer project, earned over 30 years, Huang said. It’s like Microsoft Windows for a new age: a personalized operating system of AI agents that act for their users, he said.

The moment’s significance has begun dawning on the world. OpenAI snapped up Steinberger as a new hire, hoping he could produce similar magic at the ChatGPT maker. China’s government saw so big a spike in OpenClaw usage that it warned staff that the tech could leak their data. Shares of companies related to OpenClaw blew up in the Chinese market, with MiniMax Group and Zhipu each climbing around 20% on Wednesday after Huang’s endorsement of the technology.

Huang had his own announcement, too: NemoClaw, a partnership he worked on with Steinberger to bring more security and corporate stateliness to agents for work.

“Every company in the world today needs to have an OpenClaw strategy,” said Huang. “This is the new computer.”

OUR LATEST REPORTING IN TECH AND AI

Exclusive - Meta planning sweeping layoffs as AI costs mount

Exclusive - Nvidia preparing Groq chips that can be sold in Chinese market, sources say

Exclusive - OpenAI courts private equity to join enterprise AI venture, sources say

Exclusive - Amazon CEO sees AI doubling prior AWS sales projections to $600 billion by 2036

A mystery AI model has developers buzzing: Is this DeepSeek's latest blockbuster?

JENSEN'S LAW

Who’s behind the successes of some of the world’s top technology companies, including Google, Amazon, Microsoft, Oracle, CoreWeave, Palantir and Dell?

In Huang’s view, it’s Nvidia.

On a slide in his Monday keynote, the Nvidia CEO showed a layer cake representing Google Cloud’s business, including some of its key AI services. At the top of the cake were Google’s customers, including Salesforce and Puma. At the bottom, underpinning it all, was Nvidia’s chip and software technology.

“Our relationship with cloud service providers are essentially us bringing customers to them,” he said, leaving an impression that they needed Nvidia’s help. “They're always asking us to land the next customer on their cloud.”

Huang developed this into a refrain for each of those big tech companies. Nvidia powered them all.

It powers their internal work, too, such as online search, Huang said.

His message was as much fact as rebuttal to industry skeptics, who recently have doubted how long Nvidia can stay on top.

The company is facing rival efforts by its customers Google and Amazon to bring AI-critical chips to market. AMD, meanwhile, is pushing software to lessen the grip of Nvidia’s GPUs, or graphics processing units.

Perhaps most crucially, OpenAI had found some Nvidia chips to be disappointing for inference, the point at which AI applies what it knows to answer a given prompt, Reuters previously reported.

That could be a problem as the industry has turned its focus toward having AI take longer to answer queries – in other words, doing more inference. Huang said on Monday: “The inference inflection has arrived.”

Huang’s response to skeptics represented the theme of Nvidia’s developers conference this week. He showed off how the company would pair chip technology from Groq, with which it struck a $17 billion licensing deal in December, and Nvidia’s next-generation Vera Rubin chips. Groq’s system would let Nvidia handle more AI tasks no matter the inference required, Huang said.

In a research note, analyst Richard Windsor said Nvidia’s Groq deal underlined how “Nvidia’s lock-in is not nearly as strong in inference.” Even if that changes, the past suggests any boost to Nvidia’s revenue could be hampered by price cuts, Windsor said.

Huang on Monday touted what he called lower cost and better results than any rival. He said Nvidia’s leaps far outpaced Moore’s Law, which heralded a doubling of tech performance about every two years. He then walked up to the image of a wrestling belt, looming large on a slide above his head, and he thrust his fists up to it. It appeared he was holding up his championship prize.

He told reporters the next day: “You are looking at the inference king.”

CHART OF THE WEEK

Software makers long established in the United States are not the only ones challenged by AI’s rapid advances.

In China, too, the old guard exemplified by e-commerce and mobile leaders Alibaba and Tencent are facing stiff competition, with their stocks underperforming peers. MiniMax, founded in December 2021, has now overtaken Baidu’s $42-billion market capitalization.

Behind the disruption are newer players pushing the frontier for China in AI – not just DeepSeek. Knowledge Atlas Technology, also known as Zhipu, is little known outside China, but one research outfit has claimed it has the mainland’s best large language model.

How is the old guard responding? For at least some, the move looks like what Jensen Huang proposed at GTC: embrace OpenClaw, or anything like it.

(Reporting by Jeffrey Dastin; Editing by Rosalba O'Brien)

((krystal.hu@thomsonreuters.com, +1 917-691-1815))