Nvidia unveils next generation of AI chips in bid to entrench market lead

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Nvidia announced the next generation of its artificial intelligence processors on Sunday in a surprise move less than three months after its most recent launch.

At the Computex conference in Taipei, the chipmaker’s chief executive Jensen Huang unveiled “Rubin”, the successor to its “Blackwell” chips for data centres, which are currently in production after being announced in March.

The unexpected move to reveal its next wave of products before Blackwell has even started shipping to customers shows how the world’s most valuable chipmaker is racing to entrench its dominance of AI processors, which has propelled it into the ranks of the world’s most valuable companies.

“A new computing age is starting,” Huang said, as Nvidia also unveiled new AI chip deals with PC makers.

Rubin is set to start shipping in 2026 and promises improved power efficiency, as the Silicon Valley-based company attempts to address concerns that Big Tech’s expansion of AI data centres is putting strain on the energy grid in some regions.

The announcement was light on detail but Huang said Nvidia was working on a “one-year rhythm” of building new AI platforms.

Nvidia’s pace of innovation has taken on outsized importance to the wider stock market, as traders bet on whether the huge AI-driven rally in a handful of Big Tech companies can continue.

The chipmaker added around $350bn to its market capitalisation after it reported surging revenue growth, and the company is closing in on Apple to become the second most valuable US company after Microsoft.

While Nvidia today sells the majority of the AI chips needed to train large language models, such as OpenAI’s GPT, the company faces growing competition from AMD and Intel, as well as bespoke chips developed by cloud computing providers including Amazon, Google and Microsoft.

Nvidia’s Blackwell chip is being rolled out barely a year after its current generation “Hopper” chips were unveiled.

The company on Sunday also announced a new Vera Arm-based central processing unit, as Nvidia seeks to make more of the chips that go into AI data centres.

CPUs, a market dominated by Intel and AMD, are traditionally the workhorse of any computer but Huang is attempting to reshape the server market around its AI chips, as artificial intelligence takes a growing share of data centre workloads.

Nvidia started out more than 30 years ago making graphics processing units, which acted as a sidekick to Intel CPUs in video gaming PCs. But more than 15 years ago, Huang realised that the technology inside its GPUs was also suited to other data intensive computing tasks, such as AI.

The company is now trying to boost its PC chip business by capitalising on its dominance in AI chips for data centres.

Huang on Sunday also announced deals with two PC makers, Asus and MSI, that will launch machines using Nvidia’s GeForce RTX graphics processing units to support a range of AI tasks, from running digital assistants to video editing and coding.

“Your future laptop will be constantly helping you in the background,” Huang said. “The PC will run apps that are enhanced by AI, from writing, photo editing, to digital humans that are AIs,” said Huang.

Nvidia did not specify when the Asus and MSI laptops will go on sale.

A range of PC makers and component providers are expected to use the Computex event to make announcements to position themselves as beneficiaries of an expected “AI PC” wave.

Microsoft recently unveiled a string of AI-enhanced PCs and tablets fitted out with its Copilot assistant tool, powered by Qualcomm’s chips, which will begin to launch later this month. Microsoft has said it expects to include the Nvidia chips and AMD’s Radeon graphics chips in its PCs in future. 

PC sales have slumped since the pandemic but analysts expect that when demand revives companies will increasingly opt for AI PCs embedded with powerful chips to run AI applications, rather than relying solely on the cloud.

“AI PCs will bring the most exciting innovation to the PC industry in the last two to three decades, since the creation of the World Wide Web in the late 1980s,” Morgan Stanley analysts wrote in a note last month.

They said that running AI applications on devices would be cheaper and more flexible than on the cloud and would also have benefits for data privacy. AI PCs will make up about 65 per cent of PC shipments by 2028, up from 2 per cent this year, Morgan Stanley predicted.