Intel will fab NVIDIA CPUs for data centre hardware and x86 system-on-chips (SOCs) that integrate NVIDIA RTX GPU chiplets for personal computers under a landmark new partnership.

NVIDIA is investing $5 billion in Intel under the deal, which comes after Intel Foundry last month reported a $3.2 billion operating loss.

CEO Jensen Huang said the agreement represented “a fusion of two world-class platforms” that “tightly couples NVIDIA’s AI and accelerated computing stack with Intel’s CPUs and the vast x86 ecosystem.”

(A decade ago NVIDIA was trading at less than a dollar per share and Intel was worth ~30 times more. NVIDIA is now a $4 trillion market cap juggernaut and Intel worth far less at a ~$116 billion market cap. The idea that NVIDIA might ride to “chipzilla”s rescue was unthinkable…)

See also: Commercial production of OpenTitan is a RISC-V landmark. What's next?

Lip-Bu Tan, CEO of Intel, said: “Intel’s leading data center and client computing platforms, combined with our process technology, manufacturing and advanced packaging capabilities, will complement NVIDIA’s AI and accelerated computing leadership to enable new breakthroughs for the industry. We appreciate the confidence Jensen and the NVIDIA team have placed in us with their investment…”

Analysts at Futurum Group commented: "For the last decade, CPUs were the tax you paid to plug GPUs into racks. Now NVIDIA is flipping that: the CPU becomes an extension of the GPU fabric. By taking a 5% stake in Intel and co-developing a NVLink-capable Xeon, NVIDIA isensuring that the world’s most entrenched compute base – x86 – becomes natively interoperable with its accelerators.

That has second-order effects across the competitive landscape. Arm has been positioning itself as the architecture of choice for hyperscalers building bespoke AI CPUs, with Grace as NVIDIA’s integrated showcase. But if enterprises can buy Intel chips with native NVLink that preserve x86 compatibility for legacy workloads, a major portion of Arm’s “inevitability” case gets undercut. Arm vendors will be forced to sharpen their pitch around efficiency and total cost of ownership, but in an economy where latency, determinism, and developer compatibility drive adoption, that’s a tougher hill to climb."

They added: "TSMC, importantly, remains central. Every Blackwell-class GPU and NVSwitch that makes NVLink valuable is still fabbed at TSMC. This partnership doesn’t dislodge Taiwan from the stack – it just reinforces NVIDIA’s control over the platform layer while letting Intel regain relevance in the socket layer. The digital economy that’s forming – AI workloads embedded into search, commerce, logistics, finance, andentertainment – will run on racks where the CPU isn’t the star but a bonded participant in NVIDIA’s mesh. That gives NVIDIA the tollbooth, Intel the seat back at the table, and forces ARM and AMD to recalibrate their strategies in a world where the interconnect, not the chip alone, defines who gets paid."

The agreement comes after Tan this summer axed chip factory construction around the world, terminating projects in Costa Rica, Germany and Poland.

The move was part of an aggressive restructuring at the troubled chipmaker where Tan said he is “instituting a policy where every major chip design needs to be personally reviewed and approved by me before tape-out…” 

The semiconductor fab site in Germany was set to be a €40 billion investment, including €10 billion euros in subsidies from Berlin. Like Poland’s $4.6 billion plant, production was pencilled in to start in 2027.

Tan, who was appointed Intel CEO in March 2025, said bluntly on a Q2 call: “I do not subscribe to the belief that if you build it, they will come.”

“Under my leadership, we will build what customers need, when they need it and earn their trust through consistent execution… our factory footprint has become needlessly fragmented,” he told analysts. 

More details and reaction to follow after a conference call this evening. 

Views on the partnership? Get in touch.

The link has been copied!