NVIDIA bets $5B on Intel to build custom chips

Two of the biggest rivals in silicon just pulled off one of the most unexpected alliances in tech history.

NVIDIA and Intel announced a multi-generation collaboration to co-develop custom data center and PC chips. Alongside the partnership, NVIDIA is investing $5 billion into Intel common stock at $23.28 per share, subject to regulatory approvals.

The core of the deal revolves around NVIDIA's NVLink high-speed interconnect, marrying NVIDIA's accelerated computing and CUDA ecosystem directly with Intel's x86 dominance.

Custom silicon for data centers, RTX chiplets inside PCs

The roadmap breaks down into two distinct battlegrounds.

For enterprise data centers, Intel will design and manufacture custom x86 CPUs wired with NVLink. NVIDIA will then take those CPUs, integrate them directly into its proprietary AI infrastructure platforms, and sell the complete systems.

On the client PC front, Intel will build and sell x86 system-on-chips (SoCs) that integrate NVIDIA RTX GPU chiplets. Instead of traditional discrete graphics setups, PC makers will be able to buy an all-in-one x86 processor packed with legitimate RTX firepower.

NVIDIA CEO Jensen Huang called the deal a historic fusion of two world-class platforms driving a new industrial revolution. Intel CEO Lip-Bu Tan emphasized that Intel's process technology, manufacturing, and advanced packaging will complement NVIDIA's AI leadership.

Why it matters

This partnership changes the calculus across both AI infrastructure and personal computing.

For PC buyers and creators, an x86 chip with an integrated RTX GPU chiplet could finally deliver workstation-grade performance without the thermal penalties and footprint of a bulky discrete card. However, this is explicitly a multi-generation effort, meaning you will not see these laptops on store shelves next week. There is no reason to postpone immediate PC purchases.

For the broader industry, this deal lands squarely on AMD's doorstep. AMD previously stood alone as the only vendor capable of pairing high-end x86 computing with high-performance graphics under one roof. Intel and NVIDIA just neutralized that advantage in a single stroke, while also throwing up a massive roadblock to Qualcomm and Arm-based encroachment on Windows PCs.

We think NVIDIA played its hand brilliantly here. While NVIDIA has pushed its own Arm-based Grace processors, the enterprise data center market remains stubbornly chained to legacy x86 workflows. Instead of trying to force customers onto Arm, NVIDIA is simply buying its way into the x86 ecosystem through Intel, keeping AI clusters unified via NVLink.

For Intel, securing a $5 billion endorsement and a guaranteed packaging customer from the most valuable chipmaker on earth is a massive survival victory. The only looming question is execution: both companies have historically preferred total control over their platforms, and maintaining a smooth co-design pipeline across multiple generations is notoriously difficult.