NVIDIA has sold its last shares in Arm Holdings PLC, walking away from a $140 million stake in the British chip designer. The move marks the end of a years-long saga that began with NVIDIA’s $40 billion bid for Arm—only to be blocked by regulators in the UK and EU. Yet despite the divestiture, the two companies are far from finished collaborating.
The sale of 1.1 million shares doesn’t sever NVIDIA’s relationship with Arm. Instead, it reinforces a licensing model that has already shaped NVIDIA’s product roadmap. The company continues to use Arm’s instruction set architecture (ISA) for its Grace and Vera CPUs, which power everything from data centers to embedded systems like Jetson modules for robotics and autonomous vehicles.
This isn’t NVIDIA’s first foray into chip design beyond GPUs. The Vera CPU, unveiled as a standalone server-grade processor, competes directly with AMD’s and Intel’s x86 offerings. Meanwhile, NVIDIA’s Grace supercomputing chips—paired with its own GPUs—rely on Arm-based cores to deliver high-performance computing for AI workloads. Even its embedded platforms, like Jetson Orin, use Arm-based SoCs tailored for edge AI applications.
But here’s the catch: NVIDIA’s GPUs don’t run on Arm. Instead, they use an in-house RISC-V control processor for management tasks, keeping its graphics architecture distinct. This dual approach—licensing Arm for CPUs while developing its own RISC-V IP for GPUs—shows how NVIDIA is hedging its bets in a fragmented chip ecosystem.
Why the Divestiture?
The $40 billion Arm acquisition attempt in late 2020 was the most ambitious deal in the semiconductor industry at the time. Regulators, however, saw it as a threat to competition, fearing NVIDIA’s dominance in AI and GPUs could stifle innovation if Arm’s IP fell under its control. The deal collapsed, and NVIDIA absorbed a $1.25 billion breakup fee—a costly lesson in regulatory scrutiny.
Now, with the remaining shares sold, NVIDIA’s focus shifts to what it can build with Arm, not just own. The company’s recent investments in other tech firms—like CoreWave and Synopsys—suggest a broader strategy of shaping the industry through partnerships rather than outright acquisitions.
What’s Next for NVIDIA and Arm?
- Licensing continues: NVIDIA will keep paying Arm for CPU designs, ensuring its Vera and Grace chips remain compatible with Arm’s ecosystem.
- GPUs stay RISC-V: NVIDIA’s graphics cards will not adopt Arm, sticking to its proprietary control architecture.
- Embedded dominance: Jetson and other Arm-based platforms will remain critical for NVIDIA’s robotics and automotive AI divisions.
- Server competition: The Vera CPU could challenge Intel and AMD in cloud and enterprise markets, where Arm’s efficiency is gaining traction.
The $140 million sale is symbolic—a final chapter on a deal that never was. But for NVIDIA, the real story isn’t what it’s leaving behind. It’s what it’s building next, with Arm as a partner, not a subsidiary.
