Intel takes aim at Nvidia with new GPU push

Intel is finally ready to march right into Nvidia’s fortress.

Speaking at the Cisco AI Summit on Tuesday, Intel CEO Lip-Bu Tan announced that the company will start producing graphics processing units (GPUs). It marks a direct push into the chip category powering the AI boom, an area dominated almost entirely by rival Nvidia.

Building the bench from scratch

According to Reuters, the GPU effort will be spearheaded by Kevork Kechichian, the executive vice president and general manager of Intel’s data center group, who joined in September as part of an engineering-focused hiring spree.

Intel also brought in Eric Demers in January. Demers spent over 13 years at Qualcomm, most recently serving as a senior vice president of engineering with deep roots in graphics hardware.

Despite the big names, the effort is in its infancy. Tan noted that the company plans to shape its rollout around customer demand and specific requirements.

Pivoting back to expansion

The move is particularly notable given Tan’s initial marching orders. When he took over as CEO last March, he insisted the chipmaker would consolidate and focus strictly on its core businesses.

While GPUs are still silicon, building high-end accelerators is fundamentally different from churning out x86 CPUs. With data center budgets being vacuumed up by AI accelerators, Intel simply couldn't afford to sit on the sidelines while Nvidia ran the table.

Why it matters

For enterprise customers, another GPU supplier is always welcome. Hyperscalers and AI labs have spent years dealing with Nvidia’s sky-high margins and allocation bottlenecks. Having Intel back in the graphics arena sounds great on paper, but because this effort is just getting off the ground, don't expect viable alternatives to ship anytime soon.

For the industry, taking on Nvidia requires far more than tapping veteran engineers. Nvidia’s real moat isn’t just silicon; it’s CUDA and a developer ecosystem built over two decades. AMD has spent years chipping away at that software barrier with mixed results, and Intel has stumbled on discrete GPU initiatives before.

We think this pivot exposes Intel’s core dilemma. Entering the GPU arena spreads resources thin right after promising consolidation, but ignoring GPUs meant watching the AI infrastructure boom pass by entirely. Starting with experienced hardware leaders and building around customer requests is a sensible way to begin, but turning that into software-ready enterprise silicon will be an uphill battle.