US chipmaker Qualcomm has fired a major shot in the booming AI market, unveiling a new range of data-centre processors designed to challenge Nvidia’s dominance in the field.

Shares in Qualcomm surged more than 11% to US$187.68 – its highest level in over a year – after the company revealed its AI200 and AI250 chips, which will power next-generation servers used for running AI applications.

The first systems are expected to ship in 2026, with the AI250 following in 2027.

Saudi Arabia’s AI firm Humain, backed by the kingdom’s Public Investment Fund, will be Qualcomm’s first customer.

It plans to deploy up to 200 megawatts of computing capacity based on the new chips, marking one of the largest single deals yet for the company’s data-centre ambitions.

The move represents Qualcomm’s most serious push yet beyond smartphones and into the lucrative world of AI accelerators – chips that process the enormous workloads needed to power large language models like ChatGPT and Google’s Gemini.

Nvidia currently controls more than three-quarters of that market.

Qualcomm says its approach focuses on efficiency and memory capacity, with designs derived from its low-power mobile processors.

Each AI250 chip will support up to 768GB of memory and offer what the company describes as a “generational leap” in performance for AI inference tasks, while using significantly less energy.

The company will offer the chips as standalone components, modular add-in cards or complete liquid-cooled rack systems – directly competing with Nvidia and AMD’s multi-chip servers.

Analysts say even a small slice of the US$500 billion AI accelerator market could mean billions in new revenue for Qualcomm as the global race to build AI infrastructure continues to heat up.