Qualcomm’s new AI chips challenge Nvidia’s dominance with mobile-level efficiency

Qualcomm announced on Monday that it will launch data center artificial intelligence (AI) chips in 2026, built on the same power-saving technology that powers billions of smartphones, making a major push into AI computing dominated by Nvidia.

The San Diego company’s stock jumped 11% following the October 27 announcement. Investors were cheering Qualcomm’s growth beyond its core smartphone business, which has stalled in recent years.

The AI200 and AI250 chips scale up Qualcomm’s Hexagon neural processing units found in mobile devices. These neural processing units (NPUs) handle AI tasks on phones without quickly draining batteries. Now Qualcomm is applying that efficiency to massive data centers.

“We first wanted to prove ourselves in other domains, and once we built our strength over there, it was pretty easy for us to go up a notch into the data center level,” said Durga Malladi, Qualcomm’s general manager for data center and edge.

Source: Qualcomm

The chips target AI inference —running trained models rather than training new ones. This focuses on a fast-growing market segment where power efficiency matters more than raw training power.

Qualcomm’s AI200 will ship in 2026, followed by the AI250 in 2027. Both are available as accelerator cards or as complete liquid-cooled server racks. Each rack consumes 160 kilowatts of power.

The AI200 features 768 gigabytes of LPDDR memory per card. LPDDR is the low-power RAM used in smartphones. While it offers less bandwidth than traditional server memory, it costs less and uses less energy.

The AI250 takes things further with near-memory computing. Qualcomm claims this delivers over ten times the memory bandwidth of the AI200 while consuming much less power.

Humain, an AI startup backed by Saudi Arabia’s Public Investment Fund, has already committed to deploying systems using up to 200 megawatts of power starting in 2026.

“Qualcomm’s entry and major deal in Saudi Arabia prove the ecosystem is fragmenting because no single company can meet the global, decentralized need for high-efficiency AI compute,” said Joe Tigay, portfolio manager of the Rational Equity Armor Fund.

The company will also sell components separately. Other chipmakers, potentially including Nvidia and AMD, could buy Qualcomm’s server CPUs or other parts for their own systems.

Qualcomm has diversified recently into laptop chips and automotive processors. The smartphone market accounts for most of its revenue, but growth has slowed after losing Huawei as a customer. Apple also plans to replace Qualcomm chips with its own designs.

The data center push puts Qualcomm against Nvidia, which controls over 90% of the AI chip market. Nearly $6.7 trillion will be spent on data center construction through 2030, according to McKinsey. Most of that money will go toward AI computing equipment.