Qualcomm Inc. on Monday announced plans to launch custom central processing units designed for data centers to power artificial intelligence workloads, integrating with Nvidia Corp.’s dominant graphics processing units.

The move marks Qualcomm’s strategic re-entry into the competitive data center market, aiming to capitalize on the surging demand for AI infrastructure.

Nvidia’s GPUs are pivotal in training large AI models for applications like chatbots, typically paired with CPUs from market leaders Intel Corp. and Advanced Micro Devices.

Qualcomm’s new CPUs, built to connect seamlessly with Nvidia’s GPUs and software, position the company to tap into Nvidia’s critical role in AI-driven data centers.

This integration is seen as essential for any newcomer vying for a foothold in the sector.

The announcement, made at Computex in Taipei, follows Qualcomm’s acquisition of Nuvia in 2021, a firm specializing in Arm-based processor designs.

Nuvia’s expertise has bolstered Qualcomm’s data center ambitions, which faltered in earlier attempts last decade.

The company now aims to challenge the dominance of Intel, AMD, and custom chips from cloud giants like Amazon and Microsoft.

Qualcomm also revealed a memorandum of understanding with Saudi Arabia’s Humain, an AI firm backed by the Public Investment Fund, to collaborate on data center development.

This deal aligns with a wave of U.S. tech investments in the region, underscoring Qualcomm’s global expansion strategy.

Under CEO Cristiano Amon, Qualcomm is diversifying beyond its core smartphone processor and modem business.

The company has made inroads into automotive chips and personal computers, a market long dominated by Intel.

During the Computex presentation, Amon highlighted that over 85 PC designs featuring Qualcomm’s Snapdragon X Series chips, launched in 2023, are either on sale or in development.

He also teased a new PC chip announcement for Qualcomm’s annual summit in September.

The data center CPU market remains fiercely competitive, with established players and custom solutions from hyperscalers