Qualcomm forecasts $15 billion in data center AI chip sales by 2029, signaling aggressive capture of custom ASIC and ARM server market share.

Qualcomm forecast $15 billion in data center chip sales by 2029, sending shares surging more than 12% in after-hours trading Wednesday. Chief Financial Officer Akash Palkhiwala said at an investor presentation that the data center business will generate $5 billion for fiscal 2027, with $1 billion coming from new custom-chip customers.

Qualcomm also expects $40 billion in revenue from chips outside its smartphone core by 2029, up from previous estimates of $22 billion, with handsets representing just one-third of chip revenue by then. "We will be truly diversified," Palkhiwala said.

The forecast boosted Arm Holdings, which supplies underlying technology for many Qualcomm chips, by 5%. Bank of America analysts had earlier projected more modest revenues of $2 billion to $5 billion annually from Qualcomm's data center push by fiscal 2027-2028.

Qualcomm announced that Microsoft and Meta Platforms will use its new AI chips and has secured two additional unnamed "hyperscaler" customers for custom silicon.

The company's pivot toward AI reflects mounting pressure in the smartphone market, squeezed by a memory chip shortage driven by surging AI infrastructure demand, and major customers like Apple and Samsung developing proprietary in-house chips.

Microsoft will use Qualcomm's new "High Bandwidth Compute" (HBC) category, which uses cost-effective memory chips from smartphones and laptops instead of the expensive high-bandwidth memory used by Nvidia and the SRAM memory used by Cerebras Systems. "That is a tremendous value that we deliver to the industry in terms of performance per cost advantage," said Tony Pialis, Qualcomm's data center chief.

Meta will deploy Qualcomm's Dragonfly C1000 CPU, designed specifically for AI data centers, a market where Arm Holdings and Nvidia are actively courting customers. Pialis said Qualcomm has won two major "hyperscaler" customers for custom chips, with revenue starting before year-end. "I have not had to push my way into hyperscale customers; they've been pulling us in," Pialis said, without naming the customers.

Qualcomm, which has attempted multiple times to build a data center business, is re-entering a fast-growing but intensely competitive AI chip market full of incumbents including Nvidia, Cerebras, Amazon's Graviton, and Google's Axion.

The company said in April it plans to begin shipping data center processors and AI chips by year-end. It is working with customers on three types of chips: central processing units, inference accelerators, and custom application-specific integrated circuits (ASICs)—a segment experiencing strong growth among competitors like Broadcom and Marvell.

AI inference—running trained AI models—has emerged as a critical battleground in data center chip design.