Qualcomm flexes its AI infrastructure muscles

• Chip giant Qualcomm has high hopes for its datacentre-related product lines
• It has unveiled a range of AI infrastructure products designed for the next generation of datacentre facilities
• It is aiming to maximise energy efficiency and token throughput

Qualcomm, best known for its wireless chips, has clearly signalled its intention to play a major role in the AI infrastructure sector with the launch of a range of products and a new strategy that it hopes will take its AI datacentre-related revenues from zero currently to $15bn in FY29, its financial year that ends in September 2029. 

“We are defining Qualcomm’s next chapter as we accelerate our edge diversification strategy, introduce a comprehensive roadmap for next-generation AI datacentres, and evolve into a platform company,” stated president and CEO Cristiano Amon. “Our presence across the entire compute continuum and unparalleled technology capabilities, in low-power computing, AI and connectivity put us in a strong position to capture these opportunities,” he added. 

The company told attendees at its investor day that it has four major datacentre customers already signed up, including Meta (see this announcement) and Microsoft, plus two others that were not identified.

These are in addition to the AI infrastructure agreement struck just over a year ago with Humain, the Saudi AI infrastructure and services giant that is headed up by Tareq Amin.

The push into datacentres is part of Qualcomm’s efforts to reduce its reliance on the sale of wireless device chips by expanding into markets such as internet of things (a target of at least $14bn by FY29) and automotive ($10bn in annual sales by 2029, up from the previously expected $8bn).

The company noted: “Multiple large markets are reaching inflection points as AI compute becomes increasingly distributed across devices, edge and cloud over the next three to five years, including agent-ready edge devices, datacentre infrastructure, automotive, industrial systems, networking and robotics. Together, these represent a combined total addressable market of approximately $1.7tn by 2030. Looking beyond fiscal 2029, Qualcomm sees continued secular growth across datacentre, robotics, ADAS [advanced driver assistance systems] and autonomous driving, industrial AI, personal AI and 6G, with agentic AI expected to drive a new upgrade cycle across intelligent connected devices. This next phase builds on accelerated diversification and proven operating leverage while funding new growth opportunities.”

And to support its overall push into the AI sector, it already announced this week the acquisition of AI-native software stack developer Modular.

The new platforms highlight Qualcomm’s “growing role in building full‑stack datacentre infrastructure optimised for AI, spanning agentic and datacentre‑class CPUs, AI inference accelerators, high‑performance connectivity and at-scale custom silicon solutions,” the vendor noted in this announcement.

The new products are the Qualcomm Dragonfly C1000 CPU, Qualcomm High Bandwidth Compute (HBC), Qualcomm Dragonfly AI300 inference accelerator and connectivity products, as well as custom silicon solutions, all of which have been  “engineered to maximise performance per watt and token throughput at lower total cost of ownership,” the vendor boasted. 

The new CPU launch adds to the company’s existing Dragonfly AI200 and AI250 products and the company expects to launch additional AI datacentre products on an annual basis. There’s a lot more detail. 

“Agentic AI is driving a significant increase in demand for AI inference in the datacentre,” noted CEO Cristiano Amon. “As these become the dominant workloads, infrastructure has to deliver much higher performance at lower power and cost. That plays directly to Qualcomm’s strengths, and we’re well positioned for this shift. With Qualcomm Dragonfly, we’re bringing our high-performance, low-power computing into the datacentre, with multi-year, multi-generation agreements with leading customers,” he added. 

According to seasoned tech sector analyst Richard Windsor, Qualcomm is taking a very pragmatic approach to the non-wireless sectors it is targeting that will help it reach its goals. In his latest Radio Free Mobile blog, Windsor noted that Qualcomm is “doubling down on its long-term approach of offering the entire technology stack but allowing customers to pick and choose the bits that they want, which I think is going to work especially well in the datacentre. 

“Qualcomm is sticking to what it is good at, which is executing on time and offering a horizontal solution. This means that it offers everything one needs to make a digital device, but the customer can pick and choose which bits it wants to buy and which bits it wants to make itself or buy from someone else. Given how the hyperscalers are all working on their own silicon and software, I think that this approach, which has worked extremely well in smartphones and automotive, is just as applicable in the datacentre,” added Windsor. 

Qualcomm says 35 tech and AI “global leaders” are supporting its strategy, including  Arista, Core42, Foxconn, Humain, Lenovo, NEC, Supermicro, Vast Data, Viettel IDC and VNPT Group.

- Ray Le Maistre, Editorial Director, TelecomTV