Key Takeaways

  • Amazon could purchase up to $60 billion in Qualcomm AI server chips and services over time.
  • Qualcomm is offering Amazon a warrant to acquire roughly $4 billion in Qualcomm stock if purchasing conditions are met.
  • The collaboration expands Amazon’s chip options even as it considers selling Trainium processors beyond AWS.

Reuters reports that Amazon and Qualcomm have established a multi-generation AI data center chip collaboration aimed at inference workloads inside Amazon Web Services data centers. The arrangement could involve up to $60 billion in Qualcomm AI server chip and services purchases over time, while giving Amazon the right to acquire roughly $4 billion in Qualcomm stock through an equity warrant tied to purchasing milestones.

The structure connects customer spending with ownership incentives. Amazon gains a financial interest in Qualcomm if the relationship develops at scale, while Qualcomm receives a potentially large route into hyperscale data centers. Still, the headline figures represent conditional rights and potential purchases, not an immediate $60 billion order or a completed $4 billion equity investment.

For Qualcomm, the collaboration opens another front beyond the mobile and edge markets where its technology is widely known. AI inference, the process of running trained models to generate answers, recommendations or other outputs, represents a large and expanding portion of data center demand. It can also reward different design priorities than model training, including energy efficiency, predictable latency and lower cost per query.

Amazon is not simply replacing its own silicon with Qualcomm processors. AWS already operates a broad custom-chip portfolio that includes Graviton CPUs, Trainium AI accelerators and Inferentia inference chips. Qualcomm’s products could add another architecture to that mix, giving AWS more flexibility as model types, customer requirements and infrastructure economics change. Whether that flexibility outweighs the added software and operational complexity will be a key deployment question.

Amazon is also considering taking its semiconductor ambitions beyond its cloud boundary. Amazon's senior vice president overseeing semiconductor, AI and quantum efforts has confirmed that the company is in talks to sell Trainium AI chips, previously available only through AWS, for use in other companies’ data centers. “We view AI infrastructure as rapidly evolving,” the executive said. “And we're constantly looking at ways to get to more customers.”

That potential expansion places Amazon in two roles at once: a major buyer of external chips and an increasingly assertive chip developer. The senior vice president described Amazon as “one of a very few players who have the ability to design a chip, design the physical attributes of that chip, and then do the production of that chip.” Amazon could become both a Qualcomm customer and a broader merchant-silicon competitor in parts of the market.

The timing reflects extraordinary semiconductor demand. ET CIO reported that global semiconductor revenue reached about $793 billion in 2025, an increase of 21% year over year, with AI semiconductors contributing nearly one-third of sales. Those totals encompass processors, high-bandwidth memory and networking components, illustrating that the AI buildout extends well beyond accelerators alone.

Meanwhile, Network World reported a projection of roughly $1.6 trillion in semiconductor revenue for 2026. The AI data center ecosystem is expected to exceed one-third of semiconductor revenue in 2026 and surpass 50% by 2030. Winning hyperscale business requires more than producing a capable chip. Software compatibility, compiler maturity, availability and system-level performance heavily influence adoption.

Interconnects will matter too. Qualcomm’s roadmap will enter data centers built around technologies such as PCIe, NVIDIA’s NVLink and high-speed Ethernet based on IEEE 802.3 standards. AWS will need to assess how Qualcomm processors behave across servers, memory, networking and orchestration layers, not just on isolated benchmarks. The warrant aligns long-term incentives, but execution will determine whether this becomes a major shift in AI infrastructure or one more option in Amazon’s increasingly diverse silicon portfolio.