Amazon and Qualcomm have struck a major multi-generation agreement to develop custom silicon for Amazon Web Services’ growing AI infrastructure, giving Qualcomm one of its biggest opportunities yet to challenge established data-center chip suppliers including NVIDIA.
What happened
Qualcomm announced on September 8 that it will collaborate with Amazon across multiple generations of customized silicon designed for large-scale AI data centers, with a particular focus on AI inference — the process of running already-trained AI models.
The companies will also work together on high-speed optical connectivity for Amazon’s data-center networks, including technology extending to 1.6 terabits per second. Qualcomm says its SerDes and optical DSP technologies will be used as AI infrastructure demands increasingly large amounts of bandwidth.
Reuters reported that the commercial relationship could involve up to $60 billion of Qualcomm products. Qualcomm has also granted Amazon warrants that could allow the company to acquire roughly $4 billion of Qualcomm shares, with the warrants tied to product purchases and other commercial commitments.
Why this matters
The agreement gives Qualcomm a major foothold in a market currently dominated by NVIDIA and other established AI-chip suppliers.
Qualcomm built much of its business around smartphones, but it has been expanding aggressively into data-center computing as demand for AI infrastructure accelerates. Amazon now joins other major cloud and technology companies backing that expansion.
The deal is also notable because Amazon already designs its own custom chips for AWS. Working with Qualcomm gives it another source of specialized silicon and connectivity technology as hyperscalers look for alternatives to relying on a small number of AI hardware suppliers.
Qualcomm is targeting $15 billion in annual data-center chip revenue by 2029, according to Reuters, making this partnership an important test of whether its expertise in efficient computing can translate from mobile devices to massive AI infrastructure.
It is not just about AI processors
One of the more interesting parts of the agreement is the focus on networking. As AI clusters grow, moving enormous amounts of data between processors, memory and servers is becoming nearly as important as the processors themselves.
Amazon and Qualcomm plan to develop optical connectivity reaching 1.6T as part of the collaboration. That could help AWS build larger and more efficient AI systems as demand for inference continues to grow.
The relationship also works in the opposite direction. Qualcomm says it will deepen its use of AWS infrastructure, including Amazon Bedrock, for electronic design automation workloads in an effort to shorten its own chip-development cycles.
What it means for readers
Most consumers will never buy one of these chips directly, but competition in AI data centers affects the services people use every day.
If Qualcomm, Amazon, AMD and other companies can provide credible alternatives to NVIDIA hardware, cloud providers may have more options for controlling the cost and energy requirements of running AI. Over time, that competition could influence the price and availability of generative-AI services, cloud applications and AI-powered products.
For Qualcomm, meanwhile, the Amazon agreement represents another important step in its transformation from primarily a smartphone-chip company into a broader computing and AI-infrastructure supplier.
Sources
Qualcomm — Multi-Generational Product Collaboration with Amazon
