Qualcomm and Amazon have agreed to jointly develop custom semiconductors for large-scale AI data centers across multiple generations. This agreement could mark Qualcomm’s full-scale entry into the cloud AI infrastructure market, a shift from its traditional focus on smartphone chips.
According to regulatory filings, the price cap linked to Amazon's purchase volume is $60 billionHey. Qualcomm also issued warrants to Amazon that allow it to purchase up to 25 million shares at $161.26 per share. This means the total value of the rights to purchase shares at a fixed price—not just the amount based on the current stock price—is approximately $4 billion.
The two companies are developing not only custom chips for AI inference but also solutions for quickly transferring data within data centers Optical interconnect with speeds of up to 1.6 TbpsLet's develop it together, too. The key point of this announcement is that it brings together computing and connectivity technologies into a single, long-term partnership.
| By jointly developing multiple generations of custom AI chips,
Qualcomm Technologies and Amazon have agreed to develop custom silicon for AWS's large-scale AI infrastructure across multiple product generations. This is not a one-time chip supply contract, but rather a long-term arrangement involving ongoing design and procurement.
The main goal is for the trained model to respond to actual requests AI InferenceWow. Since inference services require a large number of repetitive computations as the number of users increases, not only processing speed but also power efficiency and operating costs are important.
Amazon is already developing its own AI chips. This partnership doesn't mean that all server chips will be replaced with Qualcomm products, but rather, It’s natural to view this as a move by Amazon to secure additional design options and a technology supply chain tailored to its workloads.
| The $60 billion is not confirmed revenue, but a purchase-linked cap.
According to Qualcomm's SEC filing, the amount Amazon pays for server chips, technology, systems, and manufacturing services is up to $60 billionIt says it could reach that amount. This amount is tied to the conditions for the phased vesting of the warrants.
Therefore, It would not be accurate to say that Qualcomm has already secured the full $60 billion in orders. Actual sales will depend on the signing of commercial contracts, binding orders, and the progress of product purchases.
Still, the fact that the maximum amount is very large and applies to products from multiple generations means that This can be seen as a sign that Amazon is preparing for a long-term supply relationship beyond a short-term trial.
| Amazon has been granted the right to purchase up to 25 million shares
Qualcomm has sold up to 25 million sharesper share $161.26We issued warrants that can be exercised to purchase shares. A warrant is the right to purchase shares at a specified price, which is different from actually owning the shares.
This right is finalized in several stages based on the progress of the purchase. At the time of issuance, 3.75 million shares were initially confirmed based on the initial purchase agreement, and The rest is tied to specific contracts, orders, and actual purchase records.
The warrants expire on September 3, 2036. While this provides Amazon with an incentive to grow its supplier base and gives Qualcomm a way to secure long-term orders, there is also a possibility that existing shareholders’ stakes could be diluted if all the warrants are exercised in the future.
| The competition centers on inference performance and power efficiency.
In the AI data center market, the share of inference chips—used to run actual services—is growing rapidly, alongside training accelerators. To provide answers with low latency to a large number of users, It's important to reduce the cost per token and power consumption.
Qualcomm aims to extend its expertise in low-power computing design and system integration—gained through mobile devices—to data centers. Amazon will be able to secure custom chips that eliminate unnecessary features and enhance the processing power required for its services.
However, the product's specific name, manufacturing process, performance, power efficiency, and schedule for the first batch have not yet been disclosed. True competitiveness should be judged based on stability and cost in large-scale operations, not on prototypes.
| Jointly developed optical interconnects with speeds of up to 1.6 Tbps
On AI servers, the speed at which data is transferred between chips and racks is just as important as the computational speed of the chips themselves. If the connection is slow, even installing a large number of high-performance accelerators will increase the overall system wait time.
The two companies will 1.6 terabits per second We plan to develop high-performance optical interconnect solutions and next-generation technologies. Our vision is to leverage Qualcomm's SerDes and optical DSP technologies to support high-bandwidth connections within data centers.
The optical communication technology from Alphawave, which Qualcomm acquired in 2025, also served as the foundation for this expansion strategy. Proposing a combination of computing chips and optical interconnects offers the advantage of addressing data center bottlenecks at the system level.
| Qualcomm is making greater use of AWS in its chip design as well
As part of the collaboration, Qualcomm is focusing on electronic design automation, or EDA We plan to expand our use of AWS's AI infrastructure and Amazon Bedrock in our work. EDA refers to software used to design and verify complex semiconductor circuits.
By utilizing cloud resources as needed, you can process large-scale design verification in parallel and shorten the development cycle. This creates an interdependent structure in which we utilize our clients' infrastructure even during the process of manufacturing the chips we supply.
Qualcomm aims to reduce its reliance on smartphones and boost data center chip revenue by 2029 $15 billion...is our goal. The contract with Amazon will be a crucial test of whether this goal is achievable.
| The true value of a major contract is determined by shipment and operational performance
This partnership enables Qualcomm to position itself in the AI data center market—where Nvidia holds a strong presence—by offering both custom chips and connectivity technologies. It also gives Amazon greater flexibility in choosing suppliers and designs, allowing it to optimize costs and power efficiency.
However, The $60 billion represents the maximum purchase amount over the term of the contract, and the warrants will also be finalized in stages based on actual purchase volume. If there are delays in product development, performance shortfalls, or changes to Amazon's in-house chip strategy, actual orders could fall short of the upper limit.
From now on The key points to verify are the launch and deployment dates of the first chip, the manufacturing foundry, performance and power efficiency, the scope of application for 1.6T optical interconnects, and actual purchase orders. Only once this information is made public will we be able to specifically assess the extent to which this contract will grow Qualcomm's data center business.
Tags
#Qualcomm #Amazon #AWS #AI Chip #Data Center #Custom Semiconductors #AI Inference #Optical Interconnect #1 point, 6 assists #Warrant #Semiconductor Infrastructure #CloudAI #Alphawave
