Amazon and Qualcomm AI Chips: Custom Chips for AWS Explained
Amazon and Qualcomm are partnering to develop multiple generations of custom AI chips for Amazon Web Services (AWS) data centers. The collaboration will focus on AI inference and will also include high-speed optical connectivity designed for AI infrastructure.
The agreement is significant for both companies. Amazon is expanding its use of customized silicon for AWS, while Qualcomm is accelerating its move into data-center AI infrastructure beyond its traditional strength in smartphones and other connected devices.
According to Qualcomm and a regulatory filing, Amazon can purchase up to $60 billion of Qualcomm server-chip products, technology, systems and manufacturing services under arrangements covered by the agreement. Amazon has also received a warrant to purchase up to 25 million Qualcomm shares at an exercise price of $161.26 per share, with the vesting of the shares tied to commercial arrangements and purchases.
Amazon and Qualcomm AI Chip Deal at a Glance
| Item | Verified details |
|---|---|
| Companies | Amazon and Qualcomm |
| Primary market | AI data-center infrastructure |
| Chip focus | AI inference |
| Collaboration | Multiple generations of customized silicon |
| Optical connectivity | Solutions extending to 1.6 terabits per second |
| Potential purchases covered by warrant arrangements | Up to $60 billion in payments |
| Qualcomm shares covered by Amazon warrant | Up to 25 million shares |
| Warrant exercise price | $161.26 per share |
| Warrant expiry | September 3, 2036 |
| Qualcomm's fiscal 2029 data-center revenue target | More than $15 billion |
Sources: Qualcomm, Qualcomm SEC filing and Reuters.
What Are Amazon and Qualcomm Building?
The companies have announced a multi-generational collaboration on customized silicon for large-scale AI data centers.
The first major focus is AI inference.
Inference is the stage at which a trained AI model is used to produce an output. When an AI application generates an answer, classification, prediction or other result from a trained model, the computing involved is inference.
Qualcomm says the collaboration will combine Amazon's AI infrastructure with Qualcomm's processing, silicon-design and system-integration capabilities.
The companies will also work on optical connectivity technologies for AI infrastructure, with solutions extending to 1.6 terabits per second.
Why Is Amazon Working With Qualcomm?
Amazon already develops its own silicon for AWS.
AWS has built custom processors and accelerators for cloud workloads, including Graviton, Trainium and Inferentia.
The Qualcomm agreement adds another customized-silicon relationship to Amazon's broader infrastructure strategy.
Qualcomm's announcement specifically describes the agreement as a collaboration to enable customized silicon at scale for large AI data centers, with the companies working together on AI inference.
Reuters reported that Amazon's custom-chip business had reached an annualized revenue run rate of more than $25 billion at the end of the June 2026 quarter.
What Is AI Inference and Why Is It Important?
AI infrastructure is broadly divided into different types of workloads. One of the most important is inference: running a trained model to serve users and applications.
For a cloud provider such as AWS, inference can occur whenever customers use AI models in applications, agents, software, search, automation and other workloads.
Qualcomm and Amazon are specifically targeting this part of the AI infrastructure market rather than announcing a general-purpose replacement for every type of AI processor.
Reuters described AI inference as a fast-growing area and an important competitive market among semiconductor companies.
Why Custom AI Chips Matter for AWS
Custom silicon allows a cloud provider to design hardware around specific workloads and infrastructure requirements.
Amazon's existing silicon strategy demonstrates this approach. AWS describes its custom silicon portfolio as including processors and accelerators designed for different cloud workloads.
The Qualcomm collaboration adds another source of customized server silicon to that ecosystem.
Importantly, the announcement does not say that Amazon is replacing its existing Trainium or Inferentia chips with Qualcomm processors.
It says Amazon and Qualcomm will collaborate across multiple generations of customized silicon for AI infrastructure.
The $60 Billion Figure Explained
The $60 billion figure needs some context.
Amazon is not paying Qualcomm $60 billion upfront.
Qualcomm's SEC filing says Amazon received a warrant to acquire up to 25 million Qualcomm shares. The shares vest in tranches tied to commercial arrangements, binding purchase orders and actual purchases of Qualcomm server-chip products, technology, systems and manufacturing services.
The filing states that those arrangements can involve up to $60 billion in payments during the term of the warrant.
The warrant was issued at an exercise price of $161.26 per Qualcomm share and expires on September 3, 2036.
The filing also states that 3.75 million shares vested upon issuance based on initial purchase commitments.
Amazon Gets a Qualcomm Stock Warrant
The stock warrant is an unusual part of the agreement.
Amazon received the right to purchase up to 25 million Qualcomm shares at $161.26 per share.
However, the vesting is connected to the commercial relationship.
According to Qualcomm's filing, the warrant shares vest in tranches based on commercial arrangements, binding purchase orders and actual purchases of Qualcomm products and services.
This means the equity component is linked to the business relationship rather than being an unrelated investment in Qualcomm.
The warrant expires in 2036 and does not provide voting rights while it remains unexercised.
Why Is Qualcomm Moving Into AI Data Centers?
Qualcomm has historically been strongly associated with mobile processors, connectivity and smartphones. The company is now pursuing a broader strategy across data centers, automotive, industrial systems, robotics and other computing markets.
In June 2026, Qualcomm announced a target of more than $15 billion in data-center revenue by fiscal 2029.
The company also raised its fiscal 2029 non-handset revenue target to $40 billion.
Qualcomm said its data-center strategy would cover AI infrastructure and other technologies across the compute continuum.
The Amazon collaboration therefore arrives as Qualcomm is actively trying to build a large data-center business.
Qualcomm's Data-Center Strategy
Qualcomm's current data-center portfolio extends beyond a single processor.
The company describes its Dragonfly platform as covering AI accelerators, high-bandwidth memory, connectivity, custom silicon and infrastructure-management software.
Its data-center products include rack-scale AI inference platforms and connectivity technologies designed for 800G and 1.6T optical applications.
Qualcomm also describes custom silicon as a part of its data-center strategy for domain-optimized solutions.
These are Qualcomm's stated product and strategy positions; they should not be interpreted as guarantees of future performance or market share.
Why Optical Connectivity Is Part of the Deal
The partnership is not limited to AI processors.
Amazon and Qualcomm will also collaborate on high-speed optical connectivity.
Qualcomm says the solutions will extend to 1.6 terabits per second.
This addresses a separate part of AI data-center infrastructure: moving data between components at high bandwidth.
As AI systems use larger numbers of processors and increasingly distributed infrastructure, networking and interconnect technology become important parts of data-center architecture.
Qualcomm's announcement specifically connects the optical collaboration with the growing bandwidth requirements of AI infrastructure.
Qualcomm Will Also Use AWS for Chip Design
The relationship works in both directions.
Qualcomm said it plans to expand its use of AWS AI infrastructure, including Amazon Bedrock, for electronic design automation workloads.
The stated objective is to reduce chip-design cycles.
This means AWS is not only a potential deployment environment for the resulting technology. Qualcomm will also use AWS infrastructure as part of its own chip-development workloads.
Qualcomm disclosed this as part of the expanded collaboration announced on September 8, 2026.
How This Fits Into Amazon's AI Chip Strategy
Amazon's AI infrastructure strategy already includes custom silicon.
AWS offers Trainium for AI training workloads and Inferentia for AI inference workloads, alongside its other infrastructure processors.
The Qualcomm collaboration should therefore be viewed as an expansion of Amazon's customized-silicon ecosystem rather than a statement that AWS will use only Qualcomm chips.
Reuters reported that Amazon's custom-chip business had grown to an annualized revenue run rate above $25 billion at the end of June 2026.
Amazon, Qualcomm and Nvidia
Nvidia remains a major supplier of AI accelerators and infrastructure to the industry.
However, Amazon has been developing its own silicon for years, and other hyperscalers are also developing customized AI hardware.
The Qualcomm agreement adds another company to this broader custom-silicon ecosystem.
It would be inaccurate to describe the announcement as Amazon replacing Nvidia across AWS.
The announcement instead confirms that Amazon and Qualcomm will develop customized silicon for specific AI infrastructure requirements.
Reuters described the partnership as part of a broader effort by cloud providers to develop alternatives to Nvidia's dominant processors.
Other Companies Are Also Developing Custom AI Silicon
| Company | Known AI silicon strategy |
|---|---|
| Amazon Web Services | Trainium and Inferentia custom AI accelerators |
| Tensor Processing Units (TPUs) | |
| Microsoft | Custom AI infrastructure chips including Maia |
| Meta | Custom AI accelerator development |
| Qualcomm | Dragonfly AI infrastructure and custom silicon |
| Nvidia | General-purpose AI accelerators, networking and AI software infrastructure |
| AMD | Instinct AI accelerators and related data-center products |
The important distinction is that these companies do not all follow the same hardware strategy or compete for exactly the same workloads.
What Qualcomm Gets From Amazon
For Qualcomm, the partnership provides a large hyperscaler relationship at a time when the company is trying to expand its data-center business.
Qualcomm's own fiscal 2029 target calls for more than $15 billion in data-center revenue.
Amazon's potential purchases therefore provide a significant commercial opportunity if the purchase commitments and subsequent deployments develop as described in the agreement.
Reuters reported that Qualcomm expects the Amazon relationship to contribute to its broader data-center revenue ambitions.
What Amazon Gets From Qualcomm
Amazon gets access to Qualcomm's experience in processor design, power-efficient computing, connectivity and semiconductor systems.
The companies will work together on customized silicon rather than simply using an off-the-shelf processor.
The collaboration also includes optical connectivity and Qualcomm's use of AWS infrastructure for chip-design workloads.
These elements make the agreement broader than a conventional chip-supply arrangement.
What This Means for AI Data Centers
The deal illustrates how AI data-center infrastructure is becoming more specialized.
AI infrastructure includes more than compute processors.
| Infrastructure layer | Role |
|---|---|
| AI accelerators | Run AI workloads |
| CPUs | Handle general-purpose computing and system tasks |
| Memory | Stores and supplies data to processors |
| Networking | Connects servers and accelerators |
| Optical connectivity | Moves data at high bandwidth |
| Software | Allows AI models and applications to use the hardware |
The Amazon-Qualcomm agreement covers multiple layers, particularly customized compute and optical connectivity.
What Is the Biggest Commercial Question?
The biggest question is not whether Amazon needs AI infrastructure. Amazon is already investing heavily in custom AI infrastructure.
The question is how much of that infrastructure will ultimately use Qualcomm technology and how quickly the associated purchases develop.
The SEC filing ties the warrant's vesting to commercial arrangements, purchase orders and actual purchases.
That means the maximum $60 billion figure should be treated as a potential amount associated with the agreement rather than guaranteed revenue already booked by Qualcomm.
This distinction is important when evaluating the financial significance of the deal.
What Happens Next?
The companies describe the relationship as a multi-generational collaboration.
That means the agreement is intended to extend beyond one chip generation.
The announced areas include:
- Customized AI data-center silicon
- AI inference
- High-speed optical connectivity
- Future-generation connectivity solutions
- AWS infrastructure for Qualcomm's chip-design workloads
Qualcomm says its optical roadmap includes solutions extending to 1.6T, while its broader data-center strategy includes custom silicon and rack-scale AI inference platforms.
Why the Deal Matters for Qualcomm
Qualcomm has set an ambitious data-center revenue target while simultaneously trying to diversify its business beyond handsets.
The company expects data-center revenue to exceed $15 billion by fiscal 2029.
The Amazon relationship gives Qualcomm a major hyperscaler collaboration to support that strategy.
However, the announcement itself does not establish that Qualcomm will achieve its $15 billion target. That remains a company target rather than a guaranteed outcome.
Why the Deal Matters for Amazon
For Amazon, the agreement adds another technology partner to its custom-silicon strategy.
AWS already has internally developed AI accelerators, and its custom-chip business has become a significant part of its cloud infrastructure business.
The Qualcomm collaboration expands the number of silicon-development relationships supporting AWS infrastructure.
The agreement also gives AWS another commercial relationship in the AI hardware ecosystem while Qualcomm becomes a customer of AWS infrastructure for chip-design workloads.
Final Takeaway
Amazon and Qualcomm are not announcing a simple chip purchase.
They are establishing a multi-generational collaboration around customized AI data-center silicon, AI inference and optical connectivity.
The financial structure is also significant: Amazon received a warrant for up to 25 million Qualcomm shares, with vesting tied to commercial arrangements and purchases that can involve up to $60 billion in payments.
For Qualcomm, the deal supports its strategy to build a large data-center business.
For Amazon, it expands the company's customized-silicon ecosystem.
For the wider AI infrastructure market, the announcement is another example of cloud providers working with semiconductor companies to develop hardware optimized for specific AI workloads.
The most important number to watch from here is therefore not the headline $60 billion.
It is the actual volume of Qualcomm server-chip products and related technology that Amazon eventually purchases and deploys.
That will determine how large this partnership becomes in practice.
Key Takeaways
- Amazon and Qualcomm announced a multi-generational collaboration for customized AI data-center silicon.
- The initial focus includes AI inference.
- The companies will also work on optical connectivity solutions extending to 1.6 terabits per second.
- Amazon received a warrant for up to 25 million Qualcomm shares.
- The warrant has an exercise price of $161.26 per share.
- The warrant expires on September 3, 2036.
- Vesting is tied to commercial arrangements, binding purchase orders and actual purchases.
- The arrangements covered by the warrant can involve up to $60 billion in payments.
- Qualcomm is targeting more than $15 billion in data-center revenue by fiscal 2029.
- Amazon's custom-chip business had an annualized revenue run rate above $25 billion at the end of June 2026, according to Reuters.
- Qualcomm will expand its use of AWS AI infrastructure, including Amazon Bedrock, for chip-design workloads.
Sources
- Reuters — September 8, 2026: Reporting on the Amazon-Qualcomm AI chip partnership, the potential $60 billion purchase relationship, Amazon's warrant and the AI-inference focus.
- Qualcomm — September 8, 2026: Official announcement covering customized silicon, AI inference, optical connectivity up to 1.6T and Qualcomm's use of AWS infrastructure for EDA workloads.
- U.S. SEC / Qualcomm Form 8-K: Official filing covering the 25 million-share warrant, $161.26 exercise price, September 2036 expiry, vesting conditions and $60 billion payment threshold.
- Qualcomm Investor Day — June 24, 2026: Official fiscal 2029 data-center revenue target of more than $15 billion and broader diversification strategy.
- Qualcomm Data Center: Official information on Qualcomm's Dragonfly AI infrastructure, custom silicon, inference platforms and 1.6T connectivity roadmap.