Qualcomm is expanding deeper into the AI infrastructure market through a multi-generational collaboration with Amazon, targeting custom silicon, AI inference and high-speed optical connectivity for next-generation AWS data centers.

The battle to build the infrastructure behind artificial intelligence is moving beyond Nvidia’s GPUs.

Qualcomm and Amazon have announced a multi-generational collaboration that will see the two companies work together on customized silicon for large-scale AI data centers, while also developing high-bandwidth optical connectivity technologies for Amazon Web Services (AWS).

The agreement represents a significant expansion of Qualcomm’s ambitions in the data-center market. Qualcomm has historically been best known for powering smartphones and connected devices, but the company has increasingly been positioning itself as an AI infrastructure provider across devices, edge computing and cloud data centers.

For Amazon, the partnership provides another route toward expanding the custom computing infrastructure behind AWS as demand for AI inference continues to grow.

The companies announced the collaboration on September 8, 2026, saying they would work across multiple generations of customized silicon designed for large-scale AI inference. The agreement also covers optical connectivity solutions capable of supporting speeds of up to 1.6 terabits per second. (Qualcomm)

The deal comes as hyperscalers and chip companies increasingly look beyond traditional GPUs to build AI systems that can deliver more computing capacity while controlling power consumption and infrastructure costs.

A New Chapter for Qualcomm in AI Data Centers

Qualcomm’s move into data centers is part of a much broader strategy.

The company has spent decades developing low-power processors for smartphones and connected devices. But the rapid growth of AI has created an opportunity to apply that expertise to another problem: how to run increasingly large numbers of AI workloads efficiently.

Qualcomm announced its Dragonfly data-center portfolio in June, including the Dragonfly C1000 CPU, Dragonfly AI300 inference accelerator, High Bandwidth Compute technology and a range of connectivity products.

The company said its strategy is focused particularly on AI inference—the process of running trained AI models to generate responses or perform tasks—rather than concentrating exclusively on model training. (Qualcomm)

That distinction is becoming increasingly important.

Training a large AI model can require enormous amounts of computing power. But once a model is deployed, every user interaction requires inference.

As AI assistants, agents and enterprise applications become more widely used, inference workloads can potentially operate continuously and at enormous scale.

Qualcomm believes its experience with energy-efficient computing gives it an opportunity in that market.

Amazon Is Building Its Own AI Silicon Ecosystem

The Qualcomm deal also fits into Amazon’s existing strategy of developing custom chips for AWS.

Amazon has already invested heavily in its own silicon through its Trainium AI accelerators and Graviton CPUs.

The company’s custom chips have become increasingly important to AWS infrastructure. Amazon reported that Trainium and Graviton had reached a combined annual revenue run rate of more than $10 billion, while Trainium2 had been deployed at large scale and Trainium3 was already handling production workloads. (Amazon)

AWS is therefore not approaching Qualcomm as a replacement for its own silicon strategy.

Instead, the relationship appears to be part of a broader approach in which Amazon uses different computing architectures for different workloads.

The Qualcomm collaboration specifically focuses on customized silicon and AI inference.

That could allow AWS to develop processors optimized around particular workloads, performance requirements and power constraints.

Why Custom AI Chips Matter

For years, the AI hardware conversation was dominated by general-purpose accelerators, particularly GPUs.

GPUs remain extremely important for AI, but the economics of running AI at massive scale are creating demand for specialized hardware.

A custom accelerator can be designed around a particular workload instead of trying to accommodate every possible computing task.

That can potentially improve efficiency.

For cloud providers operating enormous data centers, even relatively small improvements in performance per watt can become significant when multiplied across thousands or millions of processors.

This is one reason companies such as Amazon, Google and Microsoft have developed their own AI chips while also continuing to work with outside semiconductor companies.

Qualcomm’s strategy is aimed at this same opportunity.

The company’s Dragonfly roadmap includes CPUs, inference accelerators, memory technologies, connectivity and custom silicon rather than a single standalone processor. (Qualcomm)

Qualcomm Is Targeting AI Inference

One of the most important elements of the Amazon agreement is its focus on inference.

Modern AI systems increasingly need to perform tasks in real time.

An AI chatbot needs to generate a response when a user asks a question. An AI coding assistant needs to process code and return suggestions. An AI agent may need to reason through multiple steps, search for information and use external tools.

These workloads can generate huge amounts of inference demand.

Qualcomm argues that the shift toward agentic AI will increase the importance of efficient inference infrastructure.

The company has specifically designed its Dragonfly portfolio around what it calls the agentic AI era, with a focus on performance per watt and token throughput. (Qualcomm)

This is where Qualcomm sees an opportunity to differentiate itself.

Instead of competing only on raw processing performance, the company is emphasizing energy efficiency and the cost of delivering AI output at scale.

The AI Infrastructure Race Is Becoming a Systems Race

The Qualcomm-Amazon deal also illustrates a larger transformation taking place in the semiconductor industry.

AI data centers are no longer simply collections of processors.

They are complex systems containing:

CPUs

AI accelerators

High-bandwidth memory

Storage

Networking hardware

Optical interconnects

Power systems

Cooling infrastructure

Software

AI orchestration platforms

A bottleneck in any one of these components can affect the performance of the entire system.

That is why Qualcomm is also working with Amazon on optical connectivity.

The companies say their collaboration will include optical solutions supporting up to 1.6T connectivity for high-bandwidth data-center networks. (Qualcomm)

The significance of this technology is easy to overlook.

AI processors need to communicate with each other constantly. As clusters become larger, moving data between processors can become a major challenge.

Faster interconnects can help reduce that bottleneck.

Why Optical Connectivity Is Important

Traditional electrical connections can become increasingly difficult to scale as data rates rise.

Optical technology uses light to transmit data and can support very high bandwidth over appropriate distances.

Qualcomm has been expanding its presence in this area, including through its acquisition of AlphaWave, which Reuters reported was worth approximately $2.4 billion. (Reuters)

The company’s work with Amazon therefore goes beyond developing processors.

It is also positioning itself as a supplier of technologies that connect AI infrastructure together.

This fits Qualcomm’s broader strategy of providing complete systems rather than individual pieces of silicon.

The Deal Could Be Worth Billions

Reuters reported that the agreement gives Amazon the ability to purchase up to approximately $60 billion worth of Qualcomm AI data-center chips and related products.

The report also said Qualcomm granted Amazon warrants valued at around $4 billion, giving Amazon the right to purchase Qualcomm shares at $161.26 per share. (Reuters)

Qualcomm’s own announcement did not disclose the $60 billion figure, instead describing the relationship as a multi-generational collaboration covering customized silicon and optical connectivity. (Qualcomm)

If the reported purchase opportunity develops at that scale, it would represent a major customer relationship for Qualcomm’s emerging data-center business.

It would also demonstrate how seriously large cloud providers are taking alternatives and complements to conventional AI accelerator architectures.

Qualcomm Wants to Reduce Its Dependence on Smartphones

There is another important reason this agreement matters.

Qualcomm has traditionally generated much of its identity and business around mobile technology.

Snapdragon processors are used across smartphones and other consumer devices, while the company also has substantial businesses in automotive, IoT and connected computing.

But the semiconductor market is changing.

The smartphone industry is mature compared with the explosive growth expected from AI infrastructure.

Qualcomm has therefore been attempting to diversify its business.

In June, the company said it was targeting more than $15 billion in data-center revenue by fiscal 2029 as part of its broader diversification strategy. It also raised its fiscal 2029 non-handset revenue target to $40 billion. (Qualcomm Investor Relations)

The Amazon partnership gives that strategy a potentially significant customer and technology platform.

Qualcomm Is Building a Full AI Stack

The company’s ambitions extend beyond chips.

In June, Qualcomm announced an agreement to acquire Modular, an AI software company focused on developing tools for running AI across different hardware architectures.

Qualcomm completed the acquisition in July.

According to Qualcomm, Modular’s technology provides a unified software environment capable of optimizing AI workloads across CPUs, GPUs, NPUs and custom accelerators. (Qualcomm)

That matters because hardware is only part of the AI infrastructure equation.

Developers need software that can make use of that hardware efficiently.

If Qualcomm can combine its processors with an increasingly capable software stack, it could make its data-center products more attractive to cloud providers and enterprises.

The Amazon relationship could therefore become one part of a larger Qualcomm strategy:

Silicon + networking + software + AI infrastructure.

Amazon Has Plenty of Reasons to Diversify Its AI Hardware

From Amazon’s perspective, developing multiple sources of AI computing capacity can provide flexibility.

AWS operates one of the world’s largest cloud infrastructure businesses.

Its customers increasingly expect AI services to be available at large scale, and Amazon is competing with Microsoft Azure and Google Cloud for AI workloads.

Amazon has responded by investing in its own chips, working with Nvidia and AMD, and developing partnerships with other semiconductor companies.

The company has also been expanding its relationship with Meta.

In April, Amazon announced that Meta would deploy tens of millions of AWS Graviton CPU cores for workloads associated with Meta’s AI systems. Amazon said the relationship reflected growing demand for CPU-intensive workloads associated with agentic AI, including reasoning, code generation, search and multi-step task orchestration. (US Press Center)

The Qualcomm agreement fits into this wider pattern.

AWS is building an increasingly diverse AI computing ecosystem rather than relying on a single architecture.

Nvidia Still Dominates the AI Accelerator Conversation

The Qualcomm-Amazon announcement also arrives during an intense period of competition in AI chips.

Nvidia remains a central supplier of AI accelerators and has expanded its business into CPUs, networking and complete AI data-center systems.

AMD is also expanding from individual chips into broader AI computing systems.

Google has its Tensor Processing Units, while Amazon has Trainium and Inferentia.

Microsoft has developed its own AI silicon, and numerous startups are developing specialized accelerators.

The result is a much more competitive AI hardware market than existed during the early stages of the generative AI boom.

Qualcomm does not necessarily need to replace Nvidia to benefit from this shift.

Its opportunity is to establish a significant role in the workloads where power efficiency, inference economics and custom silicon are particularly important.

AI Agents Could Increase Demand for Inference

The rise of AI agents may make this market even larger.

A traditional chatbot might generate a single response after receiving a prompt.

An AI agent can perform a sequence of operations.

It might interpret an instruction, search for information, call an API, analyze the results, make a decision and then perform another action.

Each additional step can create additional inference workloads.

As organizations deploy more agents, the amount of AI computation required could increase even if the underlying models become more efficient.

This is one reason Qualcomm is focusing so heavily on inference.

The company has described its Dragonfly portfolio as infrastructure designed for an AI era in which agentic systems require large amounts of real-time computation. (Qualcomm)

Energy Efficiency Is Becoming a Competitive Factor

The AI industry’s enormous appetite for electricity is also changing how processors are evaluated.

For traditional computing, performance was often the primary consideration.

For AI infrastructure, data-center operators also have to consider:

Electricity consumption

Cooling requirements

Rack density

Cost per inference

Memory bandwidth

Network bandwidth

Total cost of ownership

Qualcomm is making energy efficiency a central part of its pitch.

Its Dragonfly C1000 CPU, for example, is designed around Qualcomm’s Oryon CPU architecture and is aimed at data-center workloads including agentic AI and general-purpose computing. Qualcomm claims more than two times better performance per watt than certain existing competitive server CPU benchmarks, although those are company-provided comparisons rather than independent measurements. (Qualcomm)

That distinction is important.

As the AI infrastructure market matures, independent benchmarking and real-world deployments will provide a clearer picture of how different architectures compare.

Qualcomm’s Mobile Experience Could Become an Advantage

Qualcomm’s history in mobile computing could also influence its approach to AI data centers.

Smartphones have always operated under strict power constraints.

A mobile processor has to deliver substantial computing performance while working within a relatively small thermal and energy envelope.

Data centers have far more power available, but their scale creates another challenge.

When thousands of processors operate continuously, energy efficiency can have a substantial impact on operating costs.

Qualcomm is attempting to transfer some of the engineering principles developed for power-constrained devices into large-scale AI infrastructure.

The company’s broader strategy also connects devices, edge systems and data centers.

That could become particularly relevant as AI workloads increasingly move between local devices and cloud infrastructure.

Qualcomm Is Building Beyond the Data Center Too

The Amazon deal should therefore not be viewed in isolation.

Qualcomm has been expanding across several areas of AI simultaneously.

Its Dragonwing portfolio targets industrial and edge AI.

Its Snapdragon platforms target personal devices.

Its automotive business is expanding into advanced computing for vehicles.

Its robotics initiatives target physical AI.

Its Modular acquisition strengthens its AI software capabilities.

And now its Amazon relationship provides another route into large-scale cloud infrastructure. (Qualcomm)

The company’s overall strategy is increasingly centered around AI running everywhere—from smartphones and vehicles to factories, robots and cloud data centers.

A Shift From General-Purpose Computing to Specialized AI Infrastructure

The Qualcomm-Amazon relationship is part of a larger transition in computing.

For decades, the industry largely revolved around general-purpose CPUs.

Then GPUs became increasingly important for AI.

Now the market is moving toward highly specialized systems containing CPUs, GPUs, inference accelerators, memory technologies and custom silicon working together.

Cloud providers have strong incentives to optimize those systems because they operate infrastructure at enormous scale.

A few percentage points of improvement in efficiency can become meaningful when multiplied across large data-center fleets.

This is why custom silicon is becoming such an important part of the AI infrastructure race.

What the Qualcomm-Amazon Partnership Could Mean

The immediate impact of the agreement will be measured through actual deployments and products.

Qualcomm will need to demonstrate that its silicon can meet the performance, efficiency, reliability and software requirements of hyperscale data centers.

Amazon will need to integrate those technologies into an environment where customers expect consistent performance and availability.

The companies are describing the relationship as multi-generational, suggesting that this is intended to be a long-term technology partnership rather than a one-off hardware purchase. (Qualcomm)

That makes the announcement more significant than a conventional chip supply agreement.

It represents a strategic relationship between one of the world’s largest cloud providers and a semiconductor company attempting to establish a new position in AI infrastructure.

The AI Chip Race Is Getting Much Bigger

The competition surrounding AI hardware is no longer simply about which company produces the fastest chip.

It is increasingly about who can build the most efficient and scalable system.

Nvidia is developing GPUs, CPUs, networking and complete AI factories.

AMD is expanding its accelerator and CPU portfolio.

Amazon, Google and Microsoft are developing custom silicon.

Qualcomm is bringing its low-power computing expertise into data centers.

Networking and memory companies are developing technologies to solve the growing data-movement problem.

Meanwhile, software companies are working to make AI workloads portable across different hardware architectures.

The Qualcomm-Amazon partnership is another indication that the AI infrastructure market is entering a more diversified phase.

The Bigger Picture

Qualcomm’s agreement with Amazon marks an important step in the company’s attempt to establish itself as a major AI infrastructure provider.

The partnership combines Qualcomm’s expertise in power-efficient silicon with Amazon’s enormous cloud infrastructure footprint.

The companies plan to work across multiple generations of customized AI data-center chips while also developing optical networking technologies capable of supporting increasingly demanding AI systems. (Qualcomm)

For Amazon, the relationship adds another technology partner to an already diverse AI hardware strategy.

For Qualcomm, it provides a path deeper into one of the fastest-growing areas of the semiconductor industry.

And for the wider technology sector, it highlights a fundamental change taking place behind the AI boom.

The future of AI computing is unlikely to be defined by a single processor.

Instead, it will depend on entire systems—processors, accelerators, memory, networking, software and data centers working together.

As AI agents and inference workloads continue to grow, companies capable of delivering those systems efficiently will have an increasingly important role in determining how the next generation of artificial intelligence is built.

For Qualcomm, the Amazon partnership is a major opportunity to turn decades of experience in efficient computing into a new position in the cloud. For Amazon, it is another building block in its effort to create AI infrastructure at enormous scale. And for the broader AI industry, it is another sign that the race for the future of computing is becoming a race between complete ecosystems rather than individual chips.

Sources

Qualcomm’s September 8, 2026 announcement and its Dragonfly data-center roadmap provide the primary technical details of the partnership. Reuters provides additional reporting on the potential financial scale of the agreement and Amazon’s investment in Qualcomm. (Qualcomm)