Qualcomm and Amazon's $60 Billion AI Chip Deal, Explained
Qualcomm and Amazon announced a multi-generation partnership on September 8, 2026 that lets Amazon Web Services (AWS) buy up to $60 billion worth of Qualcomm's custom AI chips and networking gear over the coming years. To help seal it, Qualcomm handed Amazon $4 billion worth of stock warrants. It's Qualcomm's biggest signal yet that its bet on AI data centers -- and not just smartphones -- is actually landing with customers, and it gives Amazon a third in-house-ish chip option beyond Nvidia and its own homegrown silicon.
⚡ Quick facts
- Deal size: Amazon can buy up to $60 billion of Qualcomm's AI chips and related products, over multiple chip generations
- Sweetener: Qualcomm gave Amazon warrants for 25 million Qualcomm shares at $161.26 each (~$4 billion), vesting as Amazon actually orders chips
- What's being built: Custom AI inference silicon plus optical networking gear running up to 1.6 terabits per second
- Market reaction: Qualcomm shares jumped more than 4% on the announcement (Sept 8, 2026)
What's actually in the deal
This is a multi-generation collaboration, not a one-off chip order. Qualcomm and AWS will co-develop custom silicon focused on AI inference -- running already-trained AI models at scale, as opposed to training new ones from scratch -- for AWS's hyperscale data centers. Alongside the chips themselves, the two companies will jointly build optical connectivity technology capable of speeds up to 1.6 terabits per second, using Qualcomm's SerDes and optical DSP (digital signal processor) technology. That networking piece matters more than it sounds: as AI clusters scale to tens of thousands of chips working together, the connections between racks can bottleneck performance just as badly as the chips themselves.
Qualcomm CEO Cristiano Amon put it simply: "As AI demand accelerates, data center infrastructure will require advances in both computing and connectivity to deliver greater performance with more efficiency." AWS vice president Prasad Kalyanaraman said the partnership "builds on a strong foundation of partnership" between the two companies and is meant to deliver "more performant, efficient, and cost-effective infrastructure." Neither company has published exact delivery dates or named the specific chip models involved yet -- Qualcomm's existing data-center roadmap includes chips it calls the AI200, AI250, and AI300, alongside a CPU codenamed Dragonfly C1000, so those families are the likely starting point.
The $4 billion in warrants, unpacked
As part of the agreement, Qualcomm issued Amazon warrants to buy 25 million Qualcomm shares at a fixed price of $161.26 each -- worth roughly $4 billion at the time of the deal. These aren't a free handout: the warrants vest as Amazon actually places chip orders, with about 3.75 million shares vesting immediately based on Amazon's initial purchase commitments. In effect, the more chips AWS actually buys from Qualcomm, the more Qualcomm stock Amazon can pick up cheaply -- aligning Amazon's financial upside with Qualcomm's success as a supplier, similar to warrant structures other AI infrastructure deals (like OpenAI's cloud contracts) have used recently.
Why Qualcomm needed this win
Qualcomm has spent years trying to diversify beyond smartphone chips, a business facing pressure from the eventual loss of Apple's modem contract, rising component costs, and slowing handset demand. The company has publicly targeted $15 billion in data-center chip revenue by 2029, but until now that ambition rested mostly on investor-day slides. This Amazon deal is Qualcomm's second major hyperscaler win, following Meta's commitment to use Qualcomm's Dragonfly C1000 chip starting in 2028 production. Analyst Bob O'Donnell of TECHnalysis Research called it "exactly the kind of development that Qualcomm needed to reassure the market" that its data-center push is real. Investors agreed -- Qualcomm's stock rose more than 4% on the news, with rival chipmaker Broadcom's shares ticking up too on the read-through for the broader custom-AI-silicon category.
Why Amazon wants a third chip supplier
AWS already designs its own AI chips in-house -- Trainium for training AI models and Graviton for general-purpose compute -- and that custom silicon business was already running at an annualized revenue rate above $25 billion by mid-2026, growing at triple-digit percentages. AWS CEO Andy Jassy has said OpenAI has committed roughly 2 gigawatts of Trainium capacity and Anthropic up to 5 gigawatts. Adding Qualcomm to that roster -- alongside Trainium and Graviton -- gives AWS another supplier without displacing its existing chip programs, and more leverage in negotiating with Nvidia, which still dominates the AI chip market and can charge premium prices as a result. Qualcomm, in turn, will lean more on AWS's own infrastructure (including Amazon Bedrock) for its chip-design workloads, aiming to shorten its own design cycles.
Why this matters even if you'll never buy an AI chip
Deals like this are what actually decides how much it costs, and how fast, to run every AI product you use -- from ChatGPT to Gemini to whatever app on your phone quietly calls an AI model in the background. More credible chip suppliers competing for hyperscaler business (Qualcomm, Broadcom, AMD, and AWS's and Google's own silicon, alongside Nvidia) generally means downward pressure on the compute costs that get passed on to AI subscriptions and free-tier limits over time. It's also a reminder that the "AI boom" is as much a story about chips, warrants, and multi-year supply contracts between a handful of giant companies as it is about any single chatbot or model launch.
Frequently asked questions
How much is the Qualcomm-Amazon AI chip deal worth?
Amazon can buy up to $60 billion worth of Qualcomm's AI data-center chips and related products under the long-term, multi-generation partnership announced on September 8, 2026. As part of the deal, Qualcomm also issued Amazon warrants to buy 25 million Qualcomm shares at $161.26 each -- worth about $4 billion -- that vest as Amazon actually places chip orders.
What will Qualcomm actually build for Amazon?
Custom silicon for AI inference (running already-trained AI models, as opposed to training new ones) at hyperscale, plus high-speed optical networking gear -- using Qualcomm's SerDes and optical DSP technology -- capable of up to 1.6 terabits per second, to stop networking from becoming a bottleneck as AWS's AI server clusters grow. Qualcomm CEO Cristiano Amon said data-center infrastructure needs advances in "both computing and connectivity to deliver greater performance with more efficiency."
Why does this matter for Qualcomm?
Qualcomm has spent years trying to diversify away from smartphone chips, especially as it faces the eventual loss of Apple's modem business. This is Qualcomm's second major hyperscaler win after Meta agreed to use its Dragonfly C1000 chip starting in 2028, and Amazon's backing gives Wall Street real evidence that Qualcomm's AI data-center push (it's targeting $15 billion in data-center chip revenue by 2029) is more than a slide in an investor presentation. Qualcomm's stock rose more than 4% on the news.
Why does Amazon want Qualcomm as an AI chip supplier?
AWS already builds its own custom AI chips (Trainium for training, Graviton for general compute), and its custom-silicon business was running at over $25 billion annualized revenue by mid-2026. Adding Qualcomm gives AWS a third option and more negotiating leverage against Nvidia, which still dominates AI chip supply. AWS VP Prasad Kalyanaraman said the deal "builds on a strong foundation of partnership" and aims to deliver "more performant, efficient, and cost-effective infrastructure."