Qualcomm issues warrants to Amazon to acquire $4 billion worth of chipmaker's stock as part of AI infrastructure deal
By Gammatek ISPL
Last updated: September 9, 2026 | 12 min read
Why This Matters
Every hyperscaler's AI ambitions run into the same wall eventually: Nvidia's chips are the best available, but Nvidia can't be everyone's only supplier forever, and paying Nvidia's margins at hundreds-of-billions-of-dollars scale is a problem every major cloud provider wants to solve. Amazon just made its clearest move yet to solve it with a second supplier — and structured the deal so Qualcomm's incentives and Amazon's are tied directly together through equity, not just a purchase order. If you follow AI infrastructure spending, chip stocks, or cloud computing strategy, this is one of the more structurally interesting deals of the year, because of how it's built, not just its size.
What Actually Happened
Qualcomm disclosed in a filing with the Securities and Exchange Commission that it issued Amazon a warrant to acquire 25 million shares of Qualcomm stock at $161.26 per share — a stake worth roughly $4 billion at issuance. Qualcomm's stock rose about 4% on the news. The warrant itself doesn't expire until September 2036, and the shares vest in tranches tied to specific commercial milestones, not all at once.
The real number underneath the headline is $60 billion — the potential value of AI data-center chips and related technology Amazon could purchase from Qualcomm under the terms tied to the warrant. That figure is a ceiling on potential future purchases, not a signed, guaranteed order — the shares vest as Amazon actually buys.
The technical scope of the partnership centers on two things: custom AI inference silicon (chips optimized for running already-trained AI models, as opposed to training them from scratch) and high-performance optical connectivity — physical networking technology capable of moving data at up to 1.6 terabits per second between servers, which becomes a bottleneck at the scale AWS operates. Qualcomm will also lean more heavily on AWS internally, including using Amazon Bedrock for its own chip-design workflows, making this a genuinely reciprocal arrangement rather than a one-directional vendor deal.
Why Qualcomm, and Why Now
Qualcomm has spent years being defined almost entirely by smartphone chips — a business facing real headwinds, including the eventual loss of its modem relationship with Apple as Apple moves toward in-house modem chips. Data centers represent Qualcomm's most serious attempt yet to build a second major growth pillar. The company has previously said it's targeting $15 billion in data center sales by fiscal 2029, and in June introduced a new CPU (Dragonfly C1000) built for what the industry calls "agentic AI" workloads — AI systems that take multi-step actions rather than just answering single queries. Meta has already said it plans to use that chip starting in 2028.
The Amazon deal gives Qualcomm something it didn't have before in this market: a named hyperscaler commitment structure with real financial teeth, at a moment when the company needs to prove to investors that its data-center ambitions are more than a slide in an earnings presentation.
For Amazon, the logic is different but complementary. AWS's AI infrastructure capital expenditures — like Meta's — are running into the hundreds of billions of dollars annually. Every hyperscaler currently faces the same strategic problem: heavy reliance on Nvidia GPUs creates supply risk, pricing power imbalance, and limits on customization. Amazon already has its own in-house chip efforts (Trainium and Inferentia), and this deal effectively adds a third leg to its silicon strategy — in-house chips, Nvidia GPUs, and now customized Qualcomm silicon — reducing dependence on any single vendor.
How This Fits Into the Broader AI Chip Landscape
Nvidia | AMD | Intel | Qualcomm | AWS in-house (Trainium) | |
Primary role today | Dominant AI training/inference GPU supplier | Emerging GPU alternative | Contesting the market, historically weaker in AI | New entrant, inference-focused | Amazon's own custom silicon |
Relationship to this deal | The incumbent this deal is explicitly positioned against | Also contesting Nvidia's position | Also contesting Nvidia's position | Subject of this deal | Complementary — Amazon diversifying across all fronts |
Focus area | Training + inference, broad workloads | Training + inference | Broad, historically server CPUs | AI inference + optical connectivity | Training + inference, AWS-specific |
This is a pattern worth naming clearly: no major hyperscaler wants to be single-sourced on AI silicon. Microsoft, Google, Meta, and Amazon are all simultaneously building in-house chips, signing deals with multiple external suppliers, and hedging against Nvidia's pricing power — this Qualcomm deal is Amazon's most visible move in that direction so far in 2026, but it's part of an industry-wide trend, not an isolated event.
What's Genuinely Uncertain Here
Original analysis worth adding on top of the news coverage: the framing of "$60 billion deal" appearing in headlines is doing more work than the underlying structure supports. That figure is the ceiling of potential purchases tied to the warrant's vesting schedule — not a signed purchase order, not booked revenue, and not a guarantee AWS will actually spend anywhere close to that amount. Companies structure warrant deals this way specifically because it lets both sides announce an impressive top-line number while keeping the actual financial commitment flexible and contingent.
The more meaningful signal isn't the dollar figure — it's the structural choice. Tying equity vesting directly to purchase milestones means Qualcomm's incentives are now genuinely aligned with actually delivering chips AWS wants to buy more of, rather than a one-time announcement with no ongoing accountability. That's a more durable form of partnership than a standard supply contract, and it's worth watching whether other chipmakers start proposing similar equity-linked structures to win hyperscaler business going forward — Intel and AMD both have incentive to consider comparable deals with Google or Microsoft if this arrangement performs well for Qualcomm.
Implementation Consideration: What This Signals for Enterprise AI Buyers
For any organization currently planning AI infrastructure spend — not just hyperscalers — this deal is a useful data point: the AI chip market is entering a genuinely multi-vendor phase, not staying Nvidia-only indefinitely. That has practical implications for procurement timelines and negotiating leverage. A year ago, most enterprise AI infrastructure conversations assumed Nvidia GPU allocation was the primary constraint; going into 2027, buyers evaluating cloud AI infrastructure providers should increasingly ask which chip architectures a given cloud offers, since diversified backends may translate into better availability and pricing for end customers over time — even if that trickle-down effect is still early and unproven.
The Bottom Line
Qualcomm's deal with Amazon isn't the single event that ends Nvidia's dominance in AI chips — Nvidia remains the overwhelming leader in AI training workloads specifically. But it is one of the clearest signals yet that the "second half" of the AI infrastructure buildout — inference at massive scale, optical networking, multi-vendor chip strategies — is where the next round of competition is actually happening, and where companies like Qualcomm are placing their bets on relevance for the next decade. https://www.gammateksolutions.com/ https://www.gammateksolutions.com/post/fortinet-cyber-security-pricing-2026-firewall-cost-guide https://www.gammateksolutions.com/post/2026-price-comparison-of-hci-hyper-converged-infrastructure-solutions https://www.gammateksolutions.com/post/america-must-learn-ai-lessons-from-astro-boy
Comments