Forget Smartphones: Qualcomm Just Landed a Massive AI Deal With Amazon

Qualcomm announced an agreement under which Amazon could purchase up to $60 billion of its AI data-center chips, while Qualcomm will provide Amazon with approximately $4 billion in warrants exercisable at $161.26 per share. The deal, following a multigeneration Meta supply agreement, supports Qualcomm's goal of generating more than $15 billion in data-center revenue by fiscal 2029 from a stated $1 trillion addressable market; Qualcomm shares rose 3% on the announcement. The expansion offsets weakening handset exposure, where fiscal Q3 revenue fell 4% to $9.94 billion, net income declined 25% to $2.0 billion, and handset revenue dropped 20%.

Analysis

The equity re-rating case for QCOM depends less on a headline contract value than on whether it can establish independently measurable hyperscaler revenue, production ramps, and gross-margin durability. Hyperscale custom silicon typically carries lower ASPs and customer concentration risk versus merchant handset chips, so the relevant valuation question is whether incremental data-center EBIT offsets handset volatility rather than simply adds revenue. The warrant structure also aligns AMZN with volume adoption but creates dilution and an effective ceiling around the exercise price until purchase milestones are clarified.

Competitive impact is more nuanced than a direct challenge to NVDA. QCOM's most credible wedge is power-efficient inference and connectivity in workloads where total cost per token matters more than training throughput; this pressures CPU incumbents such as INTC and ARM-licensee server ecosystems more immediately than it displaces NVIDIA accelerator clusters. If AMZN adopts QCOM silicon broadly, it could also improve AWS negotiating leverage against AVGO custom-silicon programs and reduce dependence on a concentrated accelerator supply chain, though AWS will likely preserve multi-vendor optionality rather than award a winner-take-all deployment.

The near-term move is vulnerable to a familiar semiconductor failure mode: announced design wins can precede material revenue by 12-24 months, while qualification delays or weak AI infrastructure returns can eliminate expected volume. A credible 1-3 month catalyst is disclosure of product generations, deployment timing, minimum purchase commitments, and whether reported value includes connectivity, cloud spend, or multi-year contingencies. The 6-18 month bull case requires data-center revenue and segment gross margin to inflect without a commensurate increase in R&D and customer-specific engineering expense; failure to show this by FY2027 guidance would argue that the market is capitalizing optionality too aggressively.

Key Decisions for Investors

• Maintain a tactical long QCOM only on confirmation that the commercial arrangement contains enforceable volume commitments and first revenue begins within the next 12 months; target a 15-20% rerating from data-center multiple expansion, with a stop/reassessment if FY2027 guidance does not identify material data-center revenue or handset weakness worsens.
• Express the relative thesis as long QCOM / short INTC over 6-12 months, sized modestly: QCOM has a potentially differentiated inference/connectivity path while INTC remains more exposed to general-purpose server CPU share pressure. Exit if QCOM's disclosed data-center gross margin is materially dilutive or INTC demonstrates sustained server-share stabilization.
• Do not chase AMZN on the announcement alone. Monitor whether AWS commits to deploying the silicon internally versus merely retaining sourcing flexibility; meaningful deployment would be modestly positive for AWS capex efficiency but any benefit is unlikely to move consolidated AMZN earnings near term.
• Set an event-driven alert around QCOM's next earnings call for warrant share count, vesting thresholds, contract duration, and capex/customer-concentration disclosures. Absent these details, treat the announced contract value as an unverified pipeline metric rather than revenue backlog.