Qualcomm hands Amazon $4B of warrants to build custom silicon for AWS

Qualcomm issued Amazon warrants covering more than 25 million shares, valued at approximately $4B, tied to commercial milestones and up to $60B in AWS server-chip purchases under a multi-generation custom-silicon agreement. Qualcomm also signed a memorandum to supply the EU's AI gigafactory program, which has roughly EUR 1B of actual funding. The agreements materially strengthen Qualcomm's position in AI infrastructure and hyperscaler custom silicon, although the warrant dilution and purchase commitments are milestone-dependent.

Analysis

The economic question is not headline contract value but the implied customer-acquisition cost. If the full 25M-share issuance ultimately vests, dilution is roughly 2% of QCOM’s share base; against the maximum purchase commitment, that is a mid-single-digit effective rebate before considering silicon gross margin. The deal is attractive only if it establishes a reusable hyperscaler custom-silicon platform with software/IP pull-through, rather than a low-margin, customer-specific design win that displaces higher-return handset licensing capital.

QCOM’s strategic upside is an expanded valuation framework: credible data-center revenue would reduce its dependence on Android handset replacement cycles and could support a multiple closer to diversified semiconductor peers. The nearer-term beneficiary is AMZN, which gains additional bargaining leverage versus incumbent custom-ASIC and accelerator suppliers, potentially pressuring design-win economics for AVGO and MRVL at future renewals. However, AWS has historically retained substantial control over chip architecture and volume allocation; a memorandum or multi-generation framework is not evidence that QCOM has cleared qualification, yield, software-stack, or rack-scale networking hurdles.

Consensus may overvalue the nominal purchase ceiling before disclosed deployment timing, product scope, and gross-margin terms. Over the next 1-3 months, QCOM can outperform on incremental hyperscaler-design-win speculation, but the 6-18 month catalyst path requires disclosed production ramps and data-center revenue guidance. Falsify the constructive case if QCOM does not identify a production program by its next two earnings cycles, if warrant accounting/dilution exceeds expectations, or if AWS capex shifts toward internally designed silicon without corresponding QCOM content.

Key Decisions for Investors

• Initiate a small long QCOM / short SMH beta hedge over the next 1-3 months; target relative outperformance of 8-12% if management provides production-ramp detail or raises non-handset revenue outlook. Exit if QCOM fails to disclose tangible data-center revenue or milestones by two reporting cycles.
• Do not underwrite the full purchase ceiling into QCOM estimates. Treat the position as an event-driven optionality trade until management discloses annualized revenue, gross margin, and vesting schedule; monitor the next 10-Q for warrant fair value and diluted-share-count impact.
• Watch AVGO and MRVL for a better short-entry signal rather than immediately fading them: initiate a tactical short only if AWS confirms QCOM production allocation or either company flags hyperscaler custom-silicon pricing pressure. The risk is that AWS diversifies suppliers without reducing incumbent content.
• For AMZN, retain a modest long bias rather than chase QCOM’s move: supplier competition can improve accelerator procurement economics and protect AWS operating margins, but the benefit will be visible only through capex efficiency and AWS margin progression over 2-4 quarters.