This AI Chip Stock Just Signed Massive Deals With 3 Hyperscalers, and It Still Looks Like a Great Buy Right Now (Hint: Not Nvidia or Intel)

Qualcomm expects its data center chip business to rise from near-zero last year to $5B in FY2027 and $15B by 2029 after signing three major hyperscaler deals. Management projects EPS growing to more than $18 by 2029 (~18% annualized growth), while noting the stock trades at less than 18x forward earnings—implying the opportunity is not fully priced in. Article frames Qualcomm as well positioned to capture incremental CPU demand driven by agentic AI, despite competitive pressure from Intel and Nvidia.

Analysis

The economic prize here is not “AI chips” broadly; it is a shift in hyperscaler capex toward lower-power, workload-specific CPUs for inference and orchestration. That favors QCOM because it can monetize the same architecture trend across data center and edge, but the real P&L leverage will come from mix and scale, not just top-line growth. The market is likely still discounting the fact that direct silicon revenue carries lower margin than licensing, so the near-term earnings contribution may lag the headline revenue ramp even if the design wins are real.

Intel is the most obvious loser because this is another front where incumbency is less defensible than management would like the market to believe. The second-order effect is a bigger procurement shift inside hyperscalers: more multi-sourcing, more custom silicon, and more bargaining power against both CPU vendors and accelerator vendors as customers optimize cost per token. NVDA is not the primary casualty, but if inference workloads keep migrating to specialized CPUs, the growth mix in AI infrastructure could tilt away from pure GPU intensity faster than consensus expects.

The key risk is timing. The positive narrative can trade immediately on contract headlines, but the actual revenue inflection is back-end loaded and could slip if deployment cycles, node availability, or software validation stall. The contrarian view is that the stock may be under-owned, but the market may still be overestimating how much of the 2027-2029 opportunity becomes visible in 2025-2026 EPS; if the next two earnings prints do not show a rising data-center backlog or better gross-margin commentary, the rerating should fade.

Key Decisions for Investors

• Long QCOM on pullbacks over a 3-6 month horizon; thesis is that the market is underpricing design-win optionality and a potential re-rating toward a higher-growth semiconductor multiple. Falsify if management fails to show tangible data-center revenue progression or if gross margin dilutes faster than operating leverage.
• Pair trade: long QCOM / short INTC for 6-12 months. The trade expresses share capture in hyperscaler CPUs while reducing exposure to broad AI-capex beta; risk is Intel regaining credibility on server roadmap or packaging/foundry execution.
• Watch META as a long-duration beneficiary rather than a near-term trade. If custom AI CPUs materially lower inference cost, META’s operating margin could improve, but the 2028 start date means this is more of an alert than an immediate recommendation.
• Avoid chasing NVDA downside on this headline alone. The better expression is relative, not directional: if agentic AI shifts some spend from GPUs to CPUs, NVDA may underperform the semiconductor basket modestly, but training demand still anchors the core thesis.