Copilot+ laptops arrived in force at CES 2026 — just as Windows left NPUs behind

CES 2026 saw major PC OEMs and chipmakers ship Copilot+ capable NPUs—Intel Core Ultra Series 3 (50 TOPS), AMD Ryzen AI 400 series (60 TOPS) and Qualcomm Snapdragon X2 Elite/X2 Plus (80 TOPS)—exceeding Microsoft’s original 40 TOPS Copilot+ threshold. However, Microsoft’s shift to Windows AI Foundry (supporting GPU/CPU inference and replacing the NPU‑centric Copilot Runtime) and limited application support mean NPUs may offer little near‑term differentiation, creating demand and monetization uncertainty for NPU‑focused hardware investments. For investors, the story implies technical progress in silicon but a strategic software pivot by Microsoft that could reduce the premium for NPU hardware and compress the expected addressable market for Copilot+ differentiated PCs.

Analysis

Market structure: Hardware vendors (INTC, AMD, QCOM) are near-term beneficiaries as NPUs (50–80 TOPS) become standard across flagship SKUs, pushing ASPs and replacement cycles into H1–H2 2026. But Microsoft’s Windows AI Foundry reducing NPU-dependency shifts value to GPUs/CPUs (and GPU vendors) and commoditizes NPU differentiation, increasing pricing pressure on NPU premiums and OEM margins within 6–12 months.

Risk assessment: Tail risks include a fast software pivot that re-tightens NPU requirements (re-creating a hardware premium), antitrust scrutiny on bundling AI features, or a 12–24 month inventory glut if software uptake lags; any of these could swing semiconductor revenues ±20–30% vs base. Short-term (weeks) sentiment moves will be modest; medium-term (3–9 months) is where revenue recognition and channel inventory show up; long-term (12–24 months) outcomes depend on developer adoption rates and model-quantization tooling.

Trade implications: Favor semiconductor exposure tied to device ASPs but size positions conservatively—expect 10–25% idiosyncratic volatility. Use directional and relative-value plays (see decisions) that cap downside with defined option structures ahead of expected H2 2026 OEM refresh and Microsoft roadmap updates.

Contrarian angles: Consensus underestimates the risk of NPU commoditization and an inventory cycle—historically similar to smartphone modem overbuilds (12–18 month drawdown). Conversely, if Microsoft’s Foundry accelerates GPU/CPU inference adoption, software monetization could disproportionately reward MSFT over hardware makers; watch developer SDK adoption and Copilot feature telemetry as early leading indicators.

Key Decisions for Investors

• Establish a 2–3% notional long position in INTC via a 6–12 month bull-call spread (buy ATM, sell ATM+10%) to capture hardware cycle re-rating; set profit target +25–35% and stop-loss at -12% (time horizon 6–12 months) given execution risk on multithread gains.
• Initiate a 1–2% notional long in QCOM using 9–12 month calls (25–40% OTM or equivalent call spread) to play 80 TOPS NPU adoption in premium Windows laptops; target +30–50% if Qualcomm wins design wins, stop -15% if OEM share loss is announced within 3 months.
• Reduce MSFT exposure by 1–2% and buy a 3-month put spread (buy 2–3% OTM, sell 6–8% OTM) to hedge against near-term product confusion and execution risk ahead of next earnings; close if MSFT publishes concrete Windows AI Foundry adoption metrics or pledges broader Copilot rollout within 60 days.
• Pair trade: go long INTC (2% notional) vs short MSFT (1.5% notional) for 3–9 months to express hardware re-rating vs software execution risk; unwind if INTC outperforms MSFT by +15% or if MSFT reports developer adoption >25% MoM for Foundry in two consecutive updates.