AI boom means your next phone will cost more, warns Nothing boss

Nothing CEO Carl Pei warns that an AI-driven surge in demand for memory chips has broken the long-term downtrend in component costs, with some memory costs reportedly tripling and certain parts rising from under $20 a year ago to potentially over $100 by end-2026. He said the squeeze could make memory the single largest smartphone materials cost, force manufacturers to raise prices by ~30% or cut specs, and shrink entry/mid-tier segments by ~20%; Nothing, valued at $1.3bn after late-2023 fundraising and a $200m September round led by Tiger Global and others, says higher retail prices are inevitable and sees the shift as an opportunity given its design-focused positioning.

Analysis

Market structure: Memory suppliers and cloud/datacenter infrastructure (e.g., MU, NVDA, SMCI, AMZN, MSFT) are clear winners as AI-driven capacity adds turn into multi-year demand for DRAM/NAND; expect memory ASPs to remain elevated with upside of 2x to 3x in pockets cited by industry (6–24 months). Mid/entry smartphone OEMs and retail-led consumer electronics (value brands, EM OEMs) are losers — price elasticity suggests volume contraction of 15–25% in 2026 if OEMs pass costs or cut specs.

Risk assessment: Tail risks include rapid capex-driven oversupply (new fabs online in 12–24 months), US/China export controls or Taiwan geopolitical shocks that could spike prices >50% or halt supply. Near-term (days–weeks) expect sentiment volatility around industry data and quarterly reports; medium-term (3–12 months) depends on spot-price indices and fab capacity announcements; long-term (12–36 months) hinges on structural AI growth vs. cyclical memory capacity additions.

Trade implications: Tactical long semiconductor exposure (DRAM/NAND names and SOXX) and selective longs in AI compute (NVDA) with 6–12 month horizons; short/underweight EM/mid-tier smartphone OEMs and retail-focused consumer ETFs (1–2% positions). Use option structures: buy 6–12 month call spreads on MU/NVDA and buy puts on a mid-tier smartphone basket to limit downside. Rotate away from consumer discretionary into semis and cloud infra within 4–8 weeks.

Contrarian angles: Market may be underestimating speed of capacity add — memory cycles historically flip from shortage to oversupply within 12–24 months (2016–18 DRAM cycle). If AI demand growth decelerates or major fabs announce >15% capacity additions with <18-month lead times, memory longs can become overheated; conversely, a geo-disruption would sharply re-rate suppliers with onshore capacity.

Key Decisions for Investors

• Establish a 2–3% portfolio long in Micron (MU) and a 1–2% long in Nvidia (NVDA) split over 3 tranches over 4 weeks; target horizon 6–12 months, trim if DRAM/NAND spot prices fall 25% from current peaks or if company guidance lowers FY revenue by >8%.
• Initiate a 1–2% short/put position on a mid-tier smartphone basket (e.g., short 1810.HK Xiaomi 1% or buy 3–6 month puts) with strike ~10–15% OTM to hedge against a 15–25% unit volume contraction in 2026; exit if OEMs announce price increases >20% without volume decline.
• Implement options: buy 12-month MU call spread (buy ATM, sell +25% strike) sized to equal 1.5% portfolio risk and buy 6–9 month puts on XRT or a consumer electronics ETF equal to 1% portfolio to protect against retail margin compression over next 6 months.
• Rotate 3–5% from consumer discretionary (XLY/retail names) into semiconductors/cloud infra (SOXX, AMZN, MSFT) over the next 4–8 weeks; monitor memory capacity announcements — if fabs add >15% DRAM capacity with <18-month lead times, reduce semiconductor exposure by 40%.