Big Tech's $133B Q1 capex binge locks in a structural earnings drag through 2030—annual depreciation for the four largest hyperscalers is on track to exceed last year's combined net income within five years. Korea DRAM exports surged 83% QoQ to $9.25B in April. Huawei AI chip revenue targets $12B this year as China's domestic substitution accelerates past Western estimates.
Big Tech deployed $133B in capex in Q1 2026, up 70% year-over-year. Combined depreciation hit $41.6B in the quarter alone. Server equipment depreciates over 5-6 years, meaning today's record spend is a locked-in earnings headwind through the end of the decade. Consensus now pegs annual depreciation for MSFT, AMZN, META, and GOOGL exceeding $430B within five years—against just $372B of combined net income last year. Useful lives have already been extended from ~4 to 5-6 years. That accounting lever is spent. Alphabet shares surged after blowout Q1 results, adding a record single-day market cap gain, pushing total valuation to $4.65T—within striking distance of Nvidia at $4.85T.
↳ Goldman's Coviello is explicitly recommending long hyperscalers, short semis. His thesis: hyperscaler valuations already discount capex ROI skepticism via compressed multiples, while chip stocks have been fully rewarded for the cycle. If hyperscalers demonstrate positive ROI on AI spend, they re-rate significantly—while semis face asymmetric downside from any capex deceleration. The question remains whether 40% capital-intensity commodity token businesses will be rewarded by markets over a full cycle.
Preliminary April Korea semiconductor exports are staggering. DRAM: $9.25B (+15% MoM, +83% QoQ). DRAM modules: $6.14B (-17% MoM, +70% QoQ). NAND: $1.67B (-34% MoM, +17% QoQ). SSD: $3.84B (+20% MoM, +181% QoQ). The SSD surge is the standout—nearly 3x quarter-over-quarter. The breadth of the memory recovery is real.
↳ Commodity memory profitability now exceeds HBM, giving the Memory Big 3 stronger incentive to prioritize commodity DRAM/NAND over HBM shipments. This could tighten HBM supply at precisely the moment AI accelerator demand scales. Apple's Q2 FY26 call confirmed memory is not a shipment bottleneck—it's a margin cost variable. The real constraint is SoC capacity at TSMC's leading edge. Separately, Turing Award winner David Patterson flagged High Bandwidth Flash (HBF) as the probable next bottleneck after HBM, with demand surge likely.
Huawei AI chip revenue is projected at ~$12B this year, up from $7.5B in 2025. The Ascend 950PR entered mass production in March; an upgraded 950DT is planned for Q4 launch. Beijing has directed Chinese tech firms to support domestic suppliers and restrict Nvidia chip deployment to China-only operations. Nvidia H200 chips reportedly cannot clear Chinese customs as a result. Morgan Stanley sees China's total AI chip market reaching $67B by 2030, with 86% sourced domestically. This year's domestic market: ~$21B. The substitution velocity is the key takeaway.
↳ SMIC has reportedly produced early 3nm test wafers—signaling faster-than-expected domestic process technology progress. Gulf state alignment remains the geopolitical variable: U.S. leverage over China requires keeping Gulf energy producers and sovereign capital on-side, which implies escalation risk is non-trivial.
AI's impact on silicon wafer demand is no longer a rounding error. Leading-edge logic (≤7nm) wafer demand inflects next year, reaching ~1M wpm in CY28—roughly 10% of total 300mm equivalent demand. Epitaxial wafer supply-demand is tightening faster than expected. GlobalWafers, SUMCO, Shin-Etsu, and Siltronics are the primary beneficiaries. ↳ TSMC can break even at just 25% fab utilization, down from 35-40% historically—revenue outpaced capex, giving them unmatched structural margin discipline.
$MPWR: AI demand broadening from GPUs into CPU servers, optical modules, switches, storage, DDR5, and high-voltage rack architectures. $AXTI: Indium phosphide substrate demand accelerating faster than qualified supply; customers shifting to multi-year allocation agreements; China AI optical ecosystem scaling rapidly. $AIXA: Q1 revenue €59M, orders €171M; optoelectronics TAM sized at 80-100 G10 tools/year (~€300-400M annually). $COHU: Pivoting to AI thermal management and HBM inspection via Eclipse handler. Teradyne: Q1 $1.28B revenue, 60.9% GM beat, but guided Q2 to $1.15-1.25B (midpoint -6% QoQ)—entering transition in 2H26.
Low TSMC SoIC yields are forcing the industry to cut near-term CPO shipment forecasts for 2026. The packaging bottleneck cascades across the optical interconnect chain. Former TSMC packaging VP Douglas Yu has joined MediaTek, reportedly de-risking their first EMIB-T adoption. Industry source: "TSMC assumed customers would not leave, but they did." ↳ Goldman has a TP5000 on MediaTek—90% upside—driven by next-gen AI ASIC projects with ASPs multiple times higher than current gen, plus margin accretion from compute die design involvement.
Bare metal GPU capacity sold out across top providers (Vultr, DigitalOcean, TW). Vultr reportedly moved massive MI355x volume. DigitalOcean pricing H100 at $3.39 and MI300x at $1.99—a ~70% discount reflecting AMD's aggressive capture strategy. Applied Digital signed a 15-year, 300MW lease with a second IG-rated hyperscaler, adding ~$7.5B in contracted value. Total contracted revenue now exceeds $23B, with 50%+ backed by IG tenants. Initial operations mid-2027.
Full analysis, implications, trade read-throughs, positioning reads and catalysts — delivered by email at 07:00 HKT daily.
Full bull/bear breakdown, broker targets, and catalyst calendar — delivered by email at 07:00 HKT daily.