Big Tech Q1 capex hits $133B, locking in a depreciation drag that will pressure earnings through 2030. Alphabet surges to a $4.65T valuation on AI cloud demand. Inference constraints are biting as bare-metal providers sell out of AMD silicon, while memory bottlenecks shift toward HBF.
Big Tech spent $133B in Q1 capex, up 70% y/y. Combined depreciation hit $41.6B in the quarter. Server equipment depreciates over 5-6 years, locking in an earnings drag through 2030. Consensus now pegs annual depreciation for MSFT, AMZN, META, and GOOGL exceeding $430B within five years, against $372B of combined net income last year. The useful-life extension lever is fully spent.
Alphabet market cap spiked to $4.65T—closing in on NVDA’s $4.85T—after a massive earnings beat driven by cloud and AI demand. Applied Digital signed a 15-year lease for 300 MW of critical IT load with an IG hyperscaler, adding $7.5B in contracted value. Total contracted revenue for APLD now exceeds $23B.
↳ Goldman Sachs now recommends long hyperscalers, short semis. The bank argues ROI doubt is already priced into cloud multiples, while semi upside is capped if hyperscalers eventually throttle spend to protect cash flow.
OpenAI's GPT-5.5 reportedly reverses the competitive lead over Anthropic's Opus 4.7. Industry signals indicate Anthropic is hampered by compute shortages, while OpenAI claims to have cracked recursive self-improvement for its Codex agent.
The physical compute bottleneck is acutely visible in bare-metal markets. Top AMD providers (Vultr, DO, TW) are entirely sold out of capacity. One provider is raising prices on 8x bare metal boxes due to zero market availability.
↳ Pricing dynamics remain cutthroat for excess inventory. DigitalOcean is pricing H100s at $3.39 and MI300x at $1.99—a ~70% discount—indicating aggressive clearing of older generation nodes against the ramp of next-gen silicon.
Wafer ASPs are inflecting. Supply-demand for epitaxial wafers is tightening faster than expected. Leading-edge logic (7nm and below) wafer demand is modeled to reach nearly 1M wpm in CY28, representing ~10% of total 300mm equivalent demand. GlobalWafers, SUMCO, and Shin-Etsu stand to benefit.
Memory shipment mixes are actively shifting. Commodity memory profitability is currently exceeding HBM, incentivizing the Memory Big 3 to prioritize standard DRAM and NAND production. Korea April semiconductor export data reflects this volume surge: DRAM hit $9.25B (+15% MoM, +83% QoQ) and SSDs hit $3.84B (+20% MoM, +181% QoQ).
↳ UC Berkeley’s David Patterson flags High Bandwidth Flash (HBF) as the next critical compute bottleneck following HBM. TSMC SoIC yields remain low, prompting industry downgrades to near-term 2026 CPO shipment forecasts.
Huawei is rapidly absorbing domestic AI chip share. Huawei AI chip revenue is projected to reach ~$12B this year, up sharply from $7.5B in 2025. Mass production of the 950PR chip is underway, with an upgraded 950DT version planned for Q4.
Morgan Stanley estimates the China AI chip market will hit $67B by 2030, with 86% supplied by domestic vendors. Regulatory friction continues: Nvidia H200 chips are stalled at Chinese customs, unable to clear for domestic transport.
↳ SMIC has successfully produced early 3-nanometer test wafers, indicating slow but steady progress in advanced-node domestic lithography.
Aixtron ($AIXA) reported Q1 revenue of €59m but secured €171m in orders. Management sizes the AI data center optoelectronics opportunity at 80-100 G10 tools per year—roughly a €300-400m annual TAM. Demand for indium phosphide substrates (used in AI optical modules) is accelerating faster than qualified supply.
Teradyne posted strong Q1 numbers ($1.28B rev, 60.9% GM) but guided Q2 revenue down to $1.15-1.25B (midpoint -6% QoQ), signaling a transition period in 2H26.
↳ Advantest is gaining ground over Teradyne in the critical CPO testing space as the industry navigates advanced packaging yield issues.
Apple’s Q2 FY2026 earnings carried positive demand signals for premium consumer electronics but negative read-through for Android OEMs. Memory is not the bottleneck for Apple; it is merely a margin cost headwind.
The true constraint limiting Apple unit shipments is SoC supply, constrained by tight TSMC leading-edge node capacity. Mac mini and desktop Mac growth are gated by fab allocation.
↳ MediaTek received an upgrade to TP5000 (90% upside) by Goldman Sachs. The firm expects next-gen AI ASIC ASPs to be multiples higher, driven by compute-die design integration. MediaTek recently poached TSMC's former packaging VP to de-risk EMIB-T mass production. $TEAM (Atlassian) also surged +20% AH on strong operating results.
Full analysis, implications, relative trade read-throughs, positioning reads and catalysts — delivered by email at 07:00 HKT daily.