Big Tech Q1 capex surged 70% y/y to $133B, locking in an annual depreciation treadmill projected to exceed $430B within five years. Alphabet's record market cap surge to $4.65T proves the market is currently rewarding AI revenue realization, while supply chain data points to tightening constraints in leading-edge logic and optical components.
Big Tech poured $133B into capex in Q1, up 70% y/y, with combined quarterly depreciation hitting $41.6B. Server equipment depreciates over 5-6 years, meaning current record spend locks in a massive earnings drag through 2030. Consensus pegs annual depreciation for MSFT, AMZN, META, and GOOGL exceeding $430B within five years, against $372B of combined net income last year. Hyperscalers have already extended useful lives from ~4 to 5-6 years, exhausting that accounting lever.
↳ On the infrastructure side, Applied Digital signed a 15-year lease for 300 MW of critical IT load with a second IG hyperscaler, adding $7.5B in contracted value and bringing total contracted revenue to $23B+.
Apple’s Q2 FY2026 earnings confirm that leading-edge SoC capacity is the primary constraint limiting unit shipments across premium hardware, not memory. Memory is simply a margin headwind; Apple cannot ship enough Mac and wearable products due to extremely tight TSMC leading-edge capacity. The optical supply chain is also tightening rapidly. AXT reported indium phosphide substrate demand is accelerating faster than qualified supply, triggering a shift toward multi-year allocation agreements as China's AI optical ecosystem scales.
↳ TSMC’s low SoIC yields are reportedly forcing the industry to lower near-term CPO shipment forecasts for 2026.
Huawei is set to capture the largest share of China's AI chip market in 2026. Revenue from its AI chips is projected to hit roughly $12B, up from $7.5B in 2025, driven by mass production of the 950PR chip and Beijing's push for domestic alternatives. Morgan Stanley estimates China's AI chip market will reach $67B by 2030, with 86% served by domestic suppliers. In Taiwan, MediaTek sees a 90% upside to TP5000 as next-gen AI ASPs scale multiple times higher.
↳ Compute demand remains insatiable: top bare metal cloud providers (Vultr, DO, TW) are entirely sold out of AMD MI355x clusters, pushing spot pricing premiums and forcing providers to raise prices on 8x bare metal boxes.
Turing Award laureate David Patterson flags High Bandwidth Flash (HBF) as the likely center of the next major AI compute bottleneck, driven by a surge in demand. April Korean semiconductor export data shows massive strength in commodity memory: DRAM hit $9.25B (+83% QoQ) and SSDs hit $3.84B (+181% QoQ). Wafer maker ASPs are turning, with epi wafer supply-demand tightening. Models estimate leading-edge logic demand inflecting to nearly 1M wpm in CY28.
↳ Commodity memory profitability is currently exceeding HBM, giving the Memory Big 3 a stronger near-term incentive to prioritize commodity DRAM and NAND production over HBM shipment mixes.
Monolithic Power Systems highlights AI demand broadening well beyond GPUs into CPU servers, optical modules, storage, and higher-voltage power conversion. The test equipment space is seeing targeted pivots: Cohu is shifting its focus from automotive toward thermal test solutions for AI accelerators and HBM, specifically with its Eclipse handler designed for extreme power density applications.
↳ Teradyne reported a strong Q1 with $1.28B in revenue and 60.9% gross margins, beating expectations. However, the company guided Q2 revenue down to $1.15–$1.25B (midpoint -6% QoQ) as it enters a transition period in 2H26.
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.