Global semiconductor revenue hits $425bn for Q2, Omdia claims - Tech Digest

The Q2 2026 semiconductor revenue milestone of $425 billion represents far more than a cyclical upswing—it signals a structural repricing of the industry's growth model. The 31.4% quarter-over-quarter... Read original →

The $425bn Q2 print is not a cyclical bounce—it is a structural break in how the industry prices growth. Memory crossing 50% of total revenue is the visible symptom; the underlying force is HBM's gravitational pull on wafer capacity, compelling Samsung, SK Hynix, and Micron to cannibalize consumer DRAM lines for AI-grade stacking. The sharper signal sits in non-memory: logic and MPU growth above 10% sequentially (versus a 3% historical norm) confirms AI has escaped the "memory narrative" and become a full-stack demand shock across interconnect, processors, and networking silicon. The binding constraint is not demand but supply elasticity. ASML's EUV delivery cadence and CoWoS/SoIC advanced-packaging throughput are the true bottlenecks. History is instructive: the 2000 dot-com unwind and the 2022 inventory correction both followed AI-adjacent capex spikes. If 2027 sees concentrated capacity release, expect a 30%+ destocking cycle. Competitively, TSMC's 2nm yield advantage over Samsung will anchor logic pricing power for at least two more generations. Custom ASICs (TPU, Trainium) will likely capture 25%+ of AI inference workloads within 18 months, compressing the addressable market for general-purpose GPUs. Over the next 24 months, "advanced packaging is the new node" becomes the industry's operating assumption, and edge-AI inference emerges as the second growth vector. The decisive question: who locks in capacity before the supply-elasticity window closes owns the next cycle's pricing power.