At its Open Innovation Partners symposium, TSMC sharply revised its semiconductor industry forecast. In 2024, TSMC projected roughly $1 trillion in worldwide semiconductor revenue by 2030; this year, TSMC North America CEO Sajiv Dalal said the industry will hit about $1.7 trillion by the end of the year, with AI alone contributing more than $1 trillion. That is broadly in line with Gartner's forecast of about $1.6 trillion for 2026, underscoring how quickly AI spending has accelerated.
TSMC's own numbers reflect the surge. August 2026 revenue was up 53.3% year over year, and the January-to-August period rose 39.3%. The company also reported that new tape-outs in the second year of a technology node increased 4x across N5, N3, and N2, signaling a wave of new designs. On the technology front, TSMC detailed power and memory gains: A14 Nanoflex delivers a 14% speed improvement and 23% power reduction versus N2P, while HPC cell power efficiency improves 2.3x from N3P to A12 SPR. By 2029, TSMC projects memory bandwidth increases of 34x for HBM, 115x for 3D DRAM, and 830x for SRAM.
Inference is now the growth driver, with token usage up 500x over three years. TSMC is also pushing interconnect technology through its COUPE roadmap to handle higher serial data rates. AI is feeding back into chip design itself: Broadcom's Greg Dix said OpenAI's Jalapeno custom AI ASIC went from project kick-off to tape-out in just nine months, and argued that the SoC era is ending as SiP accelerates. TSMC added imec as a partner, recognized EDA partners Cadence, Siemens, and Synopsys, and is promoting AI Design Kits to enable agentic AI-driven design flows.