TSMC — storyline

Every Silicon Analysts market-intelligence brief on TSMC, newest first — each sourced and dated. The arc, not a disconnected feed.

1. HighPackagingAug 18, 2026

   TSMC CoWoS 5.5x Reticle Hits 98-99% Yield in Volume Production; 14x Roadmap Confirmed to 2029

   TSMC disclosed at the 2026 OCP APAC Summit that its 5.5x reticle-size CoWoS is now in volume production with yields consistently above 98% and reaching 99% across multiple AI customer products — a strong positive signal for advanced packaging supply stability. The company additionally confirmed a technology roadmap extending to 14x reticle-size CoWoS by 2029 and a SoIC hybrid bonding pitch reduction to 4.5 microns, signaling sustained multi-year capacity and capability expansion for AI accelerator packaging.

2. HighPackagingAug 11, 2026

   TSMC CoWoS Yield Confirmed at 98-99% in HVM; Nvidia Weighs 8-Stack HBM4E Fallback for Rubin Ultra

   TSMC disclosed at the 2026 OCP APAC Summit that its 5.5x reticle CoWoS packaging is sustaining yields of 98-99% across multiple AI customer products already in high-volume manufacturing, with a roadmap to 14x reticle size by 2029 — a landmark yield validation that materially de-risks CoWoS supply for H200/B100-class accelerators. Concurrently, Korean press reports that Nvidia is evaluating a downgrade from 12-stack HBM4E to 8-stack HBM4E for its next-generation Rubin Ultra accelerator, citing Samsung and SK Hynix high-stack yield constraints — a signal that HBM stacking yield remains a live bottleneck gating the next product cycle.

3. HighPackagingJul 25, 2026

   Nvidia-SK Hynix & Samsung-Broadcom Lock In $950B HBM Supply Agreements, Securing Next-Gen Memory Pipeline Through 2030

   Nvidia and SK Group have announced a $500B+ strategic initiative that includes a long-term HBM supply partnership with SK Hynix covering HBM4 memory for Vera Rubin-generation AI accelerators, with first AI data center deployment targeted for 2027; separately, Samsung Electronics signed a $200B collaboration with Broadcom spanning memory, foundry, and advanced packaging through 2030. Combined, South Korea's presidential adviser confirmed the two Korean memory giants have committed to $950B in memory supply partnerships with U.S. big tech — $750B from SK Hynix and $200B from Samsung — representing the most significant multi-year HBM supply lock-in ever disclosed publicly.

4. HighPackagingJul 23, 2026

   SK Hynix Commits $5.8B to Advanced HBM Packaging Facility; TSMC Confirms 2027 Wafer Price Hikes Up to 10%

   SK Hynix's board has approved KRW 7.09 trillion ($5.8 billion) in capital expenditure for a next-generation advanced packaging facility in Cheongju, directly targeting HBM packaging capacity expansion to meet accelerating AI accelerator demand. Concurrently, TSMC has confirmed wafer manufacturing price increases for 2027, with mature-node technologies (12nm, 16nm, 28nm) facing hikes of up to 10%, signaling broad-based foundry cost inflation that will flow through to AI hardware BOMs.

5. HighGeopoliticsJul 21, 2026

   China Eyes Export Controls on AI Chips & Models; AMD Helios Rack Enters Market with Azure Deal

   China's Ministry of Commerce is actively consulting with domestic AI firms on tighter export controls covering AI model weights, chip designs, and potential restrictions on overseas foundries—including TSMC and Qualcomm—from producing advanced semiconductors based on Chinese IP, per a Financial Times report cited by Reuters. Simultaneously, AMD has formally launched its Helios rack-scale AI system powered by MI455X GPUs, with Microsoft Azure confirmed as a new deployment customer for H2 2026, introducing a credible rack-level competitive alternative to Nvidia's CoWoS-packaged accelerator dominance.

6. HighMemoryJul 20, 2026

   TSMC Commits $100B+ Arizona Expansion, Accelerates 3nm Conversion to Meet AI Chip Surge

   TSMC has announced an additional $100 billion investment in its Arizona facilities — part of a broader $265 billion U.S. commitment — and confirmed it is actively converting 5nm capacity to 3nm to meet surging AI chip demand, with CFO Wendell Huang citing a 'strong, multi-year demand mega trend.' Separately, Bloomberg analysis of Moonshot AI's Kimi K3 model highlights that efficiency-optimized AI architectures continue to require large memory footprints, reinforcing sustained structural demand for HBM from suppliers including SK Hynix.

7. HighPackagingJul 16, 2026

   TSMC $265B US CapEx Commitment + ASML 30% EUV Capacity Surge Signal Multi-Year AI Supply Expansion — with Equipment Price Hike Risk

   TSMC has committed an additional $100B in US manufacturing investment (total $265B) while confirming 2026 CapEx at the high end of $52B–$56B guidance, directly supporting CoWoS and advanced-node wafer capacity for AI accelerators including Nvidia H200/B200 SKUs. Concurrently, ASML — the sole supplier of EUV lithography equipment critical to HBM and leading-edge logic fabs — raised its 2026 revenue outlook by ~$6–7B to €43–45B and announced 30% annual production capacity expansion for both EUV and DUV systems over the next two years, while separately reporting active negotiations to raise tool pricing, with TSMC reportedly resisting increases on both EUV and DUV equipment.