Week 13, 2026
The Semiconductor Newsletter
The semiconductor and AI stack is entering a new phase defined by efficiency gains, infrastructure scale, and strategic localization. This week highlights a convergence of cost deflation in large language model inference, new compression techniques that reshape memory utilization, and aggressive capital deployment across photonics, EUV lithography, and advanced materials. At the same time, governments and hyperscalers are reinforcing domestic supply chains and sovereign AI capabilities. From silicon carbide power systems to agentic AI platforms, the industry is moving toward tightly integrated compute, energy, and data architectures that will determine performance, cost, and control in the next decade.
The articles of the week:
1) Gartner Models a Sharp Decline in Frontier LLM Inference Cost Through 2030
2) NVIDIA Frames the AI Stack as a Dual Model Economy of Open and Proprietary Systems
3) Google TurboQuant Targets KV Cache Compression With Near Lossless 3 Bit Efficiency
4) Lumentum Expands Domestic InP Capacity With New North Carolina Laser Manufacturing Site
5) AIXTRON Expands Assembly and Test Footprint in Malaysia to Deepen Access to the South East Asia Equipment Ecosystem
6) Infineon and DG Matrix Position Silicon Carbide Solid State Transformers for AI Data Center Power Conversion
7) Arm Moves Into Production Silicon With AGI CPU for Agentic AI Data Centers
8) Google Expands Quantum Hardware Scope and Sets 2029 Target for Post Quantum Migration
9) Apple Expands U.S. Semiconductor Supply Chain Through New American Manufacturing Program Partnerships
10) Lace Advances Helium Atom Beam Lithography With $40 Million Funding Round
11) Air Liquide Strengthens Taiwan Materials Supply With Advanced Deposition and Etch Facility
12) Rigetti Targets 1,000 Qubit System Deployment in UK With $100 Million Investment
13) South Korea Deploys $166 Million to Scale Domestic AI Chip Startup Rebellions
14) Alibaba Scales Agentic AI Strategy Through Token Based Economics and MaaS Expansion
15) SK Hynix Commits $8 Billion to ASML EUV Systems to Scale Next Generation DRAM and HBM
