Sector Watch
What's moving in genetic research, AI/local inference, and chip manufacturing.
A Complete Human Genome, Reconstructed
Johns Hopkins-led researchers, working with NIST and the Telomere-to-Telomere Consortium, built the most complete diploid human genome sequence yet — full chromosome sets from both parents — closing gaps that could reshape genetic diagnosis and personalized genomics.
A Silicon Chip That Writes DNA
Researchers built a silicon chip capable of writing dozens of DNA sequences in parallel using electricity and water-based enzymes — a cleaner, more scalable alternative to conventional chemical DNA synthesis.
OpenAI and Broadcom Ship a Dedicated Inference Chip
OpenAI and Broadcom unveiled "Jalapeño," custom silicon co-designed around the specific memory movement and serving patterns of large-model inference, aiming for gigawatt-scale deployment starting this year.
Small Models Are Getting Genuinely Capable
Quantization and tooling like llama.cpp and the GGUF format are pushing real inference performance onto consumer hardware, as small language models close the capability gap that used to force everything into the cloud.
TSMC Pushes 2nm Toward 100,000 Wafers a Month
TSMC is accelerating its 2nm process node toward 100,000 monthly wafers by the end of 2026, pulled forward by demand from Nvidia, AMD, and other customers racing for better inference performance per watt.
TSMC Deepens Its Arizona Bet
TSMC added another $100B to its Arizona buildout, bringing total committed investment there to $265B, with leadership pointing to robust AI-driven demand persisting through the end of the decade.