Synthesis history

No new themes this pass. The five new items are secondary coverage — additional articles and social posts on the Jalapeño announcement and Etched's stealth exit — with no new claims, quotes, or stances in their structur…

The substantive additions this pass are Etched's disclosed business metrics — $800M raised and over $1B in customer contracts [^39760] — and the detail that the company recruited 400+ engineers from NVIDIA, TSMC, and ot…

Jensen Huang's direct public characterization of competitor custom ASICs as 'science projects' [^38766] is the sharpest NVIDIA statement on competition in this thread and updates the Huang/NVIDIA perspective. Etched's f…

Etched, an LLM inference hardware startup, emerged as a new actor: Andrej Karpathy publicly praised its cluster-scale memory (CSM) architecture for extreme low-voltage engineering targeting tokens-per-watt at interactiv…

New items this pass are almost entirely secondary amplification of the June 24 Jalapeño announcement — empty syndication articles, YouTube coverage, and social posts with no new substantive claims. Item 34747 (Semafor, …

All new items are secondary amplification of the June 24 Jalapeño announcement — Reddit threads, YouTube coverage, and social media posts with no new substantive claims or independent benchmarks. One minor framing addit…

New items this pass are almost entirely secondary coverage and social media amplification of the already-known June 24 Jalapeño announcement — no new structural developments, no independent benchmarks, and no new substa…

The major development this pass is OpenAI and Broadcom announcing 'Jalapeño' on June 24, 2026 — OpenAI's first custom LLM inference chip, adding a new and structurally distinct voice to the thread: a major AI model oper…

Item 32547 adds a concrete, quantified example of the CUDA software moat: GB200 NVL72 serving costs dropped 2.5x in under 70 days through CUDA kernel rewrites alone on the Kimi/Cursor Composer 2.5 architecture, sharpeni…

Samsung Foundry joining as a manufacturing partner [^32062] and SiFive integrating NVLink Fusion for RISC-V AI data center chips [^32066] expand the ecosystem picture from hyperscalers and chip designers to now include …

The primary new development is NVLink Fusion: Jensen Huang's appearance at a Marvell keynote introduced a coexistence framing that explicitly reframes 'NVIDIA vs custom silicon' as 'NVIDIA plus custom silicon within a s…

For roughly two years, market observers expected custom ASICs to erode NVIDIA's dominant share of AI compute. Mid-2026 data suggests the opposite has occurred — NVIDIA has held or grown its position [^30984], and NVIDIA…