L2 · AI ASICs
AI ASICs — Frequently Asked Questions
Hyperscaler silicon: TPU, Trainium, MTIA, Maia — and the merchant gap they create.
What is the AI ASICs topic on aiinframap?
AI ASICs sits on layer L2 of the AI physical-infrastructure stack. Google TPU v5/v6, AWS Trainium 2/3, Meta MTIA, Microsoft Maia: training-economics deltas vs NVIDIA, packaging supply contention, and what they reveal about the merchant-GPU market.
How big is the AI ASICs market?
Market size estimate is generated by the D1 Market Analysis module and may not yet be populated for this topic. See /topic/ai-asics/market when available.
Who are the top companies in AI ASICs?
aiinframap currently tracks 6 active companies in this topic. The top by Fit Score include: AMD, Groq, Cerebras Systems, Tenstorrent, SambaNova Systems. Full ranked list: /topic/ai-asics/companies.
What aiinframap tools exist for AI ASICs?
Engineer-facing decision tools: H100 vs H200 vs B200 TCO, Inference Cost Calculator, Memory Bandwidth Bottleneck Detector. Live at /tools/<layer>/<slug> with companion JSON APIs at /api/v1/tools/...
What jobseeker personas does aiinframap recognise in AI ASICs?
4 segments with weighted Fit Score profiles. ASICS-ML-SYS-ENG-01: ML systems engineer evaluating merchant AI silicon (AMD / Groq / Cerebras / etc.) vs NVIDIA defaults. · ASICS-PROCURE-LEAD-01: Procurement / supply-chain lead negotiating accelerator allocation at large AI buyer. · ASICS-RESEARCH-SCI-01: Research scientist running benchmarks on merchant AI accelerators for publication / internal eval.. Full list at /topic/ai-asics/personas.
How does aiinframap score companies in AI ASICs?
Five dimensions (growth, comp, learning, stability, culture) weighted by jobseeker segment. The persona detail page at /topic/ai-asics/personas/<segment-id> has an interactive weight tuner so you can rerank companies by your own priorities.
Where is AI ASICs headed in the next 12 months?
OpenAI's 6 GW AMD Instinct commitment plus a 160M-share AMD warrant, with Oracle adding 50,000 MI450 GPUs from Q3 2026 — open-weight gpt-oss models make the diversification off NVIDIA real.
How fresh is aiinframap data on AI ASICs?
Business + talent signals refresh daily via SEC EDGAR / LinkedIn / Greenhouse / Crunchbase collectors. Business modules (D1-D6) regenerate weekly. Last-updated timestamps are visible on every public page.
Can I get this data via API?
Yes — JSON endpoints: /api/v1/topics/ai-asics (topic-level), /api/v1/companies/<slug> (per-company), /api/v1/tools/<layer>/<slug> (per-tool). All free, no auth required, cacheable.
Where can I get weekly updates on AI ASICs?
aiinframap Weekly newsletter ships every Friday with topic-specific lead stories, top signals, and hiring radar. Subscribe at /weekly.