Daily digest · 16 November 2026
A quiet weekend for deals, so the day's work was reading partnership pages nobody had read closely. The best of them was Microsoft's: a July post about its expanded Mistral partnership says Microsoft will use Mistral's European GPU fleet to serve its own cloud, the reverse of the direction this map would have assumed, and the same post puts two Mistral models into Microsoft Foundry. Reflection AI turned out to be renting from two companies already on the map, SpaceX and Nebius. And GlobalFoundries' largest customer, Qualcomm, was named not in its annual filing but in the quote block of a capex release. Eight new relationships in all, no company marks moved, nobody joined or left the map. In the wider industry the NVIDIA succession fight kept running and Anthropic's IPO prices on Wednesday.
No sourced valuation changed today.
Anthropic holds at the $965B Series H. Its range of $70 to $78 is an offer, not a price; it prices on 11-18 and trades on 11-19. One outlet is already running a headline saying it has priced at $78 over a body that says otherwise.
microsoft -rents→ mistral - Microsoft says it will use Mistral's expanded Europe-based GPU infrastructure, thousands of NVIDIA Vera Rubin parts, to add capacity for its own cloud and AI services. Multibillion, no figure disclosed by either side. Microsoft's newsroom, 2026-07-21.azureai -builds_on→ mistral - Mistral's Medium 3.5 and OCR 4 are now available in Microsoft Foundry, Medium 3.5 also in Copilot Studio. Same post, 2026-07-21. That leaves xAI as the only provider in the set still undrawn.reflection -rents→ spacex - SpaceX sells Reflection GB300 capacity at Colossus 2 in Memphis for $150M a month from 1 July through 2029, up to about $6.3B over the full term. CNBC, 2026-06-22, reported because the figure is CNBC's.reflection -rents→ nebius - Nebius's Q2 2026 letter to shareholders says Reflection selected it for a multi-year agreement to train and run its open-source models, 2026-08-12. The letter states no value, so the edge carries no amount.qualcomm -fabricated_by→ gf - GlobalFoundries' $16B US investment release quotes Qualcomm's chief executive calling GF "a strategic supplier of Qualcomm" and names Qualcomm as GF's largest customer. GlobalFoundries, 2025-06-04.cohere -builds_on→ nvidia - Cohere builds an NVIDIA ecosystem-native model and a local North instance for privately run systems. Cohere, 2026-03-16.elevenlabs -rents→ alphabet - a multi-year extension under which ElevenLabs trains and serves its voice models on Google Cloud G4 VMs with NVIDIA Blackwell GPUs. Google Cloud's press corner, 2026-02-26.zhipu -builds_on→ ascend - Zhipu says its open-source GLM-Image was the first state-of-the-art multimodal model trained entirely on Huawei Ascend chips. Bloomberg, 2026-01-14.astera -builds_on→ nvidia onto Astera Labs' own release of 2025-12-02 announcing NVLink Fusion custom connectivity with NVIDIA quoted, and wayve -rents→ coreweave onto CoreWeave's own release rather than a trade relay. cursor -rivals→ cognition and anduril -designed_with→ meta were re-checked, left in place and date-stamped: neither has a 2026 page of the grade this map wants. No edge ended today.Reviewed: Zhipu / Z.ai, Reflection AI, Mistral, GlobalFoundries, ElevenLabs, Etched, Cohere.
Further reading: TLDR AI · The Rundown · Techmeme
Validator: 231 entities, 509 relations, 0 errors, 1 warning (40 ongoing ties over a year old). Commit: 8 new edges, 2 citation refreshes, 2 date-stamps, 4 roles rewritten, 7 company reviews.
State of AI World maps every significant AI company, what it is worth and how it is connected. Every number carries a source and an as-of date; see the methodology. Data is free to reuse with credit: open data. · All companies · Daily digest · Corrections and contact