AI lab

Thinking Machines Lab

An AI lab founded by Mira Murati, OpenAI's former chief technology officer, that builds its own models — among them Inkling, whose weights it publishes for anyone to download — and charges for fine-tuning and running models on its Tinker platform. It gets its computing from several suppliers: Google Cloud's Nvidia GB300 machines, a gigawatt-scale partnership with Nvidia announced in March 2026, since May 2026 5,000 Nvidia B300 GPUs from Boost Run under orders it cannot cancel, and, since September 2026, Crusoe, which it pays $65 million a year to serve its open models.

2 deals (2 T1) · 2 flows on the money map ·

Deals 2

sort: · · ·
  1. Sep 23, 2026 updated Nov 18
  2. May 28, 2026 updated Nov 19

Money in and out 2 ribbons

Who pays Thinking Machines Lab (left), and whom it pays (right). Width is a rough size in $B, per year for multi-year deals.

← swipe to see the whole chart →

sourced deal · 2 of 2

As a table · 2 ribbons
Notes in the record on Thinking Machines Lab

Thinking Machines Lab Inc. — so named in Boost Run's 8-K of May 28 2026 and its exhibit index; Boost Run's later registration statements spell it 'Thinking Machines Labs, Inc.'. Funding: a $2B seed round led by Andreessen Horowitz at a $12B valuation, closed in July 2025, with Nvidia, Accel, ServiceNow, Cisco, AMD and Jane Street participating (a company spokesperson to TechCrunch, Jul 15 2025); 'NVIDIA has also made a significant investment' (its own announcement, Mar 10 2026; size not stated). Models: Inkling, 'Our Open-Weights Model' (Jul 15 2026); Tinker, its platform for fine-tuning and sampling models, with a published price list. Compute it rents or buys: Boost Run — 5,000 NVIDIA B300 GPUs with storage and CPU nodes, an initial 36 months, ~$471.7M, non-cancelable (boostrun-tml); Crusoe — Managed Inference on a dedicated Nvidia HGX B200 deployment, $65M a year, no term stated (tml-crusoe); NVIDIA — 'a multi-year strategic partnership to deploy at least one gigawatt of next-generation NVIDIA Vera Rubin systems', deployment 'targeted for early next year' (its own announcement, Mar 10 2026; no record yet); Google Cloud — a new agreement for 'A4X Max VMs with NVIDIA GB300 GPUs' (Google Cloud's release, Apr 22 2026; no record). Reported, not confirmed by the company: talks in September 2026 on a new round at a valuation of at least $40B, first reported at about $1B and then at $5–6B (The Information, relayed by TechCrunch and The Next Web; The Inference has $5–6B at a $40B pre-money valuation), and a revenue run-rate known only from anonymous sources — neither is carried on the map.