Nvidia Bets $2.5B on Mira Murati’s Thinking Machines Lab

Nvidia backs Mira Murati’s AI startup with $2.5B as Thinking Machines Lab eyes $40B valuation. Plus: TikTok dodges child-safety scrutiny.

Nvidia Bets $2.5B on Mira Murati’s Thinking Machines Lab
Photo by Growtika on Unsplash

The story of the day

Nvidia is in talks to inject $2.5B into Mira Murati’s Thinking Machines Lab, a bet that could value the AI startup at $40B+ before its Series A even closes. Accel is reportedly leading the round, with Nvidia covering roughly half the target $5-6B raise. For builders, this signals Nvidia’s shift from hardware enabler to direct stakeholder in next-gen AI development—blurring the line between chipmaker and AI lab. Expect tighter integration between Nvidia’s stack and Thinking Machines’ models, with potential lock-in risks for devs relying on alternative infra.

New tools & launches

Dude — Dual-detection multi-agent system for paper-code discrepancy detection. Open-source. [arXiv:2609.03416v1]

DuplexSpeechBench-IFEval — Benchmark for implicit instruction-following in full-duplex voice agents. [arXiv:2609.03423v1]

Speculative Macro Commit (SMC) — Runtime mechanism to speed up tool-using LLM agents via predictive action chains. [arXiv:2609.03236v1]

Models & research

PlanFence introduces dependency-scoped validation for distributed LLM-agent memory, addressing "stale-plan execution" where agents act on outdated plans despite fresh data. In 30 live workflows, freshness-only executors acted on stale plans 87% of the time; PlanFence reduced this to 12%. Open-weight. [arXiv:2609.03340v1]

MasterControl proposes a governed enterprise analytics framework where LLMs interpret queries but deterministic policies execute pre-approved programs. Across 440 runs, no model fully satisfied the answer-and-evidence contract, highlighting reliability gaps in runtime planning. [arXiv:2609.03209v1]

Industry pulse

TikTok canceled a U.S. House meeting on national security, citing a desire to avoid scrutiny of its child-safety practices, according to committee chairman John Moolenaar. The move underscores the platform’s balancing act between regulatory compliance and growth—leaving AI builders leveraging TikTok data in a gray zone. [The Information]

Stat of the day

Thinking Machines Lab’s target valuation: $40B+ pre-money—before its first institutional round closes. For context, Mistral AI’s last valuation was $2B. [The Information]

Get the full analysis — subscribe to Gradient Pro.