AI startup Thinking Machines eyes up to $60b valuation

Thinking Machines Lab, an AI startup founded by former OpenAI executive Mira Murati, is in early talks to raise a new funding round that could value the company at about US$50 billion.

The valuation could also rise to as much as US$55 billion to US$60 billion.

The company, launched in early 2025, develops AI models and tools for business applications and released its first product, Tinker, earlier this year.

It previously raised US$2 billion at a US$12 billion valuation in July 2025.

Several former OpenAI staff, including co-founder John Schulman and research head Barret Zoph, have joined Murati at the startup.

Thinking Machines has paying business customers for Tinker and has received positive feedback from university researchers.

The company is also facing strong competition for AI talent, with co-founder Andrew Tulloch recently leaving for Meta.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Tinker’s valuation needs traction to hold

Valuation rose from US$12 billion in July 2025 to US$50–60B within months 1. Key metrics such as annual recurring revenue (ARR) and enterprise retention remain undisclosed, and it has not named deployments despite mentions of paying businesses 1 2.
Tinker uses Low-Rank Adaptation (LoRA) from 2022 as a managed service 3. Fast copycats could offer the method, which pressures any advantage 3.
Fine-tuning infrastructure helps teams adapt pretrained models to their own data and serves a niche of AI researchers and specialized developers 3. That focus can cap revenue versus broad enterprise software.

Multi-model infrastructure vendors can build integration tools for Tinker adopters

A new fine-tuning platform such as Thinking Machines can spur demand for multi-model orchestration tools that route and manage workflows across several LLMs. Many enterprises use more than one provider.
Tooling firms can build migration paths between Tinker-fine-tuned models and other services 1. The platform supports cloud, on-premises, or edge deployment across hardware outside data centers such as smartphones or factory gear 1.
Monitoring and benchmarking vendors track model performance, latency, and reliability 1. They can plan products once they get Tinker’s API compatibility, pricing, and Service Level Agreements (SLAs) 1. Cost platforms can target teams that use Tinker with proprietary APIs, since buyers need visibility across providers 4 5.