Thinking Machines Lab: $1 Billion Funding Talks, $40 Billion Valuation, Inkling, and NVIDIA

Thinking Machines Lab is in discussions to raise at least $1 billion at a pre-money valuation of roughly $40 billion, according to reporting on September 3. The round is still being negotiated, so the valuation, investor composition, and final amount should be treated as reported terms rather than a completed financing.

The talks place a very young AI company founded by former OpenAI CTO Mira Murati in the same financing conversation as much older frontier-model laboratories, but the company now has several elements investors can evaluate directly: an open-weights model family, a customization platform, a large NVIDIA infrastructure partnership, and reported annualized revenue already reaching the hundreds of millions of dollars.

The useful question is therefore not whether $40 billion is a large number in isolation. The relevant question is what combination of model capability, developer adoption, infrastructure access, commercialization, and future capital requirements would have to materialize for that valuation to remain economically defensible.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣

THE $1 BILLION ROUND WOULD PRICE A STILL-YOUNG LAB AT ABOUT $40 BILLION.

The reported financing terms are aggressive, but they are also lower than the valuation levels Thinking Machines explored during earlier fundraising discussions.

The Information reported that existing investor Accel is in talks to lead the new round and that NVIDIA has also discussed participating. Reuters subsequently summarized the report. No completed financing has been announced, and that distinction is important because private-market terms can change materially before signing.

........

Funding element	Reported position	Practical interpretation
New capital	At least $1 billion	Would extend runway for model training, product development, hiring, and infrastructure commitments.
Valuation	About $40 billion pre-money	Prices future execution before the new capital is added to the company.
Lead investor	Accel in talks to lead	Would deepen the role of an existing backer rather than introduce a completely new lead.
NVIDIA	Reportedly discussed participation	Would add another financial link to an already substantial infrastructure partnership.
Previous fundraising ambition	$4–5 billion at above $50 billion, discussed late last year	Shows that the current terms are below the earlier valuation target, even though they remain extremely high for the company's age.
Revenue	Hundreds of millions in annualized revenue, according to The Information	Provides a commercial reference point, although reported annualized revenue is not the same as audited recurring revenue or profit.

........

A $40 billion pre-money valuation means the company would be valued at roughly that level before the new funding is added. If the round closed at exactly $1 billion with no other structural adjustments, the simple post-money figure would be approximately $41 billion, although actual ownership outcomes depend on the security terms, option pools, secondary transactions, and any additional investors or capital.

The reported reduction from the more ambitious valuation discussed late last year is also relevant. It suggests investors may still be willing to finance Thinking Machines at a frontier-lab premium while demanding a lower entry point than the company previously sought.

That is a healthier analytical frame than treating the financing as a linear increase in value. Private AI valuations are negotiated expectations about future capability and market position, and they can move sharply even when the underlying company continues to grow.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁤

INKLING GIVES THINKING MACHINES A REAL MODEL PRODUCT, WHILE TINKER DEFINES THE CUSTOMIZATION THESIS.

The July release of Inkling changes the fundraising story because investors can now evaluate a concrete technical product rather than a research plan alone.

Thinking Machines released Inkling on July 15, 2026 as an open-weights multimodal model under the Apache 2.0 license. The company says the model uses a sparse Mixture-of-Experts architecture with 975 billion total parameters and 41 billion active parameters, while supporting a context window of up to one million tokens.

Inkling accepts text, image, and audio inputs and produces text outputs. The model card describes it as a 66-layer decoder-only transformer in which each token is routed through a subset of experts, and the company positions it for general instruction following, agentic coding, tool use, retrieval systems, and multimodal applications.

Thinking Machines also says Inkling was pretrained on 45 trillion tokens spanning text, images, audio, and video. That training figure is a vendor-reported specification, and it should be separated from independent evidence about model quality or economic efficiency.

The company is unusually explicit that Inkling is not the strongest model overall. Its strategic value lies in a different combination: open weights, multimodal capability, controllable reasoning effort, long context, and the ability to customize the model through Tinker.

Tinker is central to the business model because open weights alone do not automatically create a durable revenue stream. A platform that lets developers and enterprises fine-tune or adapt models can create recurring infrastructure usage, developer lock-in through workflow integration, and a commercial layer around otherwise portable model weights.

The availability of Inkling-Small reinforces that strategy. The smaller model uses 276 billion total parameters with 12 billion active parameters and preserves the one-million-token context target, giving the company a lower-cost option for workloads that do not require the full model.

For investors, the important distinction is therefore between model prestige and platform economics. A highly capable model can attract attention, while a customization platform can create the repeat usage and enterprise integration needed to support a large private valuation.

··········

NVIDIA IS ALREADY PART OF THE INFRASTRUCTURE STORY BEFORE ANY NEW INVESTMENT.

Thinking Machines has already tied a substantial part of its scaling roadmap to NVIDIA hardware, making any additional NVIDIA investment strategically coherent rather than surprising.

In March 2026, Thinking Machines and NVIDIA announced a multi-year partnership to deploy at least one gigawatt of next-generation NVIDIA Vera Rubin systems. The companies said deployment on the Vera Rubin platform was targeted for early 2027 and described the partnership as supporting frontier-model training and customizable AI at scale.

NVIDIA also made what Thinking Machines described as a significant investment at the time. The newly reported fundraising discussions therefore concern a company that already has a material technical and financial relationship with NVIDIA.

........

Layer	Current position	Why it affects the financing case
Frontier model	Inkling: 975B total / 41B active, 1M-token context	Gives the company a flagship technical asset that can be benchmarked, customized, and distributed.
Lower-cost model	Inkling-Small: 276B total / 12B active	Creates a broader cost-performance range for developers and enterprise workloads.
Customization platform	Tinker	Provides a potential recurring commercial layer around fine-tuning and model adaptation.
Compute partnership	At least 1 GW of NVIDIA Vera Rubin systems announced	Reduces uncertainty around access to frontier-scale compute while creating a very large capital and utilization commitment.
NVIDIA relationship	Existing strategic partnership and investment; further participation reportedly discussed	Could align capital supply, hardware access, and model deployment more tightly.
Commercial traction	Hundreds of millions in annualized revenue reported by The Information	Supports the case that the company is moving beyond pure research, but revenue quality and margins remain critical unknowns.

........

The infrastructure commitment creates both strategic strength and financial pressure. Frontier model development requires reliable access to accelerators, networking, power, data-center capacity, and software optimization, so a large hardware partnership can improve execution certainty.

At the same time, gigawatt-scale deployments require enormous capital and operating discipline. The financing case becomes stronger if model usage and enterprise revenue grow quickly enough to absorb that capacity, and weaker if infrastructure commitments expand faster than monetization.

This is why NVIDIA's role has to be read on two levels. It is a validation signal from the dominant AI infrastructure supplier, and it is also part of a capital-intensive operating model that can require repeated fundraising if utilization and revenue do not scale at the same speed.

··········

A $40 BILLION VALUATION NOW DEPENDS ON WHETHER RESEARCH, INFRASTRUCTURE, AND REVENUE SCALE TOGETHER.

Thinking Machines has moved beyond the stage where its value can be explained primarily by founder reputation. Mira Murati's OpenAI background remains important, but the company now has models, a customization platform, reported revenue, and a defined infrastructure roadmap that investors can evaluate against one another.

The strongest part of the case is the coherence between those pieces. Inkling provides an open-weights technical base, Tinker provides a path to customization and recurring developer usage, and the NVIDIA partnership provides access to the compute required for larger future systems.

The main risk is that all three layers are expensive to scale simultaneously. Model training consumes capital, large infrastructure commitments raise fixed and semi-fixed costs, and enterprise commercialization requires sales, support, security, reliability, and integration work that can grow more slowly than technical capability.

The reported hundreds of millions of dollars in annualized revenue are therefore important but insufficient on their own to validate a $40 billion valuation. Investors would still need to judge revenue quality, gross margins, customer concentration, infrastructure utilization, model differentiation, and how much additional capital the company may require before reaching durable operating leverage.

If the new round closes near the reported terms, the financing would show that private markets still assign exceptional value to AI laboratories that combine frontier research with a credible commercialization layer. It would also set a high execution threshold: Thinking Machines would need to convert technical ambition and infrastructure access into sustained economic output quickly enough to justify a valuation normally reserved for companies with far longer operating histories.

FOLLOW US FOR MORE.

·····

DATA STUDIOS

·····

[datastudios.org]