What is Neolab?

A research-first AI startup, usually founded by researchers who left an established lab, that raises large sums for a long-term breakthrough before it has a product.

Definition

A neolab is a research-first AI startup built around a technical bet. Its founders are usually researchers who left an established model maker such as OpenAI or Google DeepMind, and it raises large amounts of capital to pursue a long-term breakthrough. A product, if one comes, grows out of the research later.

Business Insider described neolabs on 22 October 2026 as "the lavishly funded, research-first AI startups founded by top talent who defected from the likes of OpenAI or Google DeepMind". Safe Superintelligence is the reference case: Dealroom reported it valued at $32 billion as of October 2026, with a stated plan to ship no product until it reaches "superintelligence".

The word sits between two other kinds of AI lab. A frontier lab already sells access to the most capable general models. An applied AI lab trains models for work businesses already do and earns from its first deployment.

Why It Matters

Neolabs are among the largest bets in venture capital in 2026. Dealroom reported on 18 October 2026, summarising The Economist and a count by Radical Ventures, that 29 neolabs had raised a combined $23 billion and fewer than 35% had released a commercial product. Investors are split: NEA argued in June 2026 that neolabs could produce the next generation of venture outcomes, while Business Insider reported in October 2026 that some venture capitalists question whether neolab valuations can be justified.

How It Works

A neolab typically recruits a small team of senior researchers, raises a large early round on the strength of that team and its research direction, and spends most of the capital on compute, talent and a training pipeline. NEA noted in June 2026 that neolabs raise $50 million to $500 million before they have revenue, so their investors hold positions for years before a commercial signal arrives. Some neolabs aim at general intelligence; others target a domain, such as tabular data, robotics or drug discovery, where they can own a specialised data corpus.

Example Use Case

An investor comparing an AI research startup with an applied AI company can use the word to frame the risk. A neolab is valued on its team and the plausibility of its research bet, with revenue years away. An applied lab is valued on its deployments, the revenue each model earns in service, and what each deployment adds to its product.

Key Takeaways

A neolab is a research-first AI startup that raises large sums for a long-term breakthrough before it has a product.
As of October 2026, 29 neolabs had raised $23 billion combined, and fewer than 35% had released a commercial product (Dealroom).
Founders are usually researchers from established labs such as OpenAI or Google DeepMind.
Investor opinion is divided between the power-law upside and the risk of heavy spending with little revenue.
A neolab differs from an applied AI lab, which earns from its first deployment, and from a frontier lab, which already sells general models.

How Ertas Helps

Ertas AI is an applied AI lab, a different kind of company from a neolab. It trains specialist models for tasks businesses and their AI agents already repeat, and it earns from its first deployment. Its research is published with dated measurements on the Ertas research page, and it runs in production through the adaptive model endpoint. The endpoint is currently invitation only, and we are developing a more streamlined version for wider access.