Why are Big Tech employees leaving to create AI startups

In 2026, there was a noticeable increase in the flow of researchers and engineers from the largest AI labs into their own startups. Former employees of Google DeepMind, OpenAI, and other companies are creating new AI laboratories, attracting hundreds of millions of dollars, and beginning to compete with their former employers.

From DeepMind to Own AI Labs

David Silver explained his departure in a January 2026 post on the Ineffable Intelligence website. He wrote that he wanted to create a space for ambitious research without the need to adapt to the requirements of short-term products and profitability.

The company describes its goal as creating a superlearner — a system capable of continuously learning from its own experience. Unlike the approach where AI is primarily trained on human data, Silver’s team is betting on learning through interaction with the environment. David Silver himself notes that before founding the company, he led reinforcement learning at DeepMind, and now heads Ineffable Intelligence.

Investors are ready to fund such projects at an early stage. In April 2026, Lightspeed Venture Partners announced investments in Ineffable Intelligence, linking the project to Silver’s research in reinforcement learning.

It later became known that the company raised $1.1 billion at a valuation of $5.1 billion. According to TechCrunch, at the time of the round, the company had existed for only a few months.

This case illustrates a shift in the market: investors are willing to invest large sums not only in a ready-made AI product but also in the team, research expertise, and specific technological hypotheses.

Knowledge from Big Tech Becomes the Foundation of New Companies

Another example is Latent Labs. The company was founded by former DeepMind researchers who worked on AlphaFold and protein research.

On its own website, Latent Labs states that its founders participated in the development of AlphaFold models, and the team itself works on generative AI for biology. In February 2025, the startup announced it had raised $50 million.

The company is not trying to create a universal chatbot. Its focus is on foundation models for biology, which can be used for antibody development, enzyme optimization, and other tasks in molecular biology.

Former OpenAI Employees Create Their Own Ecosystem

A similar situation has developed around OpenAI. One of the most notable examples is Anthropic, founded by Dario and Daniela Amodei after their work at OpenAI. The company is now itself one of the leading AI laboratories.

In 2025, former OpenAI CTO Mira Murati also launched her own company, Thinking Machines Lab. The project is built not only around the practical use of existing models. The team is developing its own AI models and exploring new ways for humans and artificial intelligence to interact.

For example, in May 2026, Thinking Machines Lab published research on Interaction Models. The company describes models that can simultaneously perceive audio, video, and text and interact with humans in real time.

And in July, the team released Inkling — its own model with open weights. According to the company, the model contains 975 billion parameters, of which 41 billion are active, and was trained on 45 trillion tokens of text, images, audio, and video.

This is no longer a classic startup that simply takes a large model’s API and builds a service on top of it. Former employees of AI laboratories are attempting to build their own fundamental technologies.

Why Specialists Are Leaving

1. The Opportunity to Pursue Their Own Research Ideas

For some researchers, the main motivation is the ability to independently define the direction of their work. This is exactly what David Silver wrote. On the Ineffable Intelligence website, he explained his decision to create a new company as a desire to work on long-term research tasks without the need to focus on short-term products and commercial goals.

This is particularly relevant for frontier AI. Research into the next generation of models can take years, whereas businesses often need faster results.

A startup allows the team to independently determine the work horizon and technical hypothesis.

2. The Ability to Faster Turn Research into a Company

Within Big Tech, a researcher works within the structure of a large organization. A startup operates under a different model: a small team can focus entirely on one technology.

Latent Labs is an example of this approach. The company formed around a team with experience in creating models for biology at DeepMind, and then built a separate business around this expertise.

This results in a transformation: a research project within a large company becomes an independent technological direction.

3. Investors Are Ready to Fund Not Only Products but Also Teams

The AI market has also changed its attitude toward early-stage funding. In a traditional startup, the founder usually has to prove the existence of a market, a product, and initial users. In frontier AI, investors may bet primarily on the team and technological expertise.

This lowers one of the barriers to leaving a large company: a specialist can gain access to capital if they have a strong research reputation and team behind them.

Big Tech Is Not Left Without AI Specialists

It is important not to conclude that large technology companies are losing all their researchers. Competition for AI talent continues in several directions simultaneously. Specialists move from one laboratory to another, create their own companies, or return to Big Tech in a different capacity.

In the case of new AI startups, another process is occurring: former employees bring their colleagues with them. This forms teams in which several participants previously worked together on major AI projects.

Latent Labs explicitly notes that its founding team consists of specialists who participated in the creation of AlphaFold, and among the company’s investors are leaders of other major AI companies, including Jeff Dean from Google and Dario Amodei from Anthropic.

Thus, the boundary between Big Tech and startups is becoming less distinct. Today, a researcher can work at Google DeepMind, tomorrow create their own company, and then collaborate again with large technology companies as a founder.

What This Means for the AI Market

The emergence of new companies founded by former employees of the largest laboratories creates additional competition for several key resources.

The first is people. Experienced researchers become founders and attract other specialists.

The second is capital. Investors are ready to fund projects at earlier stages, especially if they are backed by well-known researchers.

The third is technology. Knowledge and research approaches that were previously concentrated within a few companies are gradually spreading throughout the ecosystem.

The current wave of founders from Google DeepMind, OpenAI, and other laboratories is not just a story about changing employers. It shows how the structure of the AI industry is changing: cutting-edge research is increasingly becoming the starting point for creating new independent companies.