DeepMind Bets $75M on AI Tools for Real Filmmaking

Google DeepMind and A24, the studio behind films including 'Everything Everywhere All at Once' and 'The Backrooms,' have announced a long-term research partnership to explore how AI can be used in real film production. Google is investing approximately $75 million in A24 as part of the arrangement. This is a notably different kind of AI deal: rather than integrating a finished AI product into a workflow, the partnership is structured so that working filmmakers can test tools in the middle of actual productions and report their findings back to researchers.

Real films as a research environment

The partnership is structured as a research exploration first, not a product launch. Eli Collins, VP of Product at Google DeepMind, described the arrangement as providing 'real-world feedback from working professionals.' A24 directors, editors, and production teams will work with AI tools as part of their regular process — not in a sandbox, but on real projects with real deadlines. The specific tools and intended outcomes are not being announced upfront; what Google is buying with its $75 million investment is the kind of feedback that only comes from professionals who will actually use the tools under creative and time pressure.

A24's role in this is meaningful beyond its name recognition. The studio is known for betting on unusual and ambitious projects, and its library includes work generally considered to be artistically demanding. That makes it a different kind of test partner than a studio producing primarily franchise content. AI tools that prove useful in an A24 production would need to earn their place among filmmakers who care about outcomes that are hard to quantify — not just faster or cheaper, but genuinely better in ways that matter to the final work.

What this signals for AI in creative work

The broader pattern this fits is one of major AI labs seeking structured feedback from professional practitioners in domains where quality is subjective and workflows are difficult to replicate in a research setting. The same approach has emerged in medicine, law, and architecture. In each case, the goal is identical: get real practitioners doing real work with the tools so researchers understand what actually breaks and what actually helps, rather than what looks good in a benchmark.

For builders thinking about AI in creative tools, the signal is that the next generation of these tools is being shaped by feedback from the professionals who are hardest to please. When DeepMind eventually publishes research from this partnership, the findings will reflect what happened when AI encountered a director or editor making decisions under deadline pressure, not what happened in a controlled evaluation. That tends to produce insights about where AI assistance genuinely helps — and where it gets in the way or produces work that feels subtly wrong.

The $75 million investment is also a statement about Google's intentions. This is not a sponsorship deal or a co-marketing arrangement — it is equity capital going into a studio in exchange for access to a real-world testing environment. Google is paying to learn what AI can actually do in filmmaking at the level where it's expected to compete with human craft.