Google DeepMind Invests $75M in A24, Putting AI Inside Hollywood

In a deal that no major Hollywood studio and no major tech company had managed to pull off before, Google announced on June 22 that it is taking an equity stake in A24 — the studio behind Everything Everywhere All At Once, Hereditary, and The Backrooms — in a $75 million investment that will embed a team of Google DeepMind researchers directly inside its film productions.

The partnership is structured as a multiyear, nonexclusive research and development collaboration. DeepMind scientists will work alongside A24 filmmakers on active productions, jointly developing new AI-powered tools and workflows. The central technology is Veo 3.1, DeepMind’s latest video generation model, which can produce 4K video clips from text or image prompts with synchronized native audio and reference-image-based character consistency.

Why A24, and Why Now

Google’s choice of A24 as its first studio investment is telling. A24 has spent the past decade building a reputation for auteur-driven, awards-season cinema that trades on its distinctive aesthetic identity. It is simultaneously one of the most critically admired film studios in the world and, with estimated annual revenues of around $1.5 billion, a relatively nimble operation compared to the major conglomerates.

That combination — creative prestige plus operational flexibility — makes A24 an ideal laboratory for AI tooling that Google would not find at a Disney or Warner Bros. A studio that prizes creative control is less likely to use AI as a blunt cost-cutting instrument, which means any tools developed through the partnership will be designed around filmmaker workflows rather than around reducing headcount.

The timing reflects a maturation in AI video models that makes them practically useful for professional production for the first time. Veo 3.1 is a meaningful advance over its predecessors: it maintains character consistency across shots using reference images, generates synchronized dialogue and ambient audio natively, and produces output at 4K resolution with photorealistic quality in many lighting conditions. Earlier versions of generative video failed at professional standards in nearly every one of these dimensions.

The Deal Structure

According to terms disclosed by both parties, Google will not gain access to A24’s content library. The studio’s existing film catalog — a key piece of intellectual property — is explicitly excluded from the AI training relationship, a concession that address one of the industry’s most contentious concerns about tech-studio deals.

The DeepMind team embedded in A24 will focus on three early application areas. The first is AI-generated storyboards that can simulate how a scene will look before cameras roll, potentially allowing directors and cinematographers to catch staging and lighting problems weeks earlier in the production timeline. The second is AI-assisted previsualization — essentially animatic generation from script pages — which could compress pre-production timelines. The third is targeted use of Veo for visual effects work, particularly for environments and background extensions where traditional VFX is expensive and time-consuming.

Critically, the arrangement is designed to keep artists in creative control. The tools DeepMind is building are positioned as accelerators and aids to human filmmakers, not as replacement systems for directors of photography, production designers, or visual effects supervisors.

Reactions from the Creative Community

The announcement was met with significant skepticism from parts of the film industry. On social media, a subset of A24’s core audience — which self-identifies heavily with the studio’s artistic independence — described the deal as a betrayal of the brand. The Kotaku headline captured the sentiment: “A24 Fans Mourn Its Death After The Backrooms Studio Signs $75M Google AI Deal.”

Industry unions were more measured. The International Alliance of Theatrical Stage Employees (IATSE) issued a statement noting that the deal does not appear to directly affect current collective bargaining agreements, but called for close monitoring to ensure that AI tools developed under the partnership do not lead to displacement of below-the-line crew.

Executives at competing studios watched the announcement carefully. Three people with knowledge of discussions at major studios said privately that at least two other large content companies have held exploratory conversations with AI video model developers in the past six months, none of which have progressed to investment terms.

The Broader Strategic Picture

For Google, the deal is less about the $75 million investment — a rounding error for a company with $90 billion in annual free cash flow — and more about establishing a credible, high-profile deployment environment for Veo in a sector that remains skeptical of AI. Hollywood is a uniquely visible battleground: if Veo 3.1 produces compelling work in an A24 production, that demonstration effect reaches producers, directors, and studios globally in a way that technical benchmarks never could.

The deal also positions Google against OpenAI’s own entertainment play. OpenAI discontinued Sora, its video generation product, earlier this month following creative-community backlash and content licensing disputes. The retreat left a gap in the AI-to-Hollywood pipeline that Google is now moving aggressively to fill.

Google DeepMind CEO Demis Hassabis framed the partnership in scientific terms. “Film is one of the richest environments for understanding how humans compose meaning from moving images,” he said in a prepared statement. “Working inside A24 productions gives our researchers access to creative decisions being made in real time, at the highest level of craft.”

A24’s co-CEO Daniel Katz offered a complementary view: “The tools don’t replace instinct or vision. What they can do is remove friction between what a director imagines and what they can actually test before a shoot.”

Whether AI ends up in the credits of an A24 film — and how audiences respond when it does — may prove to be the most consequential experiment in the partnership.