Google DeepMind A24 Deal Signals AI's Hollywood Turn

The Google DeepMind A24 partnership announced on June 22, 2026 puts one of the most admired names in independent cinema inside the same research loop as one of the largest artificial intelligence labs on the planet. Google is putting roughly $75 million into A24, the studio behind Everything Everywhere All at Once, as part of a multi-year arrangement to build AI tools aimed at filmmakers. The figure and the structure were first reported by the Wall Street Journal and confirmed in coverage from TechCrunch and The Decoder, alongside a short post on Google's own blog.

What makes this more than another tech-meets-entertainment press release is who is on each side of the table. A24 is not a legacy conglomerate chasing a buzzword. It is a studio with a reputation for backing directors with strong personal vision, the kind of company whose brand depends on taste rather than scale. DeepMind, for its part, has spent the past two years pushing its video generation research into public view. Bringing them together is a bet that the friction between artistic credibility and machine generation can be worked out in a lab rather than fought over in public.

The terms of the Google DeepMind A24 deal

The shape of the agreement matters as much as the dollar amount. According to Variety, the partnership is multi-year and non-exclusive. A24 keeps the right to work with other AI companies and other models, and DeepMind can pursue its own collaborations with other studios. That non-exclusivity is a deliberate signal. Neither party is locking the other into a closed pipeline, which softens the impression that a studio is handing its creative future to a single tech vendor.

The most pointed clause is about data. Variety reports that the deal does not grant Google access to A24's content library or its data. For a studio whose value sits in its catalog and its relationships with filmmakers, that boundary is the difference between a research collaboration and a sale. DeepMind gets practical feedback from working artists. It does not get a training corpus of finished A24 films in exchange. The trade outlet World of Reel noted that the agreement is described as covering both filmmaking and distribution techniques, which hints that the scope reaches beyond the camera into how films find audiences.

The official framing from Google describes a research effort rather than a product launch. In its blog post, DeepMind called the arrangement a first-of-its-kind partnership focused on research, with A24 filmmakers set to work alongside researchers to test and refine new approaches over time. The same post is candid that specific goals and technical outputs will evolve as the work continues. In plain terms, there is no shipping product today, and no public roadmap for one.

What A24 Labs actually wants to build

The clearest sense of intent comes from A24's side of the house. The studio's technology division, described in trade coverage as A24 Labs, is led by Scott Belsky, a product executive known for founding Behance and for years of senior product work in creative software. Belsky used the announcement to draw a sharp line between this effort and the AI tools that have unsettled much of Hollywood.

Belsky added that the resulting tools will not resemble the prompted generation style of AI that people feel uncomfortable with, according to Variety. That phrasing is doing a lot of work. The fear inside film communities is not that software exists. It is that a producer types a sentence and a finished shot appears, cutting human craft out of the loop. By promising something that looks nothing like a text box that spits out footage, A24 is trying to reassure the directors and cinematographers whose trust it needs to keep.

One concrete area surfaced in reporting is AI-assisted storyboarding. A24's group is said to be developing applications for AI-generated storyboards, a pre-production stage where directors plan shots before a single frame is captured. Storyboarding is a useful test case because it sits early in the process, where iteration is cheap and where no performer's likeness is at stake. Trade coverage noted that veteran director Martin Scorsese has lent support to AI storyboard technology, which gives A24 a recognizable name to point to when skeptics push back. Whether the tools stay confined to pre-production or creep toward generating final imagery is exactly the kind of question the announcement leaves unanswered.

Why DeepMind needs a studio, not just a model

From Google's perspective, the logic is about feedback rather than footage. DeepMind has strong generative video research, but a research demo and a usable filmmaking tool are different animals. A model that produces an impressive clip in a benchmark can still be useless to a director who needs control over framing, continuity, lighting, and timing across a sequence. The only way to learn what professionals actually require is to put experimental technology in their hands and watch how it breaks.

Eli Collins, vice president of product for DeepMind, made that reasoning explicit. "We believe breakthroughs happen when you get technology into the hands of the best minds in the field," he told Variety. Demis Hassabis, the DeepMind co-founder and chief executive, struck a similar note in remarks reported by TechCrunch, saying the best way to develop tools that empower artists is to work directly with them. Stripped of the polish, both quotes describe the same arrangement. DeepMind supplies access to its research and infrastructure. A24's filmmakers supply the hard, specific complaints that turn a flashy model into something a working crew can use.

This is also a credibility play. DeepMind already has ties to individual filmmakers, including reported collaborations with directors exploring AI in their own projects. Partnering with a studio that carries genuine artistic prestige helps Google argue that its video research belongs in serious filmmaking rather than in the flood of low-effort synthetic clips that have given generative video a poor reputation. A24's name is, in part, what Google is buying.

DeepMind's road from short films to a studio deal

The A24 partnership did not appear out of nowhere. It builds on a template DeepMind has been testing for more than a year with individual filmmakers. In May 2025, the lab announced a tie-up with Primordial Soup, the storytelling venture founded by director Darren Aronofsky, with the stated goal of putting a leading video generation model into the hands of top filmmakers. As Google described in its own blog post, Primordial Soup set out to produce three short films using DeepMind's generative tools, each helmed by an emerging director mentored by Aronofsky and backed by Google's research team.

The first of those films, ANCESTRA, was directed by Eliza McNitt and premiered at the Tribeca Festival on June 13, 2025. It used DeepMind's Veo video model, in the Veo 3.1 generation, to explore how generated footage could be woven together with live action rather than replace it. Crucially, the production did not hand the work to a model and walk away. Traditional animators, visual effects specialists, concept artists, and storyboard artists worked alongside DeepMind researchers, with the stated aim of preserving emotional continuity and the director's vision.

That earlier project is the proof of concept the A24 deal scales up. The Aronofsky collaboration showed DeepMind how to embed researchers in a real production and collect feedback from working artists, while keeping humans in the creative driver's seat. The A24 arrangement takes the same feedback-driven model and attaches it to a studio with a full slate, a brand built on directorial vision, and an in-house technology team. The bet is that what worked for a handful of short films can be turned into tools used across an entire production company.

Why the equity stake is the unusual part

Plenty of studios have signed AI deals over the past year, but most have looked like vendor relationships. A studio licenses a model, pays for access, and walks away if the tool disappoints. The Google arrangement is different because it routes the money through an equity investment rather than a service contract. Reporting indicates this is among the first times Google has taken a direct stake in a film company of A24's profile, a step most large technology firms had stopped short of with studios that carry this kind of cultural weight.

An equity stake changes the incentives on both sides. DeepMind now has a financial reason to want A24 to succeed, not just to adopt its tools, which aligns the lab with the studio's creative output rather than treating it as a customer to be billed. A24 gains a deep-pocketed partner whose research budget dwarfs anything an independent studio could fund alone. The cost of that closeness is reduced independence. When a technology partner is also an investor, disagreements over how aggressively to deploy AI carry boardroom weight, not just creative weight.

The structure also tells you something about how seriously Google is taking the entertainment market. A licensing deal is a sales motion. An equity investment paired with embedded researchers is a long-term commitment to learning a specific industry from the inside. Google is treating filmmaking as a domain worth years of patient research, which is a stronger signal of intent than a one-off product launch would have been.

A Hollywood already crowded with AI deals

The A24 arrangement does not land in a vacuum. It arrives during a stretch where nearly every major player in entertainment has staked out an AI position. The Decoder and other outlets noted the broader pattern: Netflix moved to acquire an AI filmmaking tools company founded by Ben Affleck, Amazon's MGM Studios spun up a dedicated AI unit for television and film production, and studios including Lionsgate and Disney have explored their own arrangements with AI providers. Reporting from The Hollywood Reporter and Deadline placed the Google deal squarely inside that wave.

The competitive backdrop explains the urgency. If AI tools become a standard part of pre-production and post-production, the labs that shaped those tools alongside trusted studios will have a head start. Google is racing the same clock as everyone else, and a respected partner like A24 is a scarce resource. There are only so many studios whose endorsement carries weight with the artists who decide what software a production actually adopts. By locking in A24 early, Google denies that endorsement to its rivals while buying itself a live testing ground that competitors cannot easily replicate.

The labor question the announcement sidesteps

For all the talk of empowering artists, the announcement says almost nothing about the workers whose livelihoods are most exposed to generative video. That silence is conspicuous because the labor fight over AI in Hollywood is far from settled. In June 2026, SAG-AFTRA members ratified a four-year agreement with studios and streamers that, as The Hollywood Reporter detailed, builds on the digital replica protections won during the 2023 strike and adds new limits on the use of synthetic performers.

The new contract commits producers to a principle strongly favoring human performances and asks them not to swap synthetic performers for human ones unless doing so adds significant additional value to a project. Executive director Duncan Crabtree-Ireland framed the deal as a way to ensure synthetics remain the exception in the industry instead of the rule. Critics inside the union, reported by Variety, argue the significant additional value standard is too vague to enforce and that studios will test its edges. The union also cannot strike over AI again until 2030, which limits its leverage in the meantime. During bargaining, SAG-AFTRA even floated what reporters nicknamed a studio tax on digital performers, a sign of how hard the union is working to attach a cost to synthetic characters it cannot fully block.

None of that directly governs a pre-production storyboarding tool, which is part of why A24 chose storyboards as an early target. But it sets the climate the partnership operates in. Editors, concept artists, visual effects workers, and storyboard artists are precisely the roles where assistive AI shades into replacement. A tool sold as a way to plan shots faster can also be a tool that needs fewer people to plan those shots. The announcement's careful language about preserving creative control reads, in this light, as an attempt to get ahead of that worry rather than a guarantee that it will not materialize.

The gap between the promise and the proof

The honest assessment of the Google DeepMind A24 deal is that it is a statement of intent backed by money, not a product anyone can evaluate yet. Every public detail describes process. Researchers and filmmakers will collaborate. Tools will be tested and refined. Goals will evolve. The things that would let an outsider judge the work, namely which models are involved, what the tools can do, and how output ownership and provenance are handled, are absent from the announcement on purpose, because they have not been decided.

That ambiguity is not necessarily a red flag. Genuine research partnerships are open-ended by nature, and pretending otherwise would be its own kind of spin. A24's insistence that the tools will avoid the prompt-to-finished-shot model is a meaningful design choice if the studio holds to it, because it locates the work in assistance rather than automation. The risk is that commercial pressure, on both a studio chasing efficiency and a lab chasing adoption, quietly pushes the tools past the line A24 is drawing today.

For now, the most useful way to read the deal is as a marker of where the industry is heading. The biggest AI labs and the most credible studios are no longer circling each other. They are signing checks and sharing workflows. The first real test will be whether anything ships that a director chooses to use without being told to, and whether the credits on an A24 release start naming the tools and the people who ran them. Until then, the partnership is a promise about process, and process is the easiest thing in the world to announce and the hardest to hold to once a release schedule and a return on $75 million start applying pressure. The next year will show whether a deal framed around respecting artists produces tools that working filmmakers actually want, or whether the gap between the promise and the proof turns out to be wider than $75 million can close. Readers tracking how these arrangements develop can follow our ongoing coverage of AI and the creative industries as the first tools emerge.