Originally published at twarx.com - read the full interactive version there.
Last Updated: June 22, 2026
Google Investing in Backrooms Studio A24 is not a movie bet — it is the acquisition of something far more valuable than box office returns: a legally clean, culturally authoritative pipeline of real-world creative data that no synthetic dataset can replicate. The roughly $75 million going into A24 is the most important AI infrastructure deal of 2026, and almost everyone is reporting it wrong.
The deal: Google is putting about $75 million into A24 — the studio behind Backrooms, Everything Everywhere All at Once, and Hereditary — as part of an explicit artificial-intelligence research partnership. It is Google's first direct stake in a film studio. It lands while Veo and OpenAI's Sora are actively racing for video-generation dominance — which is not a coincidence.
By the end of this piece you'll understand the deal's mechanics, the data strategy underneath it, how to access Google's creative AI stack today, and exactly who wins and loses.
The Google-A24 partnership pairs a $2 trillion AI platform with a prestige independent studio — the first move in what we call the Prestige Data Moat. Source
Coined Framework
The Prestige Data Moat — the strategic acquisition of culturally authoritative, rights-clean creative content and production workflows as proprietary training infrastructure, making entertainment investment the new AI R&D spend
It names the shift from scraping the open web to buying access to high-quality, legally defensible creative reasoning. The systemic problem it solves: the best training data for generative film models was never online — it lives inside studio workflows that no competitor can scrape.
What Was Announced: The Google–A24 Deal Explained
Official Announcement Date, Sources, and Confirmed Facts
The Wall Street Journal first broke the story, reporting that the search giant is putting about $75 million into the film company as part of an artificial-intelligence research partnership. That single sentence is the confirmed ground truth — everything beyond it is informed analysis, which we clearly label throughout.
Coverage spread the same day across Yahoo Finance, TheWrap, Reuters, and Investing.com. The confirmed facts are narrow, and worth stating plainly: (1) the investor is Google; (2) the recipient is A24; (3) the amount is approximately $75 million; (4) the structure is an AI research partnership, not a passive financial play. Four facts. That's the foundation everything else is built on.
The $75 Million Figure: Investment Structure and Stake Details
A24 was valued at roughly $2.5 billion in its 2022 fundraising round. At that valuation, $75 million implies a minority equity position — likely under 5%. (Analysis: the precise stake hasn't been disclosed; this is a valuation-implied estimate, not a confirmed figure.) But the percentage isn't the point. The capital is bundled with research access. That bundle is everything.
Direct Quotes from Google and A24 Spokespeople
As of publication, neither company has released expansive public statements beyond confirming the partnership's existence to the WSJ. We'll update this section as official quotes emerge. For now, the WSJ's framing is the authoritative source and I'd treat anything else as inference.
Google didn't buy a piece of A24 for the dividends. It bought a clean-rights window into how great films are actually made — and that window is worth far more than $75 million.
~$75M
Google's investment in A24
[WSJ, 2026](https://www.wsj.com/tech/ai/google-investing-in-backrooms-studio-a24-e7585ebe)
~$2.5B
A24 valuation (2022 round)
[TheWrap, 2022](https://www.thewrap.com/)
$1.5B+
A24 lifetime global box office
[Box Office Mojo, 2025](https://www.boxofficemojo.com/)
What Is A24 and Why Does It Matter to Google's AI Strategy
A24's Content Library, Cultural Authority, and Production Model
Founded in 2012, A24 has done over $1.5 billion at the global box office across a catalog defined by narrative complexity and visual distinctiveness — Everything Everywhere All at Once, Midsommar, Hereditary, The Whale. A24's brand isn't scale. It's taste. And that taste is encoded in thousands of granular human decisions: which shot, which cut, which sound, which silence. Most studios make movies. A24 makes decisions that other studios wouldn't.
For an AI lab, those decisions are the asset. The web is full of finished films. It's almost entirely empty of the decision trees that produced them — and that's exactly what a research partnership can provide.
The Backrooms Project: What It Is and Why Google Is Attached to It
A24's Backrooms film is built on the viral internet horror mythology of endless liminal spaces. It's a rare case of AI-adjacent, internet-native cultural IP meeting prestige filmmaking — which is why it's become the headline hook for this deal. The thematic resonance is almost too neat: a film about uncanny, generated-feeling environments, financed in part by the company building the world's most advanced video-generation models. You couldn't script a better PR alignment.
Why Independent Studios Are More Valuable Than Major Studios for AI Partnerships
Major studios sit on tangled rights — guild residuals, distribution carve-outs, third-party licenses accumulated over decades. Independent studios like A24 control their IP far more cleanly, which dramatically reduces the legal friction of using production materials in an AI research context. (Analysis: clean rights are the single most underrated variable in AI training today — I've watched deals collapse over exactly this.) A24's artisan-scale, talent-driven workflow is also small enough to instrument and observe. That's a feature, not a limitation, for research purposes.
The most valuable AI training data in film isn't the finished movie — it's the 400 rejected takes, the color-grade revisions, and the editorial notes that explain why a cut works. A24 owns that. Google just rented access to it.
How prestige production workflows convert into proprietary training infrastructure — the mechanism behind the Prestige Data Moat.
How the AI Research Partnership Works: Full Capability Breakdown
What 'AI Research Partnership' Actually Means in Contractual Terms
The phrase signals reciprocity: data and workflow access flowing one way, tooling and capital flowing the other. This is structurally different from a passive venture check — the word 'research' is doing real work in that sentence. (Analysis: the WSJ confirms the partnership framing; specific contractual data-access terms aren't public.) It echoes Google's earlier strategic AI bets, including its multi-tranche investment in Anthropic. The pattern is consistent: control the pipeline, shape the model.
Which Google AI Tools and Divisions Are Likely Involved
Google DeepMind and Google Labs are the most probable internal owners, given their active work on multimodal generative video and audio. The connective tissue here resembles classic multi-agent system design: specialized models handling script understanding, shot composition, and audio, coordinated by an orchestration layer — conceptually similar to how LangGraph and AutoGen coordinate agents in enterprise pipelines. Different domain, same architecture problem. Builders mapping this pattern can study our breakdown of AI agent architecture for the orchestration primitives involved.
The Prestige Data Moat: Why This Is Infrastructure, Not Entertainment
A24's production corpus — scripts, storyboards, shot compositions, sound-design choices, editorial workflows — is a rights-clean record of elite human creative reasoning. You can't scrape it. You can't synthesize it. You can only partner for it.
Coined Framework
The Prestige Data Moat in practice
When raw web data hits diminishing returns, the next frontier is curated, authoritative, consented data behind institutional walls. Google's A24 stake is the first publicly visible move to acquire that frontier in entertainment.
The Prestige Data Moat: From A24 Workflow to Google Multimodal Model
1
**A24 Production Data (Source)**
Scripts, storyboards, dailies, edit decision lists, sound-design notes. Inputs: rights-clean, human-authored. The corpus no competitor can scrape.
↓
2
**Consent & Rights Layer (Legal Gate)**
Partnership contract defines what data is usable for research. This gate is what direct scraping or open licensing cannot replicate.
↓
3
**Annotation & Structuring (RAG-style indexing)**
Creative decisions are labeled, embedded into vector databases, and made queryable — turning tacit craft into structured training signal.
↓
4
**Multimodal Model Training (Veo lineage)**
DeepMind fine-tunes video/audio models on the structured corpus. Output: improved narrative and emotional coherence.
↓
5
**Tooling Returns to A24 (Reciprocity)**
A24 gains pre-production and VFX tools. The loop closes — capital, data, and tooling circulate inside one moat.
The sequence matters: the legal gate at step 2 is the actual competitive advantage — it is what money, not scraping, buys.
Google's AI Filmmaking Tools: Production-Ready vs Still Experimental
Google Veo: Current Video Generation Capabilities
Production-ready: Veo 3, announced at Google I/O 2025, generates high-definition clips with native audio — the most advanced publicly demonstrated video model of its generation. It's accessible through DeepMind's VideoFX and the Gemini API for select partners. I'd call it genuinely impressive within its coherence window, which is the catch we'll get to.
Gemini Multimodal Models in Creative Production Contexts
Production-ready: Gemini 2.5 Pro handles script analysis, breakdowns, and shot-planning via the Google AI Studio API. These are reliable today for pre-production reasoning tasks. Ship them with confidence for that use case.
What Remains Experimental
Experimental / unsolved: long-form narrative coherence — maintaining character, plot, and visual consistency across a 90-minute film — is not solved by any current system, including Sora. I would not ship any of these tools for autonomous long-form generation. (Analysis: the A24 partnership most plausibly targets exactly this gap, supplying the failure cases and decision trees needed to extend coherence beyond the current ceiling.)
Every AI video model today is a brilliant 8-second sprinter that collapses at minute two. Whoever solves narrative coherence first owns film production. Google just bought the training data to try.
~8s
Typical coherent clip length, current SOTA video models
[Google DeepMind, 2025](https://deepmind.google/models/veo/)
26%
Entertainment production tasks automatable by GenAI within a decade
[Goldman Sachs, 2023](https://www.goldmansachs.com/intelligence/)
$0.35/s
Approx. Veo enterprise generation pricing (Vertex AI, 2025)
[Google Cloud, 2025](https://cloud.google.com/vertex-ai/pricing)
[
▶
Watch on YouTube
Google DeepMind Veo 3 video generation with native audio
Google DeepMind • Veo multimodal demos
](https://www.youtube.com/results?search_query=google+deepmind+veo+3+video+generation)
How to Access Google's AI Creative Tools: Pricing and Step-by-Step Guide
Current Access Pathways
Veo is accessible via Google Labs (VideoFX waitlist) and Vertex AI for enterprise. Gemini models run through Google AI Studio with a free tier and paid API access. The waitlist for VideoFX is real — don't expect same-day access if you're not already in.
Pricing Tiers
Google AI Studio: free tier with usage limits; paid Gemini access from approximately $3.50 per 1M input tokens (Gemini Pro tier, 2025 pricing).
Vertex AI (Veo): pay-per-use video generation, approximately $0.35 per second of generated video (2025 public pricing) — which adds up fast if you're iterating on scenes.
Enterprise partnership (A24-class): bespoke contracts negotiated directly with Google Cloud and DeepMind — not a public tier, and not something you're going to replicate on a startup budget.
Step-by-Step: Worked Demonstration
Here's a real, runnable pre-production workflow a small studio can start today — using Gemini for a shot breakdown.
python — Gemini API shot breakdown
pip install google-generativeai
import google.generativeai as genai
genai.configure(api_key='YOUR_API_KEY') # from aistudio.google.com
model = genai.GenerativeModel('gemini-2.5-pro')
scene = '''INT. ABANDONED OFFICE - NIGHT
Maya walks the endless yellow hallway. Buzzing fluorescent lights.
She stops. A door that wasn't there before.'''
prompt = f'''You are a 1st AD. Break this scene into a shot list.
For each shot give: shot size, camera move, lens, sound cue.
Scene:\n{scene}'''
response = model.generate_content(prompt)
print(response.text)
Actual output (abridged):
output
SHOT 1 — Wide | slow dolly-in | 24mm | low fluorescent hum, footstep echo
SHOT 2 — Medium tracking | handheld follow | 35mm | breathing, flicker buzz
SHOT 3 — Insert (the door) | static | 50mm | silence drop, single bass thud
SHOT 4 — Close on Maya | push-in | 85mm | held breath, rising tone
From that breakdown, a team can pipe each shot description into Veo for animatic previews — exactly the pre-production loop the A24 partnership aims to industrialize at scale. For teams building automated production pipelines, you can explore our AI agent library for ready-made orchestration patterns, browse production-ready agent templates, and pair them with n8n for workflow automation.
A worked pre-production loop: Gemini generates the shot list, Veo previews each shot — the practical core of AI filmmaking tools today.
When to Use Google AI Tools for Film vs Alternatives
Decision Framework for Filmmakers
Choose based on three axes: coherence, cost, and ecosystem integration. No single tool wins all three right now.
ToolNative AudioEditor IntegrationStudio PartnershipBest For
Google Veo 3YesVertex AI / APIA24 (~$75M)Studio-grade AV coherence
OpenAI SoraNo (limited)StandaloneNone confirmedPhotoreal short clips
Runway Gen-3PartialPremiere / ResolveNoneWorking editors & VFX
Pika 2.0PartialStandaloneNoneFast, cheap iteration
When Google Wins
Audio-visual coherence and enterprise API reliability — both critical for studio pipelines that can't afford a generation failing mid-sequence. (Analysis.)
When Alternatives Win
For indie filmmakers under a $50K budget, Runway Gen-3 or Pika 2.0 deliver faster iteration and lower per-generation cost than Veo enterprise pricing. Runway's timeline integration with Premiere and DaVinci Resolve is, right now, the production-ready choice for working editors. That's not a knock on Veo — it's just budget reality.
If your budget is under $50K, Veo enterprise pricing will bleed you. Runway Gen-3 inside DaVinci Resolve gets you 80% of the result at a fraction of the per-second cost. Pick tools by budget tier, not hype.
Competitive Landscape: How Google's A24 Deal Compares to Rival Moves
OpenAI and Sora
As of June 2026, OpenAI has no confirmed direct investment in a film or TV studio, despite Sora's public demos. That leaves a strategic gap Google has now filled with the data-access layer Sora simply doesn't have.
Meta AI
Meta's entertainment AI focuses on social video generation and smart-glasses capture — consumer-facing, not prestige production-facing. Different game entirely. For context on how the large labs are positioning, Meta AI Research publishes its generative video direction openly.
Amazon, Apple, and Microsoft
Amazon integrates AI into its Prime Video pipeline internally but has made no comparable external studio investment. Microsoft's play runs through its OpenAI stake. Apple Intelligence has no studio move of this structure — yet.
Why the A24 Deal Is Structurally Different
It stacks three layers no rival has combined: equity investment + research partnership + access to a rights-clean prestige library. That three-layer structure is the whole point. Any one of those alone is replicable. Together, they're not.
OpenAI has the most viral video model. Google has the studio. In the long run, distribution and data beat demos — and Google just locked up both in one $75M move.
Industry Impact: What This Deal Means for Hollywood, AI, and the Creative Economy
The Precedent Effect
If Google's model works, expect Apple and Microsoft to pursue studio-level creative-data partnerships within 18–24 months. (Prediction, grounded in the competitive pattern above.) The window to be first mover is closing quickly.
Labor and Guild Implications
The WGA and SAG-AFTRA secured AI disclosure and consent clauses in their 2023 strike agreements. The Google-A24 deal is the first real-world stress test of those clauses — do they hold against a $75M equity partner with explicit AI research interests? Nobody knows yet. The answer will define how every similar deal gets structured going forward. For the governance angle, our guide to AI governance covers how consent and audit logging should be operationalized.
The Independent Film Ecosystem
A24's brand is independence. The existential question nobody's fully answered: can a studio stay artistically independent when its infrastructure is being shaped by a $2 trillion platform? Goldman Sachs estimated in 2023 that GenAI could automate 26% of entertainment production tasks within a decade — this deal materially accelerates that timeline, whether A24's creative team wants it to or not.
❌
Mistake: Reading this as a movie investment
Analysts modeling $75M against box office ROI will conclude the deal is irrational. It isn't a content bet — it's an R&D spend disguised as equity.
✅
Fix: Model it against the cost of acquiring rights-clean training data — where $75M is cheap relative to scraping litigation risk.
❌
Mistake: Assuming Veo can already make a feature film
Headlines imply AI feature films are imminent. Narrative coherence past ~10 minutes is unsolved by every model including Sora.
✅
Fix: Deploy AI for pre-production, animatics, and VFX assists today — not autonomous long-form generation.
❌
Mistake: Ignoring guild consent clauses
Studios assuming a tech partner overrides 2023 SAG-AFTRA/WGA AI terms risk legal freezes on data use mid-project.
✅
Fix: Build explicit consent logging into the data pipeline — treat the consent gate as a first-class system component, not an afterthought.
Expert and Community Reactions to the Google–A24 Investment
Industry Analyst Takes
Showbiz411's Roger Friedman framed the deal as a signal that 'A24 is going AI,' noting the cultural dissonance of an auteur-driven studio taking money from a tech giant with explicit AI research motives. That tension isn't going away.
Creative Community Response
Reactions among filmmakers split sharply — and the split is honest. One camp sees Google's capital enabling riskier, more ambitious projects that wouldn't get greenlit otherwise. The other sees it as the final commodification of independent cinema. The A24 brand paradox — prestige built entirely on human craft, now partnered with automation infrastructure — is the emotional core of that debate, and I don't think either side is wrong.
Investor Sentiment
At A24's ~$2.5B valuation, $75M implies Google holds under 5% while gaining disproportionate research access. That structure is what AI-ethics researchers are flagging as potential regulatory arbitrage: obtaining creative data through equity investment that direct scraping or conventional licensing could never achieve. It's a legitimate concern and one the guilds will almost certainly test in court.
The quiet genius of the deal: under-5% equity, but potentially near-total visibility into how prestige films get made. Influence per dollar, not ownership per dollar, is the metric that matters here.
The A24 brand paradox visualized: prestige human craft on one side, generative AI tooling on the other — the cultural fault line this deal exposes.
What Comes Next: Roadmap, Predictions, and the Future of AI in Film Production
Expected Milestones in 2026 and 2027
The most likely near-term output: a co-developed AI production toolkit unveiled at Google I/O 2026, or a Sundance 2027 premiere of an A24 film that used Google AI in pre-production or VFX — a proof-of-concept timed for maximum press. (Prediction.) Both companies have strong incentives to stage a visible win early.
The Prestige Data Moat in Motion
If DeepMind successfully trains on A24's production data, a future Veo successor could be the first model to maintain narrative coherence past the 10-minute mark — a capability that would make every competitor's current demos look primitive by comparison. That's the strategic prize. That's the whole reason to call this infrastructure, not entertainment.
Coined Framework
The Prestige Data Moat as competitive endgame
Whoever assembles the deepest rights-clean creative corpus wins the coherence race. Google moved first; rivals must now buy their own studios or cede the frontier.
2026 H2
**First co-developed A24 x Google production toolkit teased**
Grounded in Google's I/O cadence and the explicit research-partnership framing in the WSJ report.
2027 H1
**Rival tech giant announces a studio-level data partnership**
Pattern-matched to Google's Anthropic and Gemini playbook; competitive pressure forces an Apple or Microsoft response within 18–24 months.
2027 H2
**First guild legal test of AI data-use under 2023 consent clauses**
SAG-AFTRA/WGA clauses meet a $75M AI equity partner — the inevitable collision over training-data consent.
2028
**Sub-10-minute coherent AI video becomes demonstrable**
Contingent on the A24 corpus closing today's ~8-second coherence ceiling reported by DeepMind.
Three Scenarios by 2028
Scenario 1 (Bull): The partnership yields breakthrough multimodal AI; Google becomes default studio-grade film infrastructure globally. Scenario 2 (Freeze): Guild legal challenges halt data use, leaving $75M a costly, inconclusive experiment. Scenario 3 (Backfire): A24's output declines under AI-integration pressure, the prestige signal erodes, and the deal becomes a cautionary tale. Regardless of which scenario lands, the boundary between AI infrastructure and entertainment has been permanently redrawn — rivals must respond or cede the creative-data frontier entirely. For builders, the transferable lesson echoes enterprise AI strategy: data access and orchestration beat raw model size, and RAG-style structured corpora plus AI agents built on vector databases and standards like MCP are how this actually gets operationalized.
The next AI arms race won't be fought over GPUs. It'll be fought over who can legally access the best human creative data. Google just fired the first shot — for $75 million.
Frequently Asked Questions
How much is Google investing in Backrooms studio A24 and what does the deal include?
Per the Wall Street Journal, Google is putting about $75 million into A24 as part of an artificial-intelligence research partnership. The deal bundles equity with research access — it is not a passive financial stake. At A24's roughly $2.5 billion 2022 valuation, $75M implies a minority position likely under 5%. The confirmed facts are narrow: investor (Google), recipient (A24), amount (~$75M), and structure (AI research partnership). Specific equity percentage, governance terms, and exact data-access provisions have not been publicly disclosed. Treat any figure beyond the $75M as valuation-implied estimation rather than confirmed fact until the companies release detailed statements.
What is the Google and A24 AI research partnership about?
The WSJ describes it explicitly as an AI research partnership — language signaling reciprocal data and workflow access rather than a venture check. The most defensible interpretation: Google gains structured insight into A24's rights-clean production workflows (scripts, storyboards, editorial decisions) to train multimodal video models, while A24 gains AI tooling for pre-production and VFX. The likely internal owner is Google DeepMind, given its work on video and audio generation. We call this strategy the Prestige Data Moat: acquiring culturally authoritative, legally clean creative data that cannot be scraped from the open web. Exact contractual data-use terms remain undisclosed as of publication.
Is A24 using AI to make movies now?
There is no confirmation that A24 is currently generating films with AI. The deal is a research partnership, and the realistic near-term applications are pre-production and post-production assists — shot breakdowns, animatics, and VFX — not autonomous film generation. No current AI system, including Google Veo or OpenAI Sora, can maintain narrative, character, and visual coherence across a feature-length runtime; coherence breaks down well before the 10-minute mark. So while A24 now has privileged access to Google's tooling, the practical use cases today are augmentation, not replacement. Any claim that A24 is making fully AI-generated movies is, as of June 2026, speculation rather than reported fact.
What AI tools is Google developing for film production?
Google's most relevant production-ready tools are Veo 3 for high-definition video generation with native audio, and Gemini 2.5 Pro for script analysis and shot planning. Veo is accessible through Google Labs' VideoFX and Vertex AI (around $0.35 per second of generated video at 2025 enterprise pricing); Gemini runs via Google AI Studio with a free tier and paid access from about $3.50 per 1M input tokens. Still experimental and unsolved: long-form narrative coherence, emotional consistency across scenes, and fully rights-clean generative audio. The A24 partnership most plausibly aims to supply the creative decision data needed to close these gaps in future model generations.
How does Google's A24 investment compare to OpenAI's Hollywood strategy?
As of June 2026, OpenAI has no confirmed direct investment in a film or TV studio, despite Sora's high-profile demos. Its strategy is model-first and partnership-light at the studio level. Google's A24 deal is structurally different because it stacks three layers: equity investment, an explicit research partnership, and access to a rights-clean prestige content library. That combination gives Google something Sora lacks — privileged, legally defensible creative training data and a production pipeline to observe. In short, OpenAI currently leads on viral video demos, while Google has moved to secure the data and distribution layer. Over a multi-year horizon, data access and studio integration tend to outlast demo virality.
What does the Google–A24 deal mean for SAG-AFTRA and WGA AI agreements?
The SAG-AFTRA and WGA secured AI disclosure and consent provisions in their 2023 strike agreements. The Google-A24 partnership is the first high-profile case that tests those clauses against a tech investor with explicit AI research interests. The central question is whether consent and disclosure requirements meaningfully constrain how production data tied to performers and writers can be used for model training. If guilds challenge the data-use terms, the research partnership could face legal friction or freezes. Practically, studios entering similar deals should build explicit, auditable consent logging into their data pipelines and treat the consent gate as a core system requirement, not a contractual afterthought.
Will the Backrooms movie use Google AI technology?
It has not been confirmed that A24's Backrooms film will use Google AI tools, and any specific claim to that effect is speculation. What is true: Backrooms is an A24 project rooted in internet-native horror mythology, which makes it a thematically resonant flagship for an AI-focused partnership — and a likely candidate for a proof-of-concept moment such as AI-assisted pre-production or VFX. Realistically, if Google AI touches the film, it would be in augmentation roles (animatics, environment design, VFX assists), not autonomous generation, given current coherence limits. Expect Google and A24 to stage any disclosed AI usage for maximum press impact, potentially around Google I/O 2026 or a festival premiere. Until then, treat it as unconfirmed.
About the Author
Rushil Shah
AI Systems Builder & Founder, Twarx
Rushil Shah is the founder of Twarx and an AI systems builder who has spent years designing autonomous workflows, multi-agent architectures, and AI-powered business tools. He writes from real implementation experience — covering what actually works in production, what fails at scale, and where the industry is heading next. His work focuses on making agentic AI practical for builders and businesses.
LinkedIn · Full Profile
This article was originally published on Twarx. Follow for daily deep dives on AI agents and automation.
Top comments (0)