Both China and US Tech Giants Are Running the Steel Playbook on AI Film — and Most Producers Don't See It Coming

I've watched this movie before.

In 2014, Amazon was selling ebooks at $9.99 — below cost, subsidized by Kindle hardware margins — and the Department of Justice sued the publishers for trying to stop them. The predator was legally protected. The defense was prosecuted. By the time the dust settled, Amazon controlled 80% of the ebook market, set the pricing floor, and extracted 30% of every sale through a distribution channel they'd built while everyone else was bleeding.

Now I'm watching the same sequence in AI film production. Except this time there are two sets of players running the playbook simultaneously — Chinese AI labs and American tech giants — and traditional film producers are caught in the middle of both.

The Steel Playbook Has Three Phases

Amazon ran it on publishing. Spotify ran it on music. Now both Chinese AI labs and American tech giants are running it simultaneously on film production. NYU Stern professor Scott Galloway calls it "modern-day dumping." The structure is simple:

Phase 1 — Price destruction. Enter the market below cost, subsidized by state capital or strategic patience. Force incumbents to match or lose customers. Margins collapse across the industry. Weak players exit.

Phase 2 — Infrastructure capture. While competitors are bleeding, build distribution lock-in. Amazon captured the Kindle ecosystem — device, store, format, discovery algorithm. Spotify captured the listener relationship and made the label irrelevant to the fan. In film, OpenAI's Sora is capturing the creative workflow. Google's Veo is embedded in YouTube's production infrastructure. DeepSeek and Kling are capturing the enterprise API routing. The infrastructure is the real asset. The cheap product is just the bait.

Phase 3 — Extraction. Once dependency is established, pricing power returns — but only to the player who owns the infrastructure. Amazon now charges publishers 70/30 revenue splits. Spotify pays artists fractions of a cent per stream while capturing the full ad and subscription margin. The question for film isn't whether extraction happens — it's which player extracts, American or Chinese.

The film industry is entering Phase 1 right now, getting squeezed from both directions simultaneously.

What "Cheap AI Film" Actually Is — And Who's Selling It

Here's what most producers get wrong: they think this is a China problem. It's not. It's a capital concentration problem, and the capital is coming from both sides of the Pacific.

Chinese tools — Kling, Hailuo, and their successors — are priced at near-zero, subsidized by High-Flyer Capital and underwritten by the Chinese state's geopolitical mandate. There is no natural profit floor on that side. They can dump indefinitely.

But American companies are doing the same thing with different financing. OpenAI's Sora, Google's Veo, Meta's video generation tools, and Runway — all US-based — are also priced below sustainable cost, subsidized by venture capital, cloud revenue cross-subsidies, and the trillion-dollar infrastructure bets their parent companies have already committed. OpenAI lost billions in 2024 and 2025 building the model capacity that Sora runs on. That cost is already sunk. Sora can be priced near-zero because the build cost is amortized across the entire OpenAI product line.

The producer looking at cheap AI film tools and thinking "at least the American ones are on my side" has misread the situation entirely. OpenAI's business model requires capturing the creative workflow layer. So does Google's. The extraction phase looks identical whether the player is in Beijing or San Francisco.

The compute cost floor is real, and it doesn't disappear when the subsidies do — on either side.

AI film production doesn't eliminate cost. It shifts cost from human labor to capital inputs — GPUs, electricity, data center overhead, cooling. Those inputs are priced by physics and economics, not by a startup's burn rate. A 90-minute feature at cinematic quality requires enormous compute: character consistency across thousands of frames, narrative coherence, physics, lighting, emotional performance. That is not a cheap inference job.

The pricing gap you see today — "AI film = almost free, crew = expensive" — narrows dramatically once the subsidy lifts. What remains is a speed advantage on iteration, not a permanent cost advantage on final output quality. The "free film" era is a temporary Phase 1 condition.

Why 2028 Is the Real Inflection Point

Current tools (Runway Gen-4, Veo 3.1, Kling 3.0) cannot produce narrative-coherent feature-length films. Character consistency fails across 90 minutes. Temporal coherence breaks down in complex scenes. The first commercially viable, fully AI-generated features are projected to arrive by late 2028.

That's the Phase 2 to Phase 3 transition for AI film production. Before 2028, you're in the slop flood — massive quantities of low-quality AI content competing for the same FAST channel shelf space as your catalog. After 2028, you're competing with AI content that is genuinely watchable at the feature length.

The window between now and 2028 is the most valuable period for human-produced catalog. Not because the threat isn't real — it is — but because the platforms are already responding, and they're responding in your favor.

The Platforms Are Already Moving

YouTube's January 2026 enforcement wave wiped 4.7 billion views of AI-generated content in a single action. YouTube CEO Neal Mohan explicitly stated the platform would prioritize content demonstrating genuine human creativity. FAST platforms are building the same curation infrastructure.

The enforcement pattern is consistent: mass-produced AI template content, undisclosed deepfakes, AI voiceover pipelines with no human creative input — all being flagged, demonetized, or removed. The platforms have a business reason to maintain quality signal. Advertisers won't pay CPMs for AI slop, and audiences don't return to channels that feel synthetic.

This means the flood of cheap AI content is being actively separated from human-produced catalog — and that separation works in catalog owners' favor right now.

The Copyright Moat Nobody Is Talking About

The US Copyright Office's 2026 standard: AI-generated content is not copyrightable unless human authorship is "significant enough" to qualify as original work.

This is the structural advantage that doesn't erode when AI quality improves.

Your existing human-produced catalog carries full copyright. AI-generated films carry no copyright, or partial copyright at best. Every FAST channel deal, every licensing agreement, every sync placement, every international distribution contract requires a clean rights chain. AI content can't be licensed the same way. It can't be sold to a broadcaster in Germany. It can't be cleared for a streaming deal in South Korea. It can't anchor a brand partnership without a human authorship claim.

When AI film quality reaches parity with your catalog in 2028, the rights situation won't have changed. Human-produced content will still be the only content with full, licensable copyright. That's not a temporary advantage — it's a structural moat built into intellectual property law.

"Holding Out" Is the Wrong Frame

Most producers I talk to are framing this as a defensive posture: hold out until the AI threat passes, protect catalog, wait for the market to stabilize.

That's exactly backward.

The AI content flood is creating your distribution opportunity, not threatening it. Platforms enforcing quality curation are actively seeking authenticated human-produced catalog to fill the space being vacated by AI slop. The branded owned-and-operated FAST channel strategy — launching your own channel on Tubi, Pluto, or Peacock Free with 50+ hours of catalog rather than licensing piecemeal — is generating primary-market revenue right now, not secondary-market residuals.

The differentiation premium for human-produced catalog is at its maximum value in the 18 months between now and late 2027, before AI feature quality becomes genuinely competitive. That premium starts compressing in 2028.

You're not holding out. You're late to distribute.

The Convergence Nobody Expected

Here's the structural insight that ties this together: the same AI capability maturation that will eventually threaten your catalog is the same event driving the US AI infrastructure bubble toward correction.

The $725 billion in AI capex committed by US cloud providers was justified by frontier capability premium pricing. DeepSeek commoditizing 80% of AI tasks at 97% below US pricing is the event that collapses the revenue case for that infrastructure spend. When enterprise CFOs start asking why they're paying OpenAI rates for tasks a Chinese model handles at near-zero cost, the valuation correction begins.

That correction window — projected for 2027 to 2028 — is the same moment AI film quality reaches feature-length viability. The same forces that threaten your catalog are the forces that generate the capital event you can deploy into infrastructure plays in markets where the AI disruption cycle hasn't arrived yet.

The film catalog generates cash during the slop flood. The copyright moat protects licensing value through the quality parity phase. The capital deployed during the AI bubble correction funds the next position.

You don't hold out. You run the clock.

The playbook worked on publishing because publishers thought Amazon was a retailer, not a competitor. It worked on music because artists thought Spotify was exposure, not extraction. Both industries are still trying to claw back what they gave away.

The difference in film is that the threat is coming from every direction at once — Beijing and San Francisco both running Phase 1 simultaneously. The producers who survive won't be the ones who pick a side. They'll be the ones who understood that neither side was ever on theirs.