TBPN

What does Netflix tell us about AI costs in Hollywood?

July 21, 2026

Happy Tuesday.

We’re live on YouTube and 𝕏.

The current thing in tech and business is Gemini releases 3.6 Flash.

Today’s Lineup

Augustus Co-Founder & CEO Ferdinand Dabitz at 11:55 AM
Lightphone Co-Founder & CEO Kaiwei Tang at 12:05 PM
Substack Co-Founder & CEO Chris Best at 12:15 PM
Whop Co-Founder & CEO Steven Schwartz at 12:35 PM
Business Insider Editor at Large Katie Roof at 1:00 PM
Dimension Founder Zavain Dar at 1:15 PM
CuspAI Co-Founder Chad Edwards at 1:25 PM

Run of ShowWhat does Netflix tell us about AI costs in Hollywood?, by John Coogan

Netflix is down ~10% in the last week after announcing that they expect both revenue and profit to slow in the third quarter. Investors are now worried that we’re past peak Netflix growth. Now, the company is still growing, 11.7%, but this is the smallest YoY increase since late 2023. The bull case is that Netflix only has 5% TV market share globally. The bear case is increased competition from social media, vertical video, free to play games, and maybe AI content in the future.

Eric Seufert over at Mobile Dev Memo is not happy about the decrease in detailed KPI sharing from Netflix over the years (great read) and over at Stratechery there are two pieces from Ben and Andrew studying the fall off of Netflix (and of course debating it honestly).

The big question I have today is less about Netflix directly and more focused on what we can learn about the impact of AI adoption in Hollywood broadly. I still think of Netflix as a tech company first, even though they are very much a Hollywood entity at this point. They should be early adopters of AI and indeed have already been using AI in productions. Variety recently reported that Netflix used generative AI across roughly 300 movies and TV shows during the first half of 2026. Some AI is used in the early stages of storyboarding, some is sneaking into post-production. Their new docuseries about the American Revolution included 17 minutes of AI-enhanced footage that expanded the scale of the project and would not have been financially feasible using traditional production methods. It’s not the most popular thing online, but I think the revealed preference will clearly be that tasteful use of AI in filmmaking is widely accepted.

Netflix told shareholders directly: “We are increasingly leveraging these tools to deliver higher-quality output more quickly and at a lower cost than traditional methods,”

So what’s the actual impact? There were some bold predictions about AI’s impact on Hollywood. Jeffrey Katzenberg predicted that AI could reduce the labor and cost required for a major animated film by around 90%. Tyler Perry paused an $800 million studio expansion after concluding that generated environments could reduce the need for sets, locations and soundstages. But on the flip side, Ben Affleck argued that AI would first attack the “laborious, less creative and more costly” components of filmmaking—particularly VFX—rather than replacing actors, directors or the entire creative process. He also predicted that lower costs would produce more content and greater demand, basically Jevons Paradox.

This seems to be what’s playing out at Netflix. Since ChatGPT launched, Netflix has grown its overall headcount from 12,800 to 16,000. Also, costs broadly are increasing, Q2 2026 showed Technology and Development growing 22% and G&A growing 13%. Operating margin actually declined slightly from 34.1% to 33.4%. Netflix explicitly attributed the margin decline primarily to technology and sales-and-marketing expenses growing faster than revenue.

Now, Netflix doesn’t break out spend on AI tools or anything particularly granular about production expense categories so it’s hard to tell what’s going on inside the organization exactly, but it looks like a slow takeoff to me.

It does feel like the organizational design of large studios will make them slower to adopt AI tools, and this feels like a big opportunity for younger, more entrepreneurial creators. Backrooms was built on top of Blender, and Obsession was apparently fully pre-vised in CGI as well, despite ultimately being shot using traditional, albeit modern, cameras. It’s very funny, especially with the backdrop of Christopher Nolan stressing the value of “real” traditional filmmaking techniques, like… actual film (vs digital). Using CGI to build the world of Backrooms, or previs the production of Obsession is lauded. You can imagine the next Christopher priding himself on using only “real greenscreens.” Ultimately the audience will decide what’s entertaining though. - John

Substack Cracks Down on AI Writing

Substack is partnering with Pangram, an AI-content detector, to launch an AI-detection tool that will let readers scan posts and replies to estimate how much of the text was written by the actual author versus with AI.

In a recent blog post by Chris Best, Substack Co-Founder & CEO, he explains that the core problem is:

“[N]ot people using AI, or the quality of its output. Not everything made with AI is slop, and not all slop is made with AI. The problem is when there is a mismatch between a reader’s expectation and reality, especially when they unwittingly invest their attention in something with no human thought on the other end. That’s Claudefishing.”

The new tool will work with text that is longer than 100 words and will only show an analysis of how AI-generated a text is if the reader requests it. Authors will also be able to preview drafts before publishing, challenge content that is flagged as AI-generated, and add a “How I Made This” statement.

A large focus of the blog post is Substack saying they don’t oppose AI in writing. The company is mainly concerned whether readers are being deceived into thinking they’re reading something written by a human, when it was actually written primarily, if not entirely, by ChatGPT or Claude.

“When content made by no one takes over parts of the internet that are supposed to be human, it pollutes the commons and makes it hard to discover and hear human voices. When readers have to wonder if what they’re reading is real, it undermines trust in authorship and threatens the livelihood of writers—including those who use AI tools thoughtfully to produce work they believe in. Platforms that reward fakeness will create a race to the bottom.”

Clip Spotlight: Tyler Cowen thinks The Odyssey was a disappointment.

"I thought the dialoge was lame. I don't mind the cross-ethnic casting at all. I just didn't think they were the right people. Matt Damon is not Odysseus. He's just not impressive enough physically. The woman who played Calypso just seemed like a suburban housewife. I love Zendaya, but she didn't have any real role. So they screwed up many things."

"Compared to other movies, it's fantastic and it's dreamy and it's impressive. But mostly I was disappointed."

Headlines

Gemini releases 3.6 Flash

Reuters: US judge approves Anthropic’s $1.5 billion settlement of copyright lawsuit

WSJ: Top American AI Execs Sound Alarm on Chinese Models

Washington Post: The biggest winners of the American economy fear they’re sinking fast

FT: Oracle could face $7bn collateral bill for Wisconsin data centre

Bloomberg: OpenAI’s Altman to Brief US Officials on Next Wave of AI Models

TechCrunch: Colossal Biosciences reportedly in talks to raise new capital at $20B–$30B valuation

Lisan al Gaib’s Substack: Have Chinese AI Models Caught Up to the US Frontier?

OpenAI: Safety and alignment in an era of long-horizon models

FT: Anduril launches helicopter strike drone

The Atlantic: The DSA Is a Parasite