Vol. 1 · Edition 025Free · No paywall

Everyone Needs a Samwise

AI news · Synthesized · Opinionated · 🌿

$2.1M
lifetime revenue
vs est. $15M/day in costs · OpenAI Sora · 6 months
Industry
By Sam Taylor with Samwise

On how Sora's compute costs buried its revenue, why Disney walked away from a billion-dollar deal in under an hour, and what the Sora math says about the apps you're already paying for.

OpenAI's video tool made $2.1 million total. It cost an estimated $15 million a day. The math is about all of AI.

Source lean on this story
▲ avg

Anti-AI

00

Skeptic

02

Neutral

00

Pro (practical)

01

Pro (hyped)

00

← Anti-AI · Pro-AI →

If you saw someone share an AI-made video late last year — a quick clip of a cartoon dog doing something absurd, or a beach at sunset that looked almost real — there's a chance it was made with Sora. OpenAI's video tool had a big viral moment in November 2025. A lot of people downloaded the app. And then, quietly, the app shut down in April 2026.

Here's the number that explains why: Sora generated roughly $2.1 million in total lifetime revenue across its entire run. Against daily operating costs that the Wall Street Journal reported at $1 million a day — and that some internal estimates put as high as $15 million. Every day, Sora cost more to run than it earned in its entire existence.

That math isn't sustainable. And the story of why it happened is more interesting than "OpenAI made a bad product."

What actually happened

Sora launched publicly in September 2025, and then really launched in November 2025 — the version that got viral. 9.6 million people downloaded the app across iOS and Android. Revenue peaked in December 2025 at roughly $540,000 for the month.

Then it dropped. By January 2026, monthly revenue had slid to about $367,000. Downloads had fallen nearly 66% from their November peak. Active users went from over a million to under 500,000.

The two things that happened simultaneously: novelty faded, and competitors got better and cheaper. Runway Gen-4 and Kling 2.0 were producing comparable video at lower per-clip costs. Sora was competing with tools that had a head start on efficiency.

9.6M
Total downloads across iOS and Android before shutdown

→ Source: Tech Insider, citing Appfigures data

On March 24, 2026, OpenAI announced Sora would shut down. The app and website went dark April 26, 2026. The API — the version developers use to build with Sora — will shut down September 24, 2026.

The Disney part is the most revealing piece

The same week OpenAI announced the Sora shutdown, it also announced that Disney's planned $1 billion investment and character licensing deal had collapsed. No money changed hands.

Disney had pledged roughly a billion dollars to OpenAI, tied to a deal that would have let OpenAI build AI tools using Disney characters. Disney reportedly learned the Sora deal was being pulled less than an hour before the public announcement.

Read that again. Disney made a billion-dollar investment decision around a product that OpenAI killed without advance notice to the partner. Major AI partnerships can unravel in under an hour.

Sora's revenue vs. costs
MetricFigure
Lifetime revenue (full run)~$2.1M
Peak monthly revenue (Dec 2025)~$540K
Estimated daily operating cost (WSJ)~$1M/day
Estimated daily operating cost (internal est.)Up to $15M/day
Total downloads9.6M
App shutdown dateApril 26, 2026
API shutdown dateSeptember 24, 2026

What does OpenAI say?

OpenAI told the BBC it discontinued Sora to focus on other priorities, specifically mentioning robotics. Sora's underlying team has reportedly been redirected toward a robotics model codenamed "Spud," targeting a public release around July 2026.

The company has not confirmed specific revenue or cost figures publicly. The numbers above come from third-party data providers and reporting — Appfigures for revenue, and multiple outlets citing internal estimates for costs. The revenue figure is well-sourced; the cost figures have a wide range, which is itself informative. Even the lower estimate — $1 million per day — means Sora lost money every single day it operated.

There's also the IPO context. OpenAI filed its S-1 confidentially with the SEC in May 2026, with a target valuation around $830 billion. A product burning $1M or more per day with $2.1M total revenue doesn't look good on a prospectus. The shutdown timing and the IPO filing are not unconnected.

Source spread

What's real

  • Sora is a real technological achievement. AI video at that quality six months ago required specialized hardware and studios. Sora put it on a phone. The technology worked.
  • The shutdown announcement was handled honestly — specific dates, clear policy on what happens to your data, no vague "transition period." That's worth noting because many product shutdowns are deliberately obscure about timelines.
  • The Disney amount: $1 billion is a news headline, not a real loss. No money changed hands. Disney lost a planned investment, not cash it had already committed.

What deserves a side-eye

  • The cost-to-revenue ratio is staggering at any estimate. Even using the WSJ's conservative $1M/day figure, Sora cost more in a single day than it earned in its peak month.
  • OpenAI gave Disney less than an hour of warning. For a billion-dollar partner. That's not how you treat a major investor, and it reveals something about how quickly these decisions get made internally.
  • The "focusing on robotics" explanation is plausible and probably true — and also the most flattering way to describe a product being shut down for bad economics.
  • The costs range wildly in the reporting. "$1M/day" and "$15M/day" are not minor rounding differences. OpenAI has not clarified. That gap matters if you're trying to understand how much AI products actually cost to run.

What to do about it

  • Check your AI subscriptions. If you're paying for an AI video generation service right now, look at who owns it and how long it's been around. Products in this category have shorter runways than AI chat tools. Sora was six months. That's not the norm but it's not impossible.
  • Don't build a creative workflow you can't export from. If you're using any AI video tool for work, make sure you can download your outputs in a standard format. When a product shuts down, you get a deadline, not a warning.
  • The $20/month ChatGPT or Claude subscription is probably more durable. Text-based AI chat has better unit economics than video. If you're going to pay for one AI subscription, that category has proven it can sustain itself.
  • Treat "Disney-scale partnerships" as a lagging indicator, not a leading one. Big companies announced partnerships with Sora. Then Sora shut down. Enterprise deals take months to close; AI product cycles can be faster than that. The announcement of a partnership is not the same as the product being around in two years.

Further reading

🌿

Liked this? Get the weekly digest.

Free. Monday mornings. The week's stories, synthesized. Unsubscribe anytime.

Your take

How'd I do on this one?

What did I miss?

Tell Samwise (and Sam).

Disagree with the take? Spotted a fact I got wrong? Have context I should have included? Drop it here. Anonymous unless you leave an email.