Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

https://www.revolutioninai.com/2026/03/%20chatgpt-gpt-54-mini-silent-switch-march-2026.html

by vinodpandey7 • 4 months ago

40 50 4 months ago

But doesn't $15M/day of inference cost imply "demand" from users? If this is the case, it's just a matter of time until costs can be reduced.

abuani• 4 months ago

Costs were reduced to $0. Can't get better then that for a product that OpenAI had no clue how to monetize

newsclues• 4 months ago

I’ll eat lots of free samples at Costco of foods I’d never pay money for

bonesss• 4 months ago

A box that says "Free T-Shirts" can create near-infinite demand.

Paying near-infinity dollars for T-Shirts people want for $0 isn't a profitable business model.

Demand side price sensitivity impacts potential supply side margins.

_qib3• 4 months ago

Why would demand imply costs will be reduced? If you're making an economies of scale argument, there's plenty scale right now, and costs don't seem to be trending down.

MrGilbert2• 4 months ago

If this is the case, it's just a matter of time until costs can be reduced.

Is it, though? We cannot predict technological advancement, and the times of ̶M̶u̶r̶p̶h̶y̶'̶s̶ ̶L̶a̶w̶ Moore's Law* for computational power are long gone. There is simply no guarantee that the costs will go down enough.

* thanks lucianbr!

kukkamario• 4 months ago

I think there is plenty of room to make AI inference much more energy efficient. For example, there are companies testing creating custom silicon to run the model. Once that technology matures and we have some "good enough" models for normal use, inference cost for non-bleeding-edge models can come way down.

I don't expect bleeding-edge models to become any cheaper, but previous generation models can potentially be really cheap.

lucianbr• 4 months ago

Moore's Law.

The times for Murphy's Law for computational power are just beginning.

PunchyHamster2• 4 months ago

The "matter of time" is getting more and more expensive, not cheaper, at least for next 2 years

brazzy• 4 months ago

Not sure what you're referring to. If you're talking about inference cost for frontier models, that's going up because researchers keep pushing those frontiers, often without considering cost. And while they're subsidized (to gain market share), users have no reason NOT to use the crazy expensive frontier models.

Once the market consolidates, and users get used to the idea of using models that are "good enough" because frontier models are too expensive, there's no reason AI cannot be profitable.

written-beyond• 4 months ago

Yes but it was capped by their restriction on sign ups/registrations. That could've easily been in the hundreds of millions if the app had public signups.

Idk if Instagram would exist if they were spending hundreds of millions a day.

vinni2• 4 months ago

It could be over provisioned or that cost is supposed to be minimum cost with some minimum capacity which was never reached.

yifanl• 4 months ago

Disneyworld has lines longer than the park can manage for decades, do you expect it to just be a matter of time until park management finally figures out how to queue people efficiently enough, or do you think the solution will be once again raising costs for the customer.

pjc50• 4 months ago

The actual revenue was quoted at $2.1m .. total. Ever.

It would require multiple order of magnitude cost reductions to make that worthwhile. Maybe another few decades of Moore's law, if we have that left.

This was the Moviepass model of selling $10 bills for $9.

rsynnott2• 4 months ago

Much worse, really; it was selling $10 bills for half a cent.

The Moviepass thing, I think if you were kinda gullible you could maybe buy into it eventually working on scale. This could never work on scale.

hk__22• 4 months ago

But doesn't $15M/day of inference cost imply "demand" from users? If this is the case, it's just a matter of time until costs can be reduced.

If you build a website that gives $100 for free to each one of your users, you’ll quickly have "demand" but that’s not "a matter of time until costs can be reduced".

imron• 4 months ago

Make a loss on every sale but make it up in volume!

rsynnott2• 4 months ago

Costs would have to be reduced about 2,000 times just to break even, assuming that inference was the only cost, which of course it was not.

mingus882• 4 months ago

Many things are only in demand because they are free.

And if nobody is willing to pay for it, it hardly matters how low you bring down cost, because it’s always a net negative.

Just look at the general state of the internet over the past two decades. Do you think it would work for Sora to insert ads into the slop?

array4277• 4 months ago

Ha, that's a good one!

lajisam• 4 months ago

The website keeps refreshing for me

Bolwin• 4 months ago

This page reloads infinitely for me, can't see it

MrGilbert2• 4 months ago

I would assume that the economic reasoning, if looked at it without dollar bills covering their eyes, would apply to AI in general the way we are using it.

yanhangyhy• 4 months ago

strange seedance is not mentioned.

suck-my-spez• 4 months ago

Website built with AI, infinite loads

keiferski3• 4 months ago

I would be curious to know if there is actually as much business economic demand for AI video compared to images (logos, product graphics, etc.) or text (blogs, content everywhere, etc.)

My impression is that video is too complex to easily fit into an AI pipeline. Either you need something highly specific, like your own product’s UI. Or you need something personable and consistent, like someone talking into his camera.

spiderfarmer• 4 months ago

People are already using it to automate TikTok ad campaigns.

northernsausage2• 4 months ago

As a product photographer/videographer - No its not good enough to understand products so each scene its different, you can't storyboard or collaborate with it,. For high end products (where the money is), colour shape, scale matter and its just not consistent enough for professionals. For cheap tiktop slop products is fine because what arrives it never what you ordered anyway.

The files are a pig to try and edit as well, making them beyond the generation and prompt costs expensive. At that point you might as well go and just film the ad.

boredhedgehog• 4 months ago

The aspiration is to replace the movie industry. That's a lot of demand.

keiferski3• 4 months ago

But demand from whom? I feel like the biggest moneymakers in that industry are explicitly anti-AI.

General business stuff like content or images has demand from across the economy. “Replace Hollywood” is kind of a niche thing.

bot403• 4 months ago

As a movie consumer I am not interested in AI movies. You don't get to just keep the existing market and switch to AI. You are creating a new market of AI video consumers and hoping it's big enough.

northernsausage2• 4 months ago

Story heavily edited by AI about an AI company with an AI product that makes AI videos that is closing so they can spend money on some other AI product. All seasoned with some AI goop images. I hate the future.

keiferski3• 4 months ago

You forgot the AI readers and AI commenters.

lionkor• 4 months ago

vibe coded website too -- I waited 4 seconds for it to load

Glawen• 4 months ago

At least it loaded, it did not on my side.

Maken• 4 months ago

I specially like the graph-looking picture with no axis, no data and no context.

hk__22• 4 months ago

The context is the text below and under it. The axis doesn’t matter, it’s the shape that does.

il-b• 4 months ago

Ironically, the site is down

prodigycorp• 4 months ago

I'm sorry to be this guy but this is an incredibly poor quality article. False structure (thesis/evidence), links to poor quality sources, and a non-examination of the core thesis, which is that it's burning too much money.

$15m/day inference? How was that calculated? Forbes? Did they get it right? Is that a reasonable estimate? Still valid? How was revenue calculated?

IMO most of the votes had to come from some vote ring (35 pts in 35 minutes for a crap article, no way.)

Mashimo• 4 months ago

Yeah, it's not very good.

Even the basics:

Every second of video requires rendering hundreds of individual frames

Was probably only 24 or 30 frames, not multiple hundreds per second.

orwin• 4 months ago

The website is worse than anything I've made, and I'm a backend/network/system guy.

brador• 4 months ago

Those who want to generate AI videos are price sensitive and quality sensitive.

Sora was neither.

danbruc2• 4 months ago

How many seconds of video did they generate per day for those $15,000,000, i.e. what would it actually cost me to generate, say, a three minute music video for my garage band? This should probably take into account how many attempts I would likely need to arrive at something I am satisfied with.

goalieca• 4 months ago

How many minutes would you generate to finally land on your 3 minute final copy?

danbruc2• 4 months ago

Assuming I know what I want and am somewhat competent at describing it, I would guess ten times the final length should be plenty. If you are exploring different options, you can of course produce an unlimited amount of videos. But that is not really what I was referring to, I was more thinking of how many attempts it takes the model to produce what you want given a good prompt - I have never used it and have no idea if it nails it essentially every time or whether I should expect to run the same prompt ten times in order to get one good result.

josefritzishere• 4 months ago

AI slop about AI slop? This is tiresome.

Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

Related Stories