Can Infinite Slop make money?

How big is Infinite Slop right now?

Infinite Slop has already shown that people will watch it: more than 37,000 people visited during its first full day, and the live stream has since passed 2,000 concurrent viewers.

Those numbers are unusually strong for something Pieter Levels built almost immediately after seeing Rehan Sheikh's original H3 Max livestream experiment. The launch post also moved past one million views on X, giving Infinite Slop a huge burst of distribution from the start.

The 2,000 concurrent-viewer number is more interesting than the headline traffic. A viral post can send tens of thousands of curious people to a website for 20 seconds. Getting thousands of people to occupy the same live experience at once is harder.

We can also see that Levels is reacting to actual usage rather than leaving Infinite Slop as a weekend demo. After seeing that most traffic came from phones, he switched generation from landscape to vertical video. He added a queue, upvotes, a “playing next” indicator and navigation through previous generations. The site currently exposes TOP, NEWEST and OLDEST feeds plus categories such as banana, cat, space, news and game show.

So there is clearly an audience here. What we cannot tell yet is how much of that audience survives the launch.

Are people actually coming back to Infinite Slop?

We still don't know whether Infinite Slop can keep people coming back, and that is the biggest missing number in the business case today.

Infinite Slop is only a few days old. Levels has published traffic and concurrent-viewer milestones, but we have no public cohort retention, daily returning-user rate, average watch time or sessions per user.

That distinction is especially important for this product because the first visit almost sells itself. “An infinite AI TV channel where strangers decide what happens next” is weird enough to click even if someone never opens it again.

There are early signs that sessions can become more than passive curiosity. Viewers have created recurring banana, eye and character motifs because H3 Max carries visual elements from one clip into the next. People then see the emerging joke, write related prompts and reinforce it again. Levels has described these loops happening spontaneously.

The upvote queue adds another reason to stay. Someone who submits a prompt now has to watch the queue, collect votes and see whether the scene gets generated. That gives Infinite Slop a basic game loop that the original automatic livestream did not have.

Still, we would want at least a few weeks of repeat usage before calling Infinite Slop addictive in the business sense. Right now we know people are showing up. We do not yet know whether a habit is forming.

How much does Infinite Slop cost to run?

Infinite Slop costs about $207,360 per 30 days at fal's normal public 768p H3 Max price if it generates fresh video continuously.

Fal currently prices H3 Max by generated video second. The promotional 768p rate is $0.04 per second, with the standard price listed at $0.08. Infinite Slop needs 86,400 seconds of footage to fill one day without repeats.

At the standard rate, that means $6,912 of API-priced generation every 24 hours. Levels has independently said that Infinite Slop would need to make around $210,000 per month to be profitable, which lines up almost exactly with this calculation.

Levels is not currently paying that retail bill himself. Fal sponsors the stream, and fal's internal GPU cost will obviously be lower than the public price it charges API customers.

Still, $207,000 is the right benchmark if we want to understand whether somebody could run Infinite Slop as a standalone business without a special infrastructure deal.

H3 Max setup	Cost per second	24 hours of fresh video	30-day equivalent
480p promotional price	$0.025	$2,160	$64,800
768p promotional price	$0.040	$3,456	$103,680
480p standard price	$0.050	$4,320	$129,600
768p standard price	$0.080	$6,912	$207,360

Does Infinite Slop get cheaper when more people watch?

Infinite Slop gets dramatically cheaper per viewer as the audience grows because the same expensive AI clip can be shown to everyone at once.

This is the unusual part of the business.

If 10,000 people individually generate AI videos, the company has to pay for 10,000 separate generations. Infinite Slop can generate one clip and broadcast that same clip to 10,000 people.

At $207,360 per month, an average audience of 2,000 concurrent viewers produces 1.44 million viewer-hours over 30 days. The AI generation therefore costs about $0.144 per viewer-hour.

At 20,000 average concurrent viewers, the same stream produces 14.4 million viewer-hours and generation falls to $0.0144 per viewer-hour. At 200,000 viewers, it reaches roughly one tenth of a cent per viewer-hour.

Normal hosting and video delivery still grow with traffic, but those costs sit in a completely different range from H3 Max inference. Levels has previously shown that his conventional web infrastructure can serve enormous request volumes from relatively cheap VPS and CDN infrastructure. His expensive products are expensive mainly when GPUs enter the picture.

A shared stream also explains why personalized Infinite Slop channels would be dangerous economically. Give every viewer a private AI channel and the generation bill starts growing with users again. The communal stream gives Infinite Slop much better economics.

Can ads alone make Infinite Slop profitable?

Ads would struggle badly if Infinite Slop averaged around 2,000 concurrent viewers, while an audience closer to 20,000 would make the same advertising model much easier to support.

We can test this using Twitch's current revenue structure as a benchmark. Twitch gives qualifying streamers 55% of net advertising revenue when they run at least three minutes of ads per hour.

Suppose Infinite Slop showed six 30-second ads per viewer-hour. At 2,000 average concurrent viewers, the stream would generate about 8.64 million monthly ad impressions. Paying a $207,360 generation bill after a 55% creator share would require advertisers to generate roughly $43.64 per thousand impressions.

That is a very hard number.

At 10,000 concurrent viewers, the required gross CPM falls to $8.73. At 20,000, it drops to $4.36. At 50,000, only $1.75 would be required under the same assumptions.

Actual advertising revenue would move around with geography, seasonality, fill rate, ad blockers and the type of advertiser willing to appear around unpredictable AI content. The important finding is the slope. Increasing average concurrency tenfold reduces the advertising yield Infinite Slop needs per impression tenfold.

There is also a platform problem. YouTube's current monetization rules explicitly flag mass-produced, generic or repetitive AI content as potentially ineligible under its “inauthentic content” policy. Infinite Slop's interactive element makes it more original than a templated AI spam channel, but building the entire model around YouTube ad approval would still be risky.

Owning InfiniteSlop.ai gives Levels much more control over advertising, payments and the queue.

Average concurrent viewers	Monthly viewer-hours	Revenue needed per viewer-hour	Gross CPM required under this ad scenario
2,000	1.44M	$0.1440	$43.64
5,000	3.60M	$0.0576	$17.45
10,000	7.20M	$0.0288	$8.73
20,000	14.40M	$0.0144	$4.36
50,000	36.00M	$0.0058	$1.75

Could paid prompts make more money than ads on Infinite Slop?

Paid prompt boosts look like the strongest way for Infinite Slop to make money because viewers would be paying for something genuinely scarce: control over what thousands of other people see next.

The queue only has room for four new 15-second videos every minute. That gives Infinite Slop 5,760 generation slots per day, or roughly 172,800 slots over 30 days.

Each 15-second generation costs $1.20 at the standard 768p public rate.

That creates a surprisingly neat business model. If every generated scene somehow brought in $2, Infinite Slop would produce $345,600 per month from one continuous channel. Even after paying the $207,360 retail-equivalent generation bill, plenty of revenue would remain for payment fees, infrastructure and profit.

Infinite Slop does not need every scene to be paid either. If 25% of the available scenes were monetized, those scenes would need to bring in an average of $4.80 each to cover the entire public-price generation bill. At 10%, the number becomes $12.

People already pay similar and much larger amounts on livestreams simply to make messages stand out. YouTube Super Chat is a useful comparison: viewers pay to push their message into a more visible position, and YouTube says creators receive 70% of confirmed Supers revenue after applicable deductions.

Infinite Slop can offer something stronger than a highlighted chat message. A winning prompt can take over everybody's screen for 15 seconds.

Share of Infinite Slop scenes monetized	Paid scenes per month	Average revenue needed per paid scene
100%	172,800	$1.20
50%	86,400	$2.40
25%	43,200	$4.80
10%	17,280	$12.00
5%	8,640	$24.00

Would paid prompts ruin Infinite Slop?

A pure pay-to-win prompt queue would probably make Infinite Slop worse, even if it produced revenue quickly.

Levels has already pushed back on this idea. When users suggested letting people bid money for the four available generation slots, he said that sounded less fun and that he preferred the chaotic internet-community feel of the current system.

He has a point.

If the richest viewer can always jump to the front, free participants gradually lose their reason to submit anything. The queue becomes predictable, and advertisers or whales can dominate what everyone watches.

A softer system fits Infinite Slop better. Paid boosts could increase a prompt's visibility or voting weight without guaranteeing that it wins. Some generation slots could be reserved for paid prompts while most remain community-selected. A user might pay to start a themed round, revive an old prompt or give an idea another chance after it loses.

The free queue has to stay valuable because the crowd is creating the jokes and weird storylines that make people watch. Money can sit on top of that behavior without controlling every scene.

Could Infinite Slop make serious money from memberships?

Infinite Slop memberships could become useful recurring revenue, but a basic $5 subscription would need a surprisingly large paying audience to move the economics much.

Twenty thousand people paying $5 would produce $100,000 per month before payment costs. That still covers less than half of the standard 768p generation benchmark.

For comparison, YouTube currently gives creators 70% of net channel-membership revenue after applicable deductions. Even with direct payments and a better revenue share, Infinite Slop would still need a large conversion from viewers to paying members.

The membership becomes more interesting when it changes the experience. Paying users could receive extra voting weight, saved prompts, usernames or badges, access to special channels, better archives or a limited number of monthly prompt boosts.

That makes the subscription easier to understand. Someone would be paying for more influence and a stronger identity inside Infinite Slop rather than donating $5 to help pay an invisible GPU bill.

Membership revenue could then sit alongside paid prompts instead of carrying the whole business.

Is fal's Infinite Slop sponsorship already a real business model?

Fal's Infinite Slop sponsorship is already commercially meaningful because fal gets a live product demonstration that would be difficult to reproduce with a normal advertisement.

H3 Max's headline advantage is speed. Fal says its post-trained version can generate video faster than playback, which is exactly what makes Infinite Slop possible.

Showing that claim inside a benchmark chart is one thing. Letting tens of thousands of people use a 24/7 product powered by it is far more memorable.

The relationship also goes deeper than this one experiment. Levels has said that every AI model used by Photo AI runs through fal, with Photo AI generating more than one million images per month when he described the relationship. He later invested in fal himself.

So fal gets exposure from someone who already uses its infrastructure heavily, publicly talks about it and builds products that demonstrate its strongest technical advantage.

That makes the sponsorship unusually logical. It also means we should be careful when using fal as proof that Coca-Cola, Nike or another mainstream company would happily pay a similar bill.

For AI infrastructure companies, though, Infinite Slop can function as marketing. If another model provider wants to prove that its video model is faster, cheaper or visually better than H3 Max, sponsoring an Infinite Slop channel could be far more compelling than buying banners on AI websites.

Could mainstream brands actually sponsor Infinite Slop?

Mainstream sponsorship can work on Infinite Slop, but the best advertisers will probably be brands comfortable with unpredictable internet culture.

Anonymous users decide what gets generated. The model interprets those prompts. Visual themes can carry from one clip into the next. Even with safety filtering, that environment gives a brand less control than a normal sponsored YouTube video.

Levels has already said copyrighted prompts are restricted. Fal's H3 Max API also includes safety checks. Those protections help, but they cannot make a live generative stream completely predictable.

A normal logo beside the video probably leaves a lot of value on the table anyway.

The more interesting format would let sponsors become part of the game. A game publisher could sponsor a themed hour. An AI company could sponsor a channel generated by its model. A movie studio could provide an approved visual universe. Two model providers could compete side by side and let users vote on the better generations.

Internet-native brands, AI companies, GPU providers, developer tools and gaming companies fit that format much more naturally than conservative consumer brands.

Can Infinite Slop cut its $210,000 monthly AI bill?

Infinite Slop can reduce the effective AI-generation bill dramatically without waiting for H3 Max prices to fall.

The simplest move is resolution. Fal's standard 480p H3 Max price is $0.05 per second instead of $0.08 at 768p. Running entirely at 480p would cut the public-price equivalent from about $207,000 to about $130,000 per month.

The bigger opportunity comes from replaying content.

Infinite Slop currently stores previous generations and lets users browse TOP, NEWEST, OLDEST and different categories. That means a clip does not have to disappear forever after its first 15-second broadcast.

If half of the stream eventually came from previous generations, fresh 768p inference would fall to roughly $104,000 per month. At 25% fresh footage, it drops to roughly $52,000.

The shift to vertical video also helps here. Levels made that change after seeing that most usage came from mobile. A library of vertical clips can work as a scrollable discovery feed, giving every expensive generation more opportunities to earn views later.

Quiet hours offer another obvious lever. Spending $288 of public-price inference every hour makes little sense if almost nobody is watching. Infinite Slop could replay popular clips when concurrency drops and switch back to heavy live generation as the audience returns.

That still feels infinite to the viewer while making the business much less dependent on generating 86,400 completely new seconds every day.

Infinite Slop operating setup	Approx. monthly generation cost	Reduction vs. continuous 768p
100% fresh at 768p	$207,360	0%
100% fresh at 480p	$129,600	38%
50% fresh at 768p	$103,680	50%
25% fresh at 768p	$51,840	75%
25% fresh at 480p	$32,400	84%

Is Infinite Slop too easy to copy?

Infinite Slop is easy enough to copy technically that its long-term advantage will have to come from the audience and culture around it.

Levels openly credits Marc-Antoine Fon and Rehan Sheikh for the idea behind the experiment. Rehan's H3 Max livestream came first, and Infinite Slop turned that concept into an interactive website.

The model is available through fal's public API. Developers can call H3 Max themselves without needing some secret arrangement with Levels.

That means another competent developer could reproduce the basic loop: accept prompts, generate clips, put them in a queue and stream them continuously.

Copying the code still does not give somebody 2,000 people in the same room.

If Infinite Slop keeps growing, the useful assets become the shared audience, recurring jokes, user identities, voting behavior, prompt history, best-of archive and recognition of the Infinite Slop name itself. Levels also starts with a distribution advantage because nearly one million people follow him on X.

The generation technology will probably become more commoditized from here. Infinite Slop needs its community to move in the opposite direction.

Will cheaper AI video make Infinite Slop much easier to monetize?

Cheaper AI video could completely change Infinite Slop's economics because generation currently accounts for almost the entire unusual cost of the product.

H3 Max already shows how fast the technical baseline can move. Levels says a comparable 15-second generation previously took roughly two to five minutes. Fal's post-trained H3 Max version can now produce video faster than viewers consume it.

That speed jump created the product category almost overnight.

Price reductions would have a similar effect on the business side. If comparable 768p generation became 80% cheaper, the $207,360 monthly public-price bill would fall to about $41,500.

At that level, Infinite Slop would no longer need a huge advertising business or tens of thousands of paying users. A few thousand engaged viewers spending money on prompts and memberships, combined with sponsorships and ads, could support the stream much more easily.

There is a catch. Every price cut that helps Infinite Slop also makes competing streams cheaper to build. Better models will probably improve both margins and competition at the same time.

That makes the current audience land grab more important than waiting for perfect economics.

Can Infinite Slop actually make money?

Yes, Infinite Slop can make money, and paid participation looks much more convincing today than trying to fund the whole thing with normal advertising.

The scary number is roughly $207,000 per month for continuous 768p generation at fal's standard public price. That sounds ridiculous for a small website.

The structure underneath that number is much more favorable.

Generation cost barely changes as the audience grows because everybody watches the same clips. There are roughly 172,800 scarce 15-second slots available every month. Monetizing only 25% of them at an average of about $4.80 would theoretically cover the entire retail-equivalent generation bill. Sponsorship can remove another large piece of the cost, as fal is already demonstrating. Replaying previous generations can cut inference further. Falling AI-video prices would improve every one of those calculations.

Ads become interesting once concurrency gets much larger, but they look like secondary revenue for now. Memberships can add recurring revenue if they give users more influence. Brand sponsorship can work particularly well with AI, gaming and internet-native companies.

Retention remains the one question we cannot solve yet. The launch proved that Infinite Slop can get attention. We need more time to see whether people keep opening it once the novelty wears off.

If the stream falls back to a few hundred passive viewers, the current economics are ugly. If it holds several thousand engaged viewers and a small fraction starts paying to influence what airs, we think Infinite Slop has a credible path to profit.

The surprising part is where that profit would come from. The valuable inventory inside Infinite Slop is the next 15 seconds of everyone's screen. There are only four chances to control it every minute, and thousands of people can compete for them. That scarcity gives the product something ordinary AI video generators do not have.