AI is built on fake math. Nobody says it. Most won't survive.

Everyone knows the math is broken. We're all building anyway.

Amos Bar Joseph May 28, 2026

I’m Amos Bar Joseph**, co-founder of Swan, the first Autonomous Business OS. At Swan, we’re building what we call the Autonomous Business: a company that scales to $10M ARR per employee with no bloat, no assembly lines, no Cog Culture. Just humans in their zone of genius, amplified by AI agents.*** I write The Autonomous Age to share contrarian insights from that journey, on GTM, leadership, and the future of work. If you want to understand how GTM evolves beyond playbooks and assembly lines, this is where the story unfolds. Connect with me on Linkedin or X. If that's not the game you're playing, reply to unsubscribe.

We spent $113,000 on AI last month. $315k+ in the last 5 months. We're just a few employees. I'm starting to think the entire AI industry is built on economics that don't work yet.. and might not work at all.

I need to be clear about something first. I'm building Swan AI** as an autonomous business. I believe AI agents are the future of how companies scale. I've structured my entire company around humans + AI working together. I'm probably one of the most bullish people on this technology you'll meet.

Which is exactly why what I'm about to say scares the sh*t out of me.and keeps me up at night...

The entire AI industry is subsidizing token costs atm.

Meaning the prices customers pay today don't reflect the actual cost of compute.

And everyone knows it.

We're all doing the same math: charge what the market expects. absorb the difference. Hope costs come down before the money runs out.

Look at Claude Max - $100/month, unlimited usage. a single power user can burn $500-$1000 worth of tokens in that timeframe.

GitHub Copilot had to abandon flat pricing entirely because agentic coding workflows were destroying their unit economics.

The model providers are subsidizing compute. The application layer is compressing margins. And we're all building businesses on top of a foundation that doesn't have real economics underneath it yet.

It's not sustainable. Tt's a bet.

I know what you're thinking: "Uber subsidized. Netflix subsidized. you subsidize to win, then economics kick in at SCALE."

That's true. if only AI worked like those businesses. But it doesn't.

Uber got cheaper per ride as it scaled (better density, better matching).Netflix got more profitable with subscribers (fixed content costs spread across millions viewers).

But AI doesn't work that way. The millionth token costs the same as the first token.

There are no natural economies of scale in token generation. Growing your user base doesn't make the compute cheaper. Processing more requests doesn't lower your unit costs. Scale doesn't solve the problem. it just makes the problem bigger.

Which means we're all betting on something none of us control: NVIDIA figuring out how to make chips cheaper. TSMC scaling production fast enough. Data centers becoming radically more efficient. The entire compute supply chain driving costs down faster than usage grows.

This is actually a hardware problem..

and everyone building in AI right now (Swan included) is making a bet on variables completely outside our control.

The thing I'm betting my life on depends on variables i have zero influence over.

So what do I do with this? Stop building? fu*k no.

I still believe autonomous businesses are the future. I still think building a coding agent for GTM teams is exactly what the world needs.

But I'm not sleeping as well as I used to......

Is anyone else thinking about this???

[getswan.com

](https://getswan.com/?utm_source=theautonomousage&utm_medium=referral&utm_campaign=ai-is-built-on-fake-math-nobody-says-it-most-won-t-survive)

The AI industry is built on subsidized math that doesn't reflect real compute. Token pricing is fiction. When it corrects, most won't survive it.

Name the bet.

• Write down the variables your business survival depends on that you can't control. Most of them are on NVIDIA's roadmap, not yours.

• For each one: if this doesn't go your way in 24 months, does your business survive? That's your real risk profile.

• Build anyway. With full awareness of what you're betting on and why you think that bet wins.

If the correction hits next year, what makes you one of the ones still standing?

The ones who survive won't be the ones who didn't see it coming. They'll be the ones who built anyway.

-Amos

[Read more.

](https://www.linkedin.com/posts/amos-bar-joseph_we-spent-113000-on-ai-last-month-315k-share-7462541252928995330-7Gll/?utm_source=social_share_send&utm_medium=member_desktop_web&rcm=ACoAAAYm8JkB3GSdkfgU5-3_tdW6c48fsSkXvCs)