Revenue numbers are vastly inflated compared to pre-AI but these startups aren’t keeping the money. Profits are worse than ever before. Startups with 30 employees that reach $100m ARR in 6 months are not banking $90m or $80m or… they’re just passing that money straight through to OpenAI and Anthropic.

If startups aren’t just funnelling all their funds raised straight through to OpenAI and Anthropic, where is this combined $100bn in revenue coming from? Who is paying for it? My spend on software certainly hasn’t gone up in a post-AI world. My company is spending less on software now.

OpenAI have stopped being so reckless with their cash investments which is why they appear to have slowed down but they’re still investing millions in huge numbers of startups through token allowances. They invest $2 million in every YC startup (or did a few months ago). There’s an entire market of reselling these tokens!

https://mlq.ai/news/openai-and-anthropic-pour-up-to-800m-a-y...

Hell, I’ll go one step further and bet they book these credits being spent as revenue.

The problem with judging Alphabet and Meta on aggregate performance is that their advertising businesses print so much money that they can invest everything in a boondoggle and coast when it blows up. Zuckerberg burned $100,000,000,000 on the metaverse and it didn’t make a dent.

That said, Alphabet and Meta are borrowing against their future advertising profits to fund their AI boondoggles. If the advertising businesses keep growing then they’re somewhat insulated from their own missteps but the reliability of the advertising business depends on companies having money to spend on advertising.

The metaverse was mostly self-contained and the failure had zero consequence for the broader economy. AI on the other hand, every major fund is investing everything into AI companies. Every advertiser is using subsidized AI tools to generate hyper specific adverts. If this all goes south, who knows how advertising spend will be impacted.

Google specifically are backstopping billions in data centre build out costs. They’ve committed to spending, like, all of their cash to data centres. If the market gets nervous and funding disappears, Google are deep in the hole.

Anthropic “invested” $50bn in data centers to be built by Fluidstack who raised a billion dollars from Situational Awareness who used their Anthropic ownership to raise money to fund these investments. Google are backstopping much of the Fluidstack build out, i.e: if Fluidstack goes out of business then Google is on the hook for $10bn+ in costs. And Google’s Anthropic investment makes up like $100bn on the balance sheet. So, Anthropic fails to live up to expectations and fails to pay Fluidstack who can’t pay their suppliers which puts Google on the hook to hand over tens of billions in cash while at the same time their biggest investment is going down the pan.

The top line numbers don’t really do justice to the scale of the risk. You need to look at the long term commitments they’re making.

A startup raises $50 million from OpenAI and Anthropic to finance API calls to OpenAI and Anthropic that they are using at a loss who in turn spend that money on compute with Microsoft and Google who in turn invest in Anthropic and OpenAI who then invest the startup using the startup’s revenue to value it… the cycle repeats.

There are multi-billion dollar valued startups invested in by OpenAI and Anthropic with hundreds of millions in ARR that are spending 90% of their revenue with Anthropic and OpenAI.

Situational Awareness, the fund that recently imploded, invested tens of billions into AI companies using their holdings in Anthropic to help finance the investments…

This could all work out fine in the long term, we’re all just speculating at this point, but the circular financing is absolutely making it to revenue because capital invested into startups is used to fund growth which is achieved by subsidizing costs incurred with OpenAI and Anthropic.

There’s speculation on both sides and certainly Zitron is on the extreme end of the anti-AI side with the most cynical speculation but it is indisputable that none of the megascalers are open about their AI financials. Hence, we are all speculating endlessly. If only there were published financials then the speculation could end!

The obfuscation of financials doesn’t necessarily mean something bad is happening, it could be a competitive advantage for Google to be secretive about how cost effective their TPUs are or for Microsoft to hide how much revenue uplift they’ve experienced by adding AI to 365.

Facebook is reportedly the company that spent $500 million in a single month on tokens. There are individual non-enterprise users rotating multiple subscriptions incurring $10k+ in tokens per subscription. Facebook’s $500 million month… is equivalent to ~10k individual subscriptions which could be as little as a few thousand of the heaviest users. That’s $500 million when billed on usage, or ~$2 million on plans.

The reason resets are such a big deal (people have set up websites to track them, tweets announcing them get millions of impressions) is because there are huge numbers of users pushing their plan limits every single day. If there was huge demand from usage-based customers (the large enterprises) that OpenAI and Anthropic couldn’t meet, they wouldn’t be handing out resets like candy.

I think a realistic belief is that Anthropic and OpenAI have vastly overstated demand and are using resets as a way to keep usage artificially inflated at a substantial financial cost. I’d guess fixed price plan users make up at least 95% of usage.

Right now there are a small number of very very resource intensive use cases that are being subsidized by OpenAI and Anthropic. There are people generating millions of lines of code because it’s basically free at the point of use, despite the code producing very little value. Anthropic and OpenAI frequently “reset” customer limits to allow them to use even more resources at no additional cost.

The majority of use cases across business are not generating millions of lines of code per employee. The majority of businesses need just a little bit of automation to radically improve the way they operate. A software engineer making endless projects because it’s free to do so might use hundreds of billions of tokens per year, but an entire manufacturing business could be revolutionized with a few million tokens per year.

I think 2 things can be true:

1. There is very little penetration of AI across the economy and huge room to grow in the number of businesses deriving economic value from AI

2. The compute usage today is vastly overrepresented by usage outliers who are not paying the cost of their usage and will stop when forced to pay the cost

We could see AI usage 10x while seeing compute decrease 10x if the type of usage shifts. Most businesses just need smarter macros.

The business model of spending hundreds of millions on marketing insurance with cafes is silly but maybe it’ll work, gotta swing for the fences, but selling insurance that could implode at a moment’s notice is grossly irresponsible. And when it implodes, it’s going to hurt thousands of YC startups.

Risklytics will be much more expensive than Corgi but it’ll be actual reliable insurance. Anyone going with Corgi is out of their damn mind.

Corgi are taking the batshit insane approach of insuring clients directly through an esoteric insurance structure that dodges regulation.

Choosing Corgi over a typical broker is playing with fire.

https://reticulating.substack.com/p/ycombinators-corgi-insur...

My broader philosophical take is that we, programmers, lived through a golden age where our skills used on our terms were some of the most valuable skills. The golden age is over, our skills aren’t useless, they can still be applied to making things with modern tools, but it is no longer on our terms, no longer programming, no longer the meditative thinking process it once was.

For non-programmers, this is their golden age, the reign of programmer tyranny is over.

For businesses it makes sense to abandon programming in favor of delegating to agents that can do more in less time, but for programmers, it is a loss. Either be a programmer and code, or be a delegator and delegate, you aren’t going to make the life of a delegator suck any less by trying to trick yourself into thinking you’re programming.

As for product companies, like Ramp, buying domains as an investment is generally a terrible idea. Domains are not liquid, the (hypothetical) $5 million that Ramp paid is not a price Ramp could achieve when selling the domain tomorrow. The previous owners owned Router.com for over 20 years before Ramp came along, willing to pay a lot of money for it.

Sometimes a company will buy a domain, use it, shut down whatever it was, then end up with the domain on their balance sheet for years, even decades, before someone tracks them down and offers to buy it. In those cases, the company can make money off the domain, but for every time that happens, there’s thousands more domains on balance sheets that nobody wants to pay a good amount for.

More broadly, domain name investors make their money by buying lots of domains and holding them over the long term, making their money from the very small number that sell. Part of why good domains are so expensive is that 1 sale needs to cover the cost of hundreds of even thousands of other domains that don’t sell.

AI is a tool, a useful tool, an innovative tool, but technology has introduced innovative tool after innovative tool for years. Apache, Twitter’s Bootstrap, Pull to Refresh, React, Kubernetes, all innovative, all delivered real business value. Nobody invested hundreds of billions of capital into infrastructure for React.

AI has endless use cases. Theres no evidence that these use cases deliver the value necessary to justify any capital expenditure. A technology can change the world and have no standalone business case.

Business is boring. Money in, money out. We are still waiting for Anthropic and OpenAI to show how LLMs have moved the needle on business.

Every single thing that an LLM can be used to deliver value for a business today can be achieved by a cheap to run open weight model that companies could host in house. Where does that leave the trillion+ worth of dollars being buried in the ground?