← Back to issue
The Twenty Minute VC

Mercor CEO on Why Application Layer Companies Have No Defensibility & Token Spend Exceeds Salaries

1h 14m · Transcribed via assemblyai · Watch on YouTube

Brendan Foody (co-founder/CEO, Mercor — **valued >$10B, >$1B revenue over two years**) with Harry Stebbings. The throughline is a self-interested but sharp short thesis on the AI stack: **'the next 12 months will be dramatically better for infrastructure companies upstream of Anthropic and OpenAI than for application-layer companies downstream of them'** — because *'building defensibility in the software layer on top of the models is going to be incredibly difficult.'* Foody's mechanism is brutal and specific: **'2025 was the year of how do you get a model to make a PR in a code base, and 2026 is the year of how do you get the model to clone Slack end to end'** — a savvy customer paying *'a million dollars a year on the SaaS product'* realises they *'could just tell Claude to copy it.'* Mercor sells the opposite of the doomed layer: a vertically-integrated data/RL-environments business **paying out 'over $3 million a day in the fastest job category ever created in history'** off a **5-million-person talent network**, at **30-40% gross margin**, **profitable, with >$500M cash**, having **added $300M net new ARR in the last 60 days** after a security incident. The most load-bearing number in the cohort lands here: **'right now we're spending more on tokens for our internal agents than we are on employee headcount'** — Foody confirms token spend > salaries, *'and I would bet that in five years the average enterprise spends more on compute than headcount.'* The cross-current that complicates the bull case: he expects the **model API layer to commoditise** — *'the switching costs are zero… there's a new frontier model every two months'* — with the **majority of inference in five years on open-source/distilled models, not frontier ones**, even as he calls OpenAI and Anthropic *'incredible investments'* and could *'definitely see one of them being a $10 trillion company.'* On Nvidia he plants the same multi-chip caveat as the rest of the cohort: even at *'30 or 40% market share… that is the world's most valuable company.'* The labour-market tell for Jack: AI researchers now cost *'in the tens of millions of stock per year,'* and one hire had a *'$20 million in cash per year'* offer from Meta's superintelligence group.

Key points

Notable quotes

The next 12 months will be dramatically better for infrastructure companies upstream of Anthropic and OpenAI than for application layer companies downstream of them.

Harry Stebbings · 32:33

And so I feel like building defensibility in the software layer on top of the models is going to be incredibly difficult. Whereas on the other side of things, in the infrastructure side, it feels like there are meaningful moats that are getting built.

Brendan Foody · 32:50

Now we're building out an eval set that measures how effectively agents can build end to end SaaS applications. Where 2025 was the year of how do you get a model to make a PR in a code base and 2026 is the year of how do you get the model to clone slack end to end and those capabilities are going to exist in the models in the next 12 months.

Brendan Foody · 35:52

Right now we're spending more on tokens for our internal agents than we are on employee headcount.

Brendan Foody · 0:42

I would bet that in five years the average enterprise spends more on compute than headcount. And the reason for that is that the models are just becoming so capable that it seems like there is just enormous ROI to being able to have models do something for 100k a year that is going to continue compounding at an exponential rate in a way that human intelligence is not going to.

Brendan Foody · 46:50

Because the switching costs are zero. When the switching costs are zero, that means that and there's a new frontier model every two months that means that we very quickly are going to swap them out. And ultimately the decisions that we make boil down to the score on the eval corresponding to that workflow. And so it's very easy to compare model to model one for one in a perfectly hot swappable way, which is almost the definition of a commodity.

Brendan Foody · 45:50

Ultimately I think OpenAI and Anthropic are incredible investments and it seems like there's starting to be consensus around that in a way that there wasn't just a couple of years ago. But at the same time I think that majority of inference in five years is going to be using a open source or custom fine tuned or distilled model, not using a frontier model.

Brendan Foody · 50:03

I could definitely see one of them being a $10 trillion company, maybe even significantly higher. It feels like the opportunity associated with being the frontier model is so large that it will just eat up so much of the other demand within the economy.

Brendan Foody · 51:10

Like a great example is what we do in that now we're paying out over $3 million a day in the fastest job category ever created in history. And I expect that's going to continue growing exponentially from here.

Brendan Foody · 14:05

Oftentimes it would be in the tens of millions of stock per year for the really good people.

Brendan Foody · 1:05:33

Themes

Mentioned

People

Companies

Ideas