🧮 The Winner-Takes-Most Equilibrium

👋 Hey, it’s Remy.

February was a record month for VC. $189B. 83% went to three companies.

Most people read that as an AI story. I think it's a game theory story - and it has implications for anyone building in a market that's starting to tip.

Reply with which position you're in (1, 2, or 3 — you'll see what I mean) and I'll read it.

You have a TAM slide. You have a competitor matrix. You have a differentiation story.

What you probably don't have is an answer to this question: does my market tip?

Not "is my market large?" Not "can I win customers?" But: is this a market where one player eventually captures most of the value, and am I on track to be that player?

This question changes everything: how you hire, how you raise, when you push, and when you sell. Most founders never ask it explicitly. They operate with an implicit assumption that markets are competitive, that quality wins, that the best product gets rewarded proportionally.

Sometimes that's true. Often, the maths say otherwise.

Last month handed us a useful data point to make this concrete.

In February 2026, global venture investment hit $189 billion. The largest startup funding month on record. 83% of that capital went to three companies: OpenAI raised $110B, Anthropic raised $30B, and Waymo raised $16B.

The instinct is to call this a distortion. Megadeals skewing the numbers. AI exceptionalism.

It is not an anomaly. It is a Nash equilibrium resolving in real time.

In game theory, a Nash equilibrium is a state where no player can improve their outcome by unilaterally changing their strategy. What we watched in February was rational capital allocation converging on the dominant position in markets that had already tipped. Investors weren't being irrational. They were playing the equilibrium correctly.

To understand why, and how to use this as a founder, you need the underlying maths.

Start with Metcalfe's Law.

In a network of n users, the number of possible connections V is n(n−1)/2, which scales asymptotically as n². The value of the network, in Metcalfe's formulation, scales with the square of its users:

This is not just a metaphor. It is a structural fact about how certain products work. A messaging app with 10 users has ~45 possible connections. One with 100 users has ~4,950. The second network is not 10x more valuable. It is roughly 110x more valuable, per the same number of connections.

Now consider two competing networks, one with n₁ users and one with n₂ users, where n₁ > n₂. The value ratio is:

A network with 30% more users is not 30% more valuable. It is approximately 70% more valuable. A network with twice the users is four times more valuable.

This is the engine of winner-takes-most dynamics. Small linear advantages in user count compound into enormous non-linear advantages in value. The gap doesn't close. It accelerates.

Important note: we’ve used network size above to illustrate a compounding but it’s only one out of dozens of other compounding business structures.

We call this kind of feature an exponentiality: a structural property of a product that causes its value to scale super-linearly with growth. Metcalfe-style network effects are the canonical exponentiality, but they're not the only one. Data flywheels, switching cost accumulation, and ecosystem lock-in are all exponentialities. Each one changes the shape of the value function from linear to curved, and that curve is what drives markets to tip.

(A note on precision: Metcalfe's n² is debated. Some researchers argue value grows as n log(n) for mature networks, which is more conservative. The qualitative conclusion, that value grows super-linearly, creating compounding structural advantage, holds either way. The exact exponent changes the speed of tipping, not the direction.)

The formal mathematical concept behind market tipping is a bifurcation: a qualitative change in the behavior of a dynamical system caused by a smooth change in a parameter crossing a critical threshold.

Think of it like this. A market is a dynamical system. It has a state (who holds what share) and dynamics (how share shifts over time). Below the tipping threshold, the system is in a regime where multiple players coexist. The dynamics are competitive. No single player has an insurmountable advantage.

At the bifurcation point, the system's behavior flips. What was a stable multi-player equilibrium becomes unstable. The market reorganizes around a dominant position. The prior equilibrium collapses.

Mathematically, a bifurcation occurs when a parameter of the system crosses a critical value λ₀, and the stability of the equilibrium changes. What was attracting becomes repelling, and a new attractor (the concentrated equilibrium) takes over.

For startup markets, the parameter λ is roughly: the ratio of the leader's exponentiality advantage to the cost of switching for customers. When that ratio crosses the threshold, the market tips. The transition is not always visible from the outside. The system can appear competitive right up to the bifurcation, then concentrate rapidly.

This is why founders are often surprised. They were watching growth rates, not the underlying parameter.

Before you apply this: not every market has a tipping point.

The bifurcation only happens when exponentialities are present — when there exists ways where value compounds with scale through network effects, data accumulation, switching costs, or ecosystem lock-in. Remove those, and you have a normal competitive market where #2 can close the gap by working harder or spending more.

Commodity markets don't tip. Fragmented service markets don't tip. Local markets often don't tip, because geography resets the network to zero. Restaurants don't tip. Law firms don't tip. Plumbers don't tip.

The question isn't "is my market competitive?" It's: does my product have a structural feature that makes it worth more to user N+1 because user N already chose it? If yes, you're in a tipping market. If no, classical strategy applies — and this framework is the wrong lens.

Everything that follows assumes exponentialities are present.

After a bifurcation, the market settles into a power law distribution of outcomes.

A power law says the k-th ranked player has value proportional to k^(−α), where α is the scaling exponent. For most winner-takes-most tech markets, α is large enough that the leader captures an order of magnitude more value than the second player, and the tail falls off sharply.

Venture returns follow this distribution. So do market caps within categories. So does the February funding data.

This has a concrete implication: being number two in a power-law market is not half as good as being number one. It is a categorically different business. Different capital access, different talent gravity, different ability to set the pace of the category.

The expected value of your position is not determined linearly by your market share. It is determined by which side of the bifurcation you're on, and your rank in the power law tail.

The maths define three possible positions in a tipping market. Each has a structurally different optimal play.

Position 1: You are the leader approaching the bifurcation.

Your exponentialities are compounding. Your n² advantage is widening. Capital is starting to concentrate toward you.

The GTO play: accelerate aggressively. Do not optimize for capital efficiency. The objective function here is not "maximize runway." It is "cross the bifurcation threshold before the equilibrium can shift against you."

In a market governed by super-linear value scaling, the cost of losing the tip is higher than the cost of over-investing. Every dollar of capital, every key hire, every partnership either widens the gap or delays a competitor from closing it.

This is not a vibes-based "move fast" argument. It is the correct rational play given the payoff structure. OpenAI raising $110B is not irrational exuberance. It is a leader executing GTO strategy in a tipping market, deploying capital to accelerate past λ₀ before anyone else can.

Position 2: You are second or third in a market that has already bifurcated.

This is the most dangerous position, and the most common one founders find themselves in without realizing it. The equilibrium has resolved against you. The leader's exponentialities are already compounding. You are in the unfavorable tail of the power law.

Growing faster will not close the structural gap. The leader is also growing, and their value function is non-linear. You are not playing a race. You are playing from a structurally worse position in an already-settled game.

The GTO play is one of two things.

Find a flanking position. Identify a sub-market where the leader's exponentialities are weak or absent, where n resets to near zero and the bifurcation hasn't happened yet. The goal is to find a game that hasn't tipped, not to fight the one that has.

Exit from a position of strength. A strategic exit when you are a credible second player, before the power law compresses your position further, is often the highest-EV outcome. Most founders treat this as failure. The maths say otherwise. The option value of your position declines as the tail falls.

Position 3: You are building in a market that hasn't tipped yet.

This is the highest-leverage position, and the one where explicit thinking about bifurcation dynamics pays off most.

The strategic questions are: what is the tipping parameter λ in this market? How far is it from the bifurcation point? And what moves shift it in your favor?

The answer almost always traces back to the same principle: identify your exponentialities and front-load them.

If the compounding variable is data (as in AI), accumulate it faster and more defensibly than anyone else. If it's network density (as in marketplaces), pick the geographic or vertical subset where you can cross the bifurcation threshold with the resources you have, then expand. If it's switching costs (as in infrastructure), embed yourself into workflows early, before the cost of the integration matters to the customer.

In each case, the move is to engineer an early n² advantage, even a small one, that can compound before the market tips around someone else. Early movers in tipping markets are not just ahead on a linear timeline. They are acquiring the exponentiality that the post-bifurcation equilibrium will be organized around.

Most startup advice optimizes within the wrong frame. Growth rate, unit economics, product-market fit: these are useful in competitive markets. In tipping markets, they are insufficient. You can have better unit economics than the leader and still be in the unfavorable tail of the power law.

The prior question, the one the maths force you to ask, is:

Does my market allow the leveraging of some exponentialities? If so, has it already bifurcated? And which side of that bifurcation am I on?

If your honest answer is that your market has network effects, compounding data advantages, or rising switching costs, the value function is non-linear. The equilibrium will eventually concentrate.

In that case, the $189B February is not a story about AI. It is a preview of what rational capital allocation looks like when any market resolves.

The founders who read it as an anomaly will keep optimizing for the wrong variable.

The ones who understand the structure will ask the harder question: in the game I am playing, what are my exponentialities, how far is λ from λ₀, and what is the GTO move from where I am standing?

That is one of the shift from art to science. Not just building faster. Seeing the shape of the equilibrium you are building toward.

One question before you go:

Which of the three positions are you in right now - and do you think your market has actually tipped yet?

Hit reply. I read everything and respond to everyone.

If you want to help shape future essays like this, we're opening a small number of TechGames Fellowship spots for people who think about startups through the lens of games, maths, and systems.

Know someone building in a tipping market? Forward this to them.