How Ideas Rule the Market

It's been a while since I posted anything under the TLG Research umbrella, but I've been busy making Polymarket more efficient (more on that in the coming days/weeks) and building Imago.

Over the next few weeks, I'll be laying out research I've been doing on the AI Edge device build-out. We'll go into every aspect of it: what needs to happen, the timeline, and the stocks to look at. This doesn't mean I'm going fully long those stocks right now - it just means I think this is where opportunities will lie.

But today we're talking about something that's been on my mind when it comes to valuing the exact stocks we'll be looking at. It got me thinking about why certain moves are happening in markets, and how the environment has shifted in a way most people didn't predict it would a few years ago.

Valuing companies in tech right now is increasingly like sticking a finger in the wind. The fundamental thesis is clearly relevant, the industry numbers stack up - but when looking at the actual valuation it's at levels not inspiring confidence. If you still want to trade these stocks, you have to trust the thesis more than the multiple. Whether that's comfortable or not is a different question.

This isn't a new phenomenon. Markets have always had cycles where narratives run ahead of numbers. What's different now is the degree and speed at which it's happening. And to understand why, you have to understand what's actually moving these stocks.

Ideas move faster than numbers. A valuation is anchored to data like earnings, margins, growth rates - all things that change slowly. An idea has no such anchor. It can be shaped, framed, and spread at a pace that fundamentals simply can't match. And markets, right now, are increasingly pricing ideas rather than earnings. That's what opens the door to the kind of violent, rapid re-ratings we've been seeing. Names doubling and tripling on no new (yet materialised) fundamental information, just a shift in how the story is being told and who's listening. It becomes not about what the company earns and the cash it generates but about the strength of the conviction behind the idea, and how many people can be made to believe in it quickly.

The AI boom is the clearest live example of an idea that has outrun the numbers. Take the structure of how capital is actually flowing through the space. Microsoft invests in OpenAI. OpenAI, flush with that capital, signs cloud contracts. With Microsoft. The investment partially funds the revenue. Valuation gains from private funding rounds at Anthropic and OpenAI feed back into how the market perceives the hyperscalers. It's circular, and it works as long as the idea holds.

All of this is fine, if the idea actually works out. But if the valuations currently being placed on AI companies were to be believed, the productivity gains implied would be transformational. We're talking about a level of disruption that shows up in employment, meaningfully. Either AI delivers on that, and we have a structural unemployment problem society hasn't begun to prepare for, or it doesn't - and then the short-term economics collapse.

Then there's the cost layer. OpenAI and Anthropic are burning through capital at a rate the current business model simply cannot sustain. Retail inference is extraordinarily expensive to run at scale, and both companies keep returning to markets for fresh funding just to stay operational (maybe an AI edge build-out is the answer here? More on that in this series). The race to IPO is about accessing the next pool of capital before the last one runs dry. The idea is what's keeping the lights on. Whether that idea is the right one is what we will see when the S-1s get released and the lock-up periods end.

And beyond the structural dynamics, there's the way information flows have changed and what that means for markets.

Retail participation in markets is growing, and a large share of it is now being guided by AI research tools. The issue imo isn't hallucination - frontier models are reasonably good at avoiding that nowadays. The deeper problem is that LLMs conform to the framing of the question you give them, which means they tend to reflect your bias back at you. Ask a bullish question and you get a bullish answer, dressed up in the language of analysis. Everything becomes groundbreaking, every company becomes a prime beneficiary, every development becomes integral to the build-out. The idea gets amplified. Crowded trades get more crowded.

And people are willing to take more risk right now. Part of that is structural. The traditional long-run just doesn't feel like it's going to work for a growing part of society. Wages, savings rates, the cost of housing make a high-variance bet on AI stocks look rational when the expected value of the conventional path is declining. The recent abolition of the pattern day trading rule adds more fuel to this as retail can now lever up in ways that weren't previously accessible (an idea I talked about in November last year). Betting markets are evolving in the same direction, blurring the line between investing and gambling in a way that makes participation feel more sophisticated than it probably is. We'll get into that further in this series.

Put all of this together and you get something that I think is consistently underestimated as a reason for a trade: pure emotion on the back of a plausible idea. FOMO, anxiety, the feeling of being left behind. Contrary to what you'd expect in an era of algorithms and quant models, emotions are becoming more relevant again, not less. Ideas spread faster, reach more people, and land in a population that is more financially anxious and more willing to gamble.

This is the environment we're operating in. I'm not saying that valuations are useless. What I'm saying is simpler: clinging to them too rigidly is as fatal as ignoring them entirely. If you're sitting out of this market purely because something looks overvalued on a traditional multiple, you will spend a lot of time in cash. (Which isn't necessarily wrong - yields are spiking, so you can earn 4.5% sitting in T-bills right now.) But that's not what we're here for.

We're not looking for super cheaply valued companies in this series. We're looking for positions that can profit from the current environment, ideas that are still early in their spread, narratives that haven't fully been priced in yet. Don't hate the player, hate the game. If the game is being played on emotion and ideas rather than multiples, then the edge comes from understanding that dynamic better than the next person.

Keep that in mind. That's the frame for everything that follows in this series.