Valuations and Bubbles

Thoughts about momentum and ceilings

So this was written over a week ago in my finance newsletter, but given recent news I thought I’d post it here as well due to relevancy. It is no surprise that within a week, a lot of the below numbers are already out of date.

If you’re still early in your investing journey, one of the hardest things to do is figuring out whether valuations make sense. The tricky part is that if you are in a high momentum industry, valuations can shift high very quickly, historical data can be inaccurate, and in general you probably may not understand what it means when something in crypto or AI is valued at x billion dollars for example.

One problem though is how do you figure this out? Let’s walk through what’s happening in AI, where valuations are moving up because ceilings are getting moved up from the market leader, and in crypto, where projects that falter might drag other projects down.

For AI, this is a very interesting case study. Say you meet a founder who is pre-product and says that his company is valued at $100M. Does this make any sense at all? Well the way you would define this is to see if there are similar category companies and drawing a line of progression of where the valuation should be at. It is entirely possible for example that several other similar companies are valued at $200M, but if they have inferior teams and are at similar pre-product stages, then maybe $100M is very reasonable since a later round could drag upward for you.

Let’s take a more definitive use case. The AI private market leader right now in the private market is OpenAI. Each year, OpenAI has constantly pushed the ceiling for the highest valuation in the market. In 2023 it was valued around $30B, in 2024 it was valued at $157B, and now in 2025 it’s rumored to be raising at $340B. Since this time period its revenue has gone from $200M to $3.7B last year, and expects to hit $11.7B this year. We’re talking about a 92x revenue multiple with a forward looking 30x revenue multiple.

Meanwhile, the other frontier model companies are not staying still. Anthropic just raised a round at a $61.5B valuation. They had $200M last year and are going to try to hit $2B this year. If you do the multiple game, that means they had a 300x multiple from last year but a 30x forward multiple. Again, the numbers seem to match here based on revenue…but then you get stuff like Elon Musk’s xAI, which is rumored to be raising at $75B on pretty much no revenue. However, if you take a step back, you can see that OpenAI is the one continuously defining what the ceiling is for frontier model companies. Now, if a model company wants to start from scratch, has an amazing team, they can probably demand pretty crazy valuations.

Don’t believe me? Former OpenAI co-founder and chief scientist Ilya Sutskevar’s company is raising at a $30B also preproduct. So if you draw a line, OpenAI: $340B for $3.7B farthest along, $60B for Anthropic at $200M which is catching up, $75B for xAI which I suppose is because of Elon, and $30B for SSI because of Ilya. One now forgets that the addressable market now has so many players which could eat each other or cause pricing pressures, yet they do not know what the ceiling is. Now, if OpenAI were to come tomorrow and raise at a $500B valuation, no doubt that it would drag up all the valuations of the other names listed above. Not only that, even unrelated AI companies would be dragged up as well.

If you look elsewhere in AI, the same thing is happening to each segment, valuations of Coding Assistant startups, Support Agent startups, Video Gen startups are all being lifted up. There are revenue signals to justify this, but even pre-revenue or pre-product companies are still commanding huge raises. This is where we could be potentially entering bubble territory. We’ll see some unicorns get funding without anything to show for it, and winners emerge from the noise. The losers however will be the startups that shouldn’t have been highly valued in the first place.

Now let’s turn back to crypto, where the opposite effect is happening. When you have meme coins rise like crazy, one thing I’ve explained in the past is that people are diversifying risk to get higher returns. They make a lot of money from Bitcoin rising so they take those gains into Solana, then they take those Solana gains and throw it into ecosystem coins, then they take those gains and throw them into meme coins. Basically the higher the risk the higher the reward, and the cycle repeats over and over.

But another way of looking at this is with actual examples, think about when AI Agent coins were the narrative only a few months ago. We had VIRTUALS and AI16Z skyrocket and lift up other AI Agent coins as well. This meant that seeing coins valued in the multiple billions meant other smaller agent coins suddenly had potential. That meant as an investor it wouldn’t be “unreasonable” to see some of the coins at $100M, or to buy coins in the $10M range because you could see them get “lifted” as momentum was pushing this wave. However, as I mentioned earlier, this is also where bubbles are formed. There cannot be too many winners after all, and bubbles typically get popped.

When these categories fall, for example when Virtuals fell below $1B from nearly $4B, you need to similarly start discounting all the other AI Agent coins and cut their valuations, sometimes way more than the ratio of where the leader fell. So even if Virtuals fell 75%, you could expect other 0 value coins to fall 95%. That is the nature of this trend. In order to value a company or project, you can look at the momentum to help determine what’s reasonable, however, the direction is paramount to your valuation.

In summary: Remember, when momentum gets too fast and valuations on the market leader lift followers to sky high valuations, this is where bubbles are formed. If the market leader drops in momentum, the lesser players will drop way faster. In AI this is still rising, in crypto we are seeing the reverse.