Video · 2026-07-20 · 39m · 6 moments

Is This a Bubble or a Melt-Up?

✦ AI generated

timeline · colored by role

Claim

The market is broadening without damage to the headline index, which is a win — 71% of S&P stocks are above their 200-day moving average, and the equal-weight index is up while the AI-heavy space is down.

Yurion argues that the current broadening of the bull market is healthy because it's happening without the headline S&P 500 declining — a classic sign of a sustainable rally rather than a late-cycle top.

transcript

Yurion Timmer: right now the market is broadening without really any loss to the headline index. And just to give you an example, since June 2nd, the S&P has not made a new high. It's down about half a percent. Which is really not a big deal, of course. Um, but the uh the AI space within the S&P is down 15%. The Cosby index in Korea is down 22%. uh the equal weighted index is up two and the X ai space within the S&P is up five. So uh and 71% of stocks in the S&P are above their 200 day moving average.

extends · 1provides context · 1supports · 3

Claim

We are in an era of fiscal dominance where shocks are rate scares rather than growth scares, and structurally higher inflation means the 10-year yield has more upside risk to 5% than downside.

Yurion argues the macro regime has shifted from the post-GFC deflationary growth-shock era to one where fiscal policy dominates and inflation is structurally higher — making rate scares the new normal and pushing bond yields higher.

transcript

Yurion Timmer: we're in an era of fiscal dominance and fiscal policy dominates as the as the name would imply. Um, and we're, you know, we're no longer in the era where the main concern in a 6040 type portfolio are growth scares or growth shocks, right? Think about the GFC. That was a growth shock. COVID was a growth shock. ... Now we're on the other side and when we get a shock and and shock is a big word. Um but when we get a scare, it's it's a rate scare, not a growth scare.

Mechanism◆

Semiconductor earnings have tripled but the sector is trading at 14x the next two years' earnings — whether that's a value or a trap depends on whether AI-driven demand is structural or cyclical.

Yurion highlights the central tension in the AI trade: semiconductor PEs have collapsed because earnings exploded, but semiconductors are historically cyclical — buying a low-PE cyclical is a classic value trap. Investors must decide if AI demand is structural enough to sustain those earnings.

transcript

Yurion Timmer: Semiconductor earnings have tripled over the last year. Um, and their PE is actually at 14 times the next two years of earnings, which is super low. ... Semiconductors historically are a cyclical sector. And so anyone in the value space will know that, you know, buying a cyclical sector because the PE is low is the ultimate trap because it just tells you that earnings are high. And if it's cyclical, they're not going to stay high. So we have to assume right now that semiconductors are not cyclical but more structural because the AI boom is structural.

supports · 2

Data

Fast money chases momentum and amplifies moves but is not loyal — it shifts from Bitcoin to gold to semiconductors as each catalyst fades, and the current AI correction is a healthy sign that critical questions are being asked.

Yurion traces the fast money rotation from Bitcoin (ETF catalyst) to gold (central bank buying) to semiconductors (earnings boom). He notes the current AI correction is healthy because people are asking critical questions — the hallmark of a boom, not a bubble.

transcript

Yurion Timmer: the fast money is not loyal to anything or you know they fast money just wants price to go up. uh and when and when liquidity is ample. ... And so right now you know AI is in a correction. the semis are in a correction ... And again, that doesn't mean that it's a bubble or that it's a it's it's the end of a boom. It just tells you that it's a little crowded and as long as the fundamentals are good, you know, a crowded trade doesn't have to mark the end of of a boom.

gives example · 2

Prediction

The antidote to concentration risk and an overcrowded boom is diversification into high-payout-ratio sectors like European banks and US financials — which offer 5-7% yields, low correlation to MAG7, and potential AI beneficiaries downstream.

Yurion argues the best defense against a crowded MAG7/AI trade is to diversify into sectors with high payout ratios — European banks (88% payout, 7% yield, 11% correlation to MAG7) and US financials (84% payout, 5% yield). These are potential downstream AI beneficiaries and offer income that protects against over-earning beta in a long-in-the-tooth secular bull market.

transcript

Yurion Timmer: European banks, okay, uh are are they they are they are like one of the best assets right now. They are um only 11% correlated to the MAG7. They have a payout ratio of 88% which means you get almost every euro of earnings back to you. They have a yield of 7%. um and their payout is growing faster than the Max 7. ... And for me, the transition um could well be underway that you know the MAG 7 was yesterday's leader and tomorrow's leader will be the actually the downstream companies that benefit from this.

supports · 2

Claim◆

The current AI boom cannot be called a bubble because valuations are actually falling — the tech sector PE went from 22 to 21, unlike the 1998-2000 dot-com bubble where PE doubled from 36 to 70.

Yurion draws the key distinction between the current market and the dot-com bubble: earnings are growing so fast that valuations are compressing, not expanding. Without unsustainable valuations, you can't call it a bubble — though the cyclical vs. structural nature of semiconductor earnings remains an open question.

transcript

Yurion Timmer: going back to 98 to 2000, the PE on the tech sector went from like 36 to 70. It doubled basically, right? This time it's gone from 22 to 21, right? So it's a totally different story. That last episode of the internet bubble was entirely valuation uh unsupported by anything else. And this one the earnings are so big that the valuations are actually going down instead of going up. So you can't I can't really call it a bubble under those circumstances

supports · 1

Highlight slides

The Semiconductor Valuation Paradox✦ from: Semiconductor earnings have tripled but the sector is trading at 14x the next two years' earnings — whether that's a value or a trap depends on whether AI-driven demand is structural or cyclical.Cyclical Value Trap or Structural Opportunity?✦ from: Semiconductor earnings have tripled but the sector is trading at 14x the next two years' earnings — whether that's a value or a trap depends on whether AI-driven demand is structural or cyclical.AI Boom: Valuations Compressing, Not Expanding✦ from: The current AI boom cannot be called a bubble because valuations are actually falling — the tech sector PE went from 22 to 21, unlike the 1998-2000 dot-com bubble where PE doubled from 36 to 70.Why This Time Is Different✦ from: The current AI boom cannot be called a bubble because valuations are actually falling — the tech sector PE went from 22 to 21, unlike the 1998-2000 dot-com bubble where PE doubled from 36 to 70.

Related episodes

Video · a16z Podcast

The New Rule for Picking AI Winners | The a16z Show

→

Video · a16z Podcast

The $1 Trillion Firm That Refuses The Private Equity Label | a16z

→

Audio · No Priors

Pax Silica: Inside the Trump Administration’s Tech Strategy with US Under Secretary of State for Economic Affairs Jacob Helberg

→