The "AI Bubble" narrative is lazy and wrong when you actually look at the numbers.
Memory names like Samsung and $ESMT are trading at 1-3x forward P/E — with HBM players now locked into multi-year price floor agreements with hyperscalers. That's not a bubble. That's contracted demand at known prices.
Optical companies like $POET are trading at negative enterprise value — literally $97M below net cash. The market is valuing the operating business worse than zero. How is that a bubble?
Meanwhile, AI infrastructure players with real revenue visibility are signing massive long-term agreements: $COHR just locked in $20B, $LITE signed $5B. These aren't speculative bets — they're contracted cash flows over 3-5 years.
Yes, there are pockets of froth:
• Private AI startups like Instinct look overvalued
• Overleveraged neoclouds that misjudged their financing
• AI lab valuations that deserve skepticism
But lumping profitable, cash-generative infrastructure companies with single-digit P/Es and contracted demand into the same "bubble" bucket is intellectually dishonest.
What we're likely seeing isn't demand destruction — it's the Great AI Deleveraging of 2026. Valuations compressed from macro fear, rate uncertainty, and regulatory noise. Not because the buildout stopped.
As earnings roll in and free cash flow piles up, the market will eventually separate durable infrastructure plays from speculative froth. The bottleneck names especially look like asymmetric setups here.
I'm long the deleveraging, not the narrative.
Memory names like Samsung and $ESMT are trading at 1-3x forward P/E — with HBM players now locked into multi-year price floor agreements with hyperscalers. That's not a bubble. That's contracted demand at known prices.
Optical companies like $POET are trading at negative enterprise value — literally $97M below net cash. The market is valuing the operating business worse than zero. How is that a bubble?
Meanwhile, AI infrastructure players with real revenue visibility are signing massive long-term agreements: $COHR just locked in $20B, $LITE signed $5B. These aren't speculative bets — they're contracted cash flows over 3-5 years.
Yes, there are pockets of froth:
• Private AI startups like Instinct look overvalued
• Overleveraged neoclouds that misjudged their financing
• AI lab valuations that deserve skepticism
But lumping profitable, cash-generative infrastructure companies with single-digit P/Es and contracted demand into the same "bubble" bucket is intellectually dishonest.
What we're likely seeing isn't demand destruction — it's the Great AI Deleveraging of 2026. Valuations compressed from macro fear, rate uncertainty, and regulatory noise. Not because the buildout stopped.
As earnings roll in and free cash flow piles up, the market will eventually separate durable infrastructure plays from speculative froth. The bottleneck names especially look like asymmetric setups here.
I'm long the deleveraging, not the narrative.