RS

@MorningBrew ahh yes, I remember the absolute economic wasteland of the 1990s and mid-2000s when everyone was famously broke

THIS AM: The 30-year Treasury Yield has hit 5.18%, its highest level since July 2007 Don't look up what happened next

Cerebras (solana:4yEjcMiy6GAgrpWpUvhUXfaP1vQmJXfqJjEyxBSZpump) indicated to open ~$360. That is ~$110B market cap on $510M of 2025 revenue. 216x trailing P/S. ~48x on 2027E (est. $2.3B). For reference: $NVDA at its peak AI-froth multiple in mid-2024 traded around 40x trailing. At that point NVDA was a $100B< revenue business growing 200%+ YoY with the CUDA moat fully baked in. CBRS: $510M revenue, 76% growth, two-customer base. The market is paying more for projected revenue than it paid Nvidia for delivered. Customer concentration is also not being priced in. Cerebras pulled its 2024 IPO because 80% of revenue came from one customer. In 2025, 86% came from two UAE entities (G42 + MBZUAI). Going forward, the story flips to OpenAI: $24.6B backlog, majority tied to one buyer Of that backlog, only $3.7B is contractually scheduled across 2026 + 2027. The remaining ~$21B stretches through 2028 - 2030+ and depends on physical infrastructure being built The OpenAI relationship is also circular. OpenAI is an equity holder AND lent Cerebras $1B in working capital against the chip commitment At this price, you are underwriting: flawless execution by Cerebras and its customers, no macro deterioration, 30-40% beats on growth and guidance, and material unforeseen upside on top.

Great data point. I would bet we’re early in adoption rn and after 12 months, looking back at p&ls, uncapped compute spend without clear demonstrable roi becomes a real point of convo in boardrooms. Lots of employees are being told to token max and experiment. This is going to prove very costly over time and CIOs are going to have to optimize this.

@amir @_Reeesh I think we're within 6 months where this flips. We went from cursor to claude code to codex to now seriously planning for open source in the last 6 months. The growth demanded by these companies naturally is not aligned with users

something interesting happening as Anthropic raises prices & its models cost more to use: customers are just eating it.

@NovusOrderM Objectively The single digit fwd p/e signal as a no brainer has been a mistake many have made in the past. Today's counterargument is more that there has been a structural realignment in the memory buying circle, which I can buy

@NovusOrderM This isn't the first time Micron ($MU) has traded at a single digit fwd P/E ratio. The Micron trade has always been "buy when Fwd P/E is high and sell when its low" due to buying cycles. Investors have been wary that we are in cyclical peak, as we seen many times before.

In 1999, a company like Micron would have been trading at a far higher forward PE ratio. 20/30/40x. Today, it's trading in the single digits. During the dot-com bubble, you had a whole cohort of market participants buying stocks simply because the prices were skyrocketing. These retail investors weren't scrutinizing financials - they were succumbing to human irrationality and emotion (FOMO). This was, in part, because retail investors didn't have easy access to information (company financials, competing ideas, etc.), and so they weren't rigorously scrutinizing investments. Today, you have retail investors who can use AI to scrutinize any company and its financials. You have AI that interrogates assumptions and will surface second-order effects, which forces these investors to actually defend their thesis. And you have online town halls (x, reddit, etc.) where competing ideas can collide. AI isn't perfect, and the AIs probably do converge on "rational" conclusions, but it does raise the floor of market discipline in a way that 1999 lacked. In 1999, the Nasdaq was trading at ~200x PE and a forward PE of ~60x. Today, the Nasdaq is trading at ~37x PE and ~24x forward PE. I believe that AI + social platforms + accessible information are keeping the market far more rational than any prior economic revolution.