Howard

How im playing the blokcbuster IPOs. $CBRS – yes $SpaceX – no $Anthropic – no $OpenAI – no $CRBS: Nvidia competitor for inference, new architecture, memory advantage (maybe), OpenAI weirdness, sensible valuation(ish), enormous tailwinds. Hottest thing in the world right now. People always whinge about companies staying private too long and reckon they'd have been buying $MSFT and $GOOG at IPO. It's not that easy. This is what it looks like. Hard decisions on unproven companies, trades on promises and vision, not hard data. If you want hard data you can buy $WMT. Buying a small position. Enormous tailwinds + (probably) retail hype push this higher. OpenAI deal is $20bn over 3 years, so $26bn valuation puts it at 4x P/S on that contract alone. Not crazy in a potential wonder stock. Scam Altman concentration risk though. SpaceX: Too expensive. Starlink is the jewel. $11.4bn revenue, $7bn EBITDA, growing 50%. But $1.75tn valuation = 150x sales on the only profitable bit. Going to face competition from $AMZN's Kuiper and ASTS soon. Their whole launch thesis was costs coming down. Thats the bear case bro... Data centres in space aren't real. xAI hasn't hit the heights of the other labs. Starlink's nice, just not nice enough to pay $1.75tn for the whole package. Ant + OAI: I don't believe the models have a defensible moat. Anthropic's overtake was the harness (Claude Code), not the model. OpenAI launched Codex and people are flocking back. No stickiness. Maybe a bit more on Anthropic's business side, but most IDEs swap models easily, and soon price sensitive businesses will move to open source when good enough is good enough for most tasks. Huge training + inference costs flow to hyperscalers and neoclouds, on a massively depreciating asset. How much is Sonnet worth now? Who cares about the second-best model? Mythos in 6 months is yesterday's news. The models are enormously valuable in the sense they will create lots of value. I'm not convinced that value is captured by the lads themselves. Did Google make money off Gmail? Off Maps? Huge value created, captured elsewhere.