Return to sell-side pending
Joined January 2018
- Tweets97
- Following1.3K
- Followers784
- Likes985
Commodity Adjusted EBITDA retweeted
Signed on for a new role in institutional equities based in NY. Will be moving in the next month or so. If you're based in NY and like talking stocks (particularly Canadian cyclicals) would love to connect. DM's open
Commodity Adjusted EBITDA retweeted
No position in the stock, but $CRWV flagged an A100 recontracting out to 2029. I've been noodling on who the counterparty could be, and my theory is it's $NVDA. A couple of reasons why I think this makes the most sense.
It's clear the hyperscalers are trending away from GPUs as they scale their TPU ecosystems. That creates both urgency (see NVDA's $500B deal with the mega-cap PE firms) and incentive for NVDA to help the neoclouds scaling inside the NVDA ecosystem to defend its moat. Recontracting A100s out to 2029 can dramatically improve the neoclouds' cost of capital (equity and debt) especially when CRWV management is light on the details.
On IR callbacks, CRWV said the A100s are "inferencing for a customer." That could just be NVDA running internal research workloads. Who else could be so certain they'll still want an A100 in 2029? How about the guys best placed to optimize for it. It would make sense that NVDA keep building more efficient versions of Nemotron that can best leverage older-vintage GPUs, and point them at the internal coding workflow.
If NVDA does turn out to be the counterparty, the market shouldn't read this recontracting as a broader uplift in GPU terminal values because NVDA has a clear incentive to prop up the neoclouds to defend against the $AMZN and $GOOG TPU ecosystems.
Commodity Adjusted EBITDA retweeted
Around a month ago, a couple of hedge fund friends pitched me on $SNDK, $MU, $SKHY , and the rest of the memory names as their best ideas. The thesis was compelling, and the guys are very smart. Their main argument was simple: these companies were incredibly cheap. P/E of 10. P/E of 8. Some even lower.
The interesting part is that not a lot has changed fundamentally since then, yet many of these stocks are down around 50%.
We can extrapolate that, since nothing fundamentally changed and the stocks have been cut in half, sentiment is more important than multiples, and you never want to buy at peak sentiment, no matter how cheap the deal looks.
Read more about it in my post below, “The Myth of P/E Ratios.”
Commodity Adjusted EBITDA retweeted
I’m long some these AI stocks like the rest of you degens but my feed is currently flooded with people rationalizing that a 30-50% drawdown in some AI supply chain names make no sense because jevons paradox etc etc. I’ll make two market structure observations.
1. In cyclical industries like memory just like in commodities the stocks cannot simply be viewed as conventionally “cheap”. Stocks can lead the commodity price on the downside of the cycle. This happened in energy in 2018 when the $XOP peaked in mid-May while oil prices didn’t peak until early October but then promptly declined 40%+ by Christmas. All to say, stock prices are forward looking into perpetuity and will not always peak with/after underlying commodity prices (looking at you DRAM bulls)
2. People claiming a 30-50% drawdown is too severe are implicitly suggesting that pre-decline share prices were the efficient prices. This is hard to assert in isolation because as we’ve observed, a lot of the recent price action was driven by leveraged trading out of Korean retail. This was not fundamentals driven buying going in and an unwinding of the trade does not then necessarily represent a stock “decoupling” from its fundamentals. Retail’s ability to re-lever and buy back in will be a big part of determining if and when we can take out the previous highs.
