←
Matt Dratch2,987 posts
Matt Dratch avatar✓

Matt Dratch ✓

@DratchCap

Macro Equity PM | Dartmouth Fball ’08 “Be bold, and mighty forces will come to your aid.” Views expressed are solely my own and not on behalf of my employer.

◷ Joined December 2010▧ 223 Photos and videos
Matt Dratch avatar
Matt Dratch@DratchCap·Oct 19

DratchCap Mkt Update. Some thoughts and self criticism. Been a brutal couple of weeks. An equal-weight AI basket I track fell ~24% in seven sessions vs ~6% for QQQ, which was the worst 7-day window in five years, worse than DeepSeek or July. Thinking about my own mistakes, I did not respect price action and rates enough, especially given the financing reflexivity broadly. Index shorts did start to work but I did not press the way I should have. Another lesson re-learned. In these types of environments I take things day by day and look for ways to improve the shape of the portfolio. Sometimes options structures line up well for that and can be good stock replacements. Stay in the game, reduce the tail. Prior to this AM, things looked aggressively oversold. De-grossing has been material (see MS HF data today, last week 3-sigma event) w/ AI long/short near 5yrs lows. From here, I think the bigger near-term risk is macro. People have been hiding in quality, and the systematic community is still long because the index has held up much better. If the quality rolls, there's plenty of room for the index to catch down. Meanwhile, the tech demand indicators I'm tracking remain extremely strong. GPU availability is near its lows, rental rates are rising across chips and tenors, and TSMC is stepping up multiyear investment. Recent buildout data points (@SemiAnalysis_ today) are also encouraging. And my recent lab checks have been very strong too, with newer models driving demand and more capacity finally coming online. At the same time, the market is treating funding as broken across the board. But the big platforms can still raise at prices that make sense given their ROI / ROE and strategic imperatives. A lot of the neoclouds are also financing against contracts and prepayments from those same customers. Some valuations are also getting extremely attractive. Memory is a good example. Demand is compounding much faster than supply, pricing is going up, and yet the stocks trading near historical trough on fwd book w/ 30-50% type FCF yields. As for the rest of the year, there is a Mountain of Worry to climb... Strong megacap prints (soon) and better lab data could get people front-running the November seasonal (predictable AI era pattern). And that loop can feed FOMO into the Anthropic IPO. Expectations there have reset and HF positioning is now light in AI. Add in the potential for the Fed to hold, power deals to start clearing post midterms, or some relief on the war and rates. We don't need everything to go right but a few will do. I like to think of the AI buildout era as a super boom w/ mini busts along the way, often when price drives narrative. The mkt is more vulnerable to that than ever bc of new fast info channels (X), and the intangible and exponential nature of the tech. Human's aren't wired to think that way, and skepticism is always more cosmopolitan. On top of that a lot of the people making the big decisions are removed from the details and don't use the tech day to day, so when the tape turns they have limited first principles to fall back on. Staying the course during these moments has been right each time and I think that will prove true again. Good luck out there!

Matt Dratch avatar
Matt Dratch@DratchCap·9h

Let’s step in the Time Machine and go back to another crazy period, Feb 2025. DeepSeek drops R1, trained for peanuts (supposedly), NVDA loses $600Bn in a day, and everyone says efficiency just killed the AI stars. The entire world had maybe ~15 GW ish of AI compute at the time. And since then all the fears came true: the cost of doing the same hard task fell ~99% and cost per token at fixed capability fell 10x a year. Effective compute is up something like 15x by end of this year. By the bear logic nobody should have been building anything from that moment on given the ‘supply monsoon’ from stacked compute and efficiency. Meanwhile here we are on the verge of ANT’s public market debut, a co that went from a $1B run rate to $81B in 19 months and is potentially on its way to ~$150+ in 2027. We literally ran this exact compute supply experiment just under two years ago and are living in the results. So, are we finally overbuilding? The prices certainly suggest that! Or do the price setters have trouble w/ exponentials? We’ll know soon enough... Back to the Red Room for me.

Post attachment
Matt Dratch avatar
Matt Dratch@DratchCap·13h

Cue the HF vibe shift: 'What is this, an allocation for ANTs?' 😜

Matt Dratch avatar
Matt Dratch@DratchCap·18h

Right so 1/2 went to China and we didn’t know… or it’s customers billing subs. Rorschach test! Dell was 9bn alone thru its Dutch sub.

Matt Dratch avatar
Matt Dratch@DratchCap·19h

Luckily *and thankfully*, you’re a citizen in the land of second chances…

Matt Dratch avatar
Matt Dratch@DratchCap·20h

This is all just modern Paganism and a Tower of Babel replay. History rhymes indeed. Good heads up from @bgurley

Matt Dratch avatar
Matt Dratch@DratchCap·Nov 17

First Ai stocks and now this, have to come to grips with my ranking in the household

0:18
Matt Dratch avatar
Matt Dratch@DratchCap·Nov 16

The move today on remote access is more aggressive geopolitically than it seems on the surface. IFP has remote access at ~32% of China’s 2027 AI compute when counting flops (h/t JPM's Cembalest). But the rented compute is the newer NVDA in giant, coherent clusters. A lot of what China has at home is older / scattered / and short HBM. Adjust for quality and the rule likely hits China’s frontier work harder than 32% implies, and it covers inference too. While some argue this makes us safer, many forget the other side gets a vote and China may respond.

Matt Dratch avatar
Matt Dratch@DratchCap·Nov 16

On the negative side, we are now in a 4 front war Iran. China escalation. Inflation (hikes!). AI buildout. On the positive side, GTA 6 launches Thursday.

Matt Dratch avatar
Matt Dratch@DratchCap·Nov 16

As I sit in the Red Room today contemplating my life choices, I am heartened by the AMZN / BXDC deal. It's not as click baity as the Jensen drama but it's quite interesting. AMZN is selling 2026 vintage DCs in meh areas at a 7% cap, i.e. BXDC is underwriting a lot of terminal value. Why is AMZN doing this? Bc you sell an asset at a 7% yield and recycle the capital into much higher ROIC opportunities. You pick someone like BXDC bc they have a real pipeline to keep doing these. And AMZN is leasing these back. They would NOT sign up for years of fixed rent if a capex pullback was on the table.

Matt Dratch avatar
Matt Dratch@DratchCap·Nov 16

Plus the sweet distillation juice

2:21
Matt Dratch avatar
Matt Dratch@DratchCap·Nov 16

The US press (and parts of the Admin?) are not doing their job on NVDA and Singapore. OR they have an alternative agenda 🤔. It is fairly easy to find out what a "Singapore sale" to a large US customer actually looks like. This is all gettable public info… I had agents request filings in Singapore for known subs of the big US buyers. They got some but not all. Then a curious friend passed along the Dell file (h/t @RandianCapital). This of course begs the question of why anyone is pushing this narrative when it's this easy to check. NVDA may well have real questions to answer on China or Megaspeed, but this Singapore angle is likely just not it… So here is one example. Dell's global buying hub is Dell Global B.V. (Singapore Branch). It's a Dutch co w/ a Singapore address and a US parent, also audited by PwC. In the yr ended Jan ‘25 it did $56.6bn of rev (59% of all of Dell), nearly all of it selling to other Dell units, and 55% of that was for the US. By my math it bought about $30.8bn from outside suppliers, up from $20.2bn the year before. The filing says it holds "AI commodities" in inventory, and raw materials doubled. Dell shipped $9.8bn of AI servers that year. If GPUs are ~2/3 ish of the price of an AI server, that's $6-7bn of GPUs. NVDA reported $23.7bn billed to Singapore in the same fiscal year. So **one US customer** could plausibly be about a quarter of the "Singapore" line by itself. I’ll let others go pull the Singapore filings for other large known GPU buyers… To be clear, the filing doesn't name NVDA or any supplier, the GPU share is my estimate, and I don't know who the other billers are. But this is a US company whose Singapore branch buys the parts, holds AI components in inventory and sells most of it on to the US. Under NVDA's old billing location table that is a Singapore sale, the same ones the press say “disappeared”. And thanks to @RandianCapital for the spirited and thoughtful debate. @GavinSBaker @elonmusk @Jason @balajis @TimSweeneyEpic