QuantPM

Coatue getting long semis had to be the signal…. too obvious.

@yacineMTB respect

she is an emacs user. or was i guess. coding is for plebs

my mom is using codex

The nasdaq has become a giant long semis short software trade. This is the main driver now. It is no longer the mag7.

@RihardJarc This take is wrong IMHO. If we’re in an overbuild situation it’s not like compute services would be selling like hot cakes. Plus you can’t really sell compute overnight in a balanced market. You need sales, you need other things. There are better bull cases for $META than this.

The market reaction of punishing $META because of the increase in CapEx will turn out to be very wrong IMO. In the $META Q325 earnings call, Zuck already explained well that if $META overbuilds their compute infrastructure for internal needs, they can sell it to external parties. We are in one of the biggest compute demand/supply imbalances that seems to be getting worse, not better. $GOOGL GCP CEO Kurian expects demand to be bigger than supply for 10 years. $META's enormous AI compute capabilities will either return ROI in the form of their products (which they are already showing) or/and give great ROI by selling compute to external companies. $META investing in AI data center compute is a big asset, not a cash burn, similar to what metaverse investments were in large. The market is viewing it as a negative, instead of an enormous strategic advantage in a compute-constrained world.

@zephyr_z9 Does AAPL compete for capacity with NVDA, or are those two different foundry processes?

Less to do with derisking, more to do with Apple not getting enough capacity at TSMC Btw, they are outsourcing lower end SKUs

@TheEdgeStreet Jesus man 2022 Metaverse was a diasater. The stock was down 70%+. The reason it went up was because Zuck stopped out going all in on the metaverse. Terrible example.

$META is down 10% this week despite beating earnings. It’s still 30% off its all-time high. Trading at $614 right now. Here’s why that’s interesting. Meta has 3.27 billion daily active users. Ad revenue growing. AI tools making those ads more effective. Operating margin above 40%. The market sold it off on capex concerns — $115–135 billion in AI infrastructure spend guided for 2026. But here’s the thing: Every time Meta has been punished for investing aggressively in the future — it has paid off. 2022 — everyone said Zuckerberg was destroying the company. The stock fell 75%. Then it went up 800%. The pattern: market hates the spend. Rewards the returns. We’re in the hate the spend phase right now. At $615 — that’s worth paying attention to. I will be adding to my $META Position today. 📈

@MetacriticCap You’re uninformed. It’s still the best multimodal model. Try eg claude for anything involving images. Or try gpt for anything that proper measurements and aspect ratio is important. Gemini is behind in coding for sure, but still SOTA in many things.

Gemini is SOTA in the sense that everyone thinks that using it within Cursor, Antigravity, or Opencode would be great. However, no one actually uses Gemini with these tools, but they still respect Google DeepMind too much to remove it from the SOTA conversation.

A sensible trade is to buy Meta and hedge with something cyclical and/or exposed to Trump’s shenanigans in the ME. Airlines for example.

@R_and_Invest That’s exactly right. Meta, unlike the CSPs doesn’t benefit directly from the build out and as such is more vulnerable to the macro headwinds. Also it has 22% of revenue coming from Europe.

Probably a very smart tweet. But unfortunately, people dont realize oil kills 5B consumer's spend outside US which directly impacts 60% of $META's business. The day crude touches $60, $META whether RL is shut down or not, is going to leap frog to $1,000.

👀

Yep

@eric_seufert In answering that question he asked in the end if Susan Li wanted to add anything… she could and probably should but crickets.

Despite beating consensus on Q1 revenue and Q2 guidance in its earnings release last night, Meta's stock is trading down nearly 10% pre-market this morning. Meta has proven that its AI investments into content and ads ranking have delivered real commercial value: 19% impression growth and 12% growth in average price per ad in the quarter, with ad revenue up 33%. An analyst asked on the earnings call what "signposts" Meta uses to determine if its CapEx (which guided up by $10BN for the full year) is generating ROI. I think a useful answer could have been, "If our impression growth is accelerating while our price per ad growth is accelerating and DAP is essentially flat, it means our AI investments are generating return." Obviously, the break-even calculation is more nuanced, but to suggest that these investments are still speculative strikes me as detached from reality. These investments are clearly delivering commercial value. The problem is that Meta has lost control of the narrative: Mark Zuckerberg only wants to discuss AGI and 'superintelligence,' not the high-growth, $50BN-per-quarter business that sustains the company. That's fine, but it's redolent of the Metaverse fiasco, and Meta needs another voice that is willing to champion its ads business. It had that in Sheryl Sandberg, but no one plays that role now. Until someone can articulate the very concrete current returns on these investments, analysts will continue to focus exactly on what Mark Zuckerberg seems anchored to: superintelligence, personal AI, foundation models, etc. And these are interpreted by outsiders through the most superficial lens.

The winner AI consumer app will have to be monetized with ads directly or indirectly. Two companies can do that today, $Googl and $meta (OAI can’t do ads). One is pushing 30x and the other 15x. Seems like a blend of the two is a good bet.

Not so fast… problem with $META is they’re playing catch up on compute and market doesn’t like it one bit. But roadmap much clearer now with muse. At 36 eps estimate (bbg) with $612 premkt puts them @17x fwd which is cheap given the topline growth.

Insane

@dwarkesh_sp @reinerpope This is awesome

Did a very different format with @reinerpope – a blackboard lecture where he walks through how frontier LLMs are trained and served. It's shocking how much you can deduce about what the labs are doing from a handful of equations, public API prices, and some chalk. It’s a bit technical, but I encourage you to hang in there - it’s really worth it. There are less than a handful of people who understand the full stack of AI, from chip design to model architecture, as well as Reiner. It was a real delight to learn from him. Recommend watching this one on YouTube so you can see the chalkboard. 0:00:00 – How batch size affects token cost and speed 0:31:59 – How MoE models are laid out across GPU racks 0:47:02 – How pipeline parallelism spreads model layers across racks 1:03:27 – Why Ilya said, “As we now know, pipelining is not wise.” 1:18:49 – Because of RL, models may be 100x over-trained beyond Chinchilla-optimal 1:32:52 – Deducing long context memory costs from API pricing 2:03:52 – Convergent evolution between neural nets and cryptography

I can sense $META got it’s mojo back.

If you need any further signs of froth in this market, $QCOM going up 24% in 2 days on two different rumors: going to dabble in CPUs on Fri and allegedly working on smartphone with OpenAI today.

@Steve_Yegge Have it crossed your mind that perhaps GOOGL wants its engineer to use their tools to generate data to improve their RL to eventually catchup with rivals? That would make total sense actually.

My tweet last week about Google's AI adoption drew a lot of pushback, to say the least. Since then, Googlers from multiple orgs have reached out to me independently and anonymously. They've expressed fear of being doxxed, concern about what they saw as bullying of me, and general corroboration of my original tweet. I haven't verified each person's story, but the picture these Googlers paint is consistent across sources. It is more specific than what I originally wrote, and somewhat bleaker. What they describe is a two-tier system. DeepMind engineers use Claude as a daily tool. Most of the rest of Google does not. When the question of equalizing access came up internally, the proposed response was to remove Claude for everyone — which DeepMind objected to so strongly that several engineers reportedly threatened to leave. Non-DeepMind engineers get pushed onto internal Gemini variants behind router-style names that obscure which underlying model is actually serving a request. Multiple engineers describe regressions and reliability problems severe enough that some senior people have stopped using the tools. A senior manager on a major product line reportedly flagged attrition concerns over exactly this issue. Googlers say leadership knows the gap is real. The response has been to mandate AI usage in OKRs and individual expectations, and to stand up an internal token-usage leaderboard. Unfortunately, managers have been told both that the leaderboard won't be used for performance reviews and, separately, that it absolutely will. And I hear other stories that Google's culture is not adapted properly yet for high-volume coding. Addy Osmani's reply on behalf of Google said over 40,000 SWEs use agentic coding weekly. I don't doubt the number. But weekly use of a thin tool is precisely the box-checking I described in the original post. Volume of opens isn't adoption — and "weekly" is a low bar that includes a lot of people who tried it once and went back to writing code by hand. The clearest thing I'm hearing is that Googlers do want to use high-quality agentic tools. They are asking repeatedly for better ones. But overall, this is not a picture of an engineering org that is fine. My goal in the first tweet, and now, is always the same — get more people using AI and agentic coding. Nobody is as far ahead as they might look from the outside, and none of you are as far behind as you might be worried you are. To all the Googlers who've reached out: thank you. You took a real risk and I appreciate you. Be safe. And good luck getting good models!

lol