Totally get the sentiment here, this feels like a necessary smack of reality in a world where everyone’s selling the future as something that’s already arrived. The real work and risk is in execution and discipline, not the shiny slogans or rallying cries, and if we ignore that we’re just painting over the tough bits with louder language.
Three items here connect into one macro thesis: we are in a capex-driven regime where compute, power, and industrial capacity become the limiting factors.
1. Meta x NVIDIA
A multi-year infrastructure buildout anchored in Blackwell and Rubin, plus networking, plus a meaningful CPU posture shift, signals that hyperscalers are optimizing systems-level performance per watt, not just buying GPUs. Also, the size of the 2026 capex guidance is the tell. This cycle is not cooling.
2. The Fed, AI, and prediction markets
The minutes read like an acknowledgement that AI is both a productivity tailwind and a forecasting risk. The Kalshi angle matters because it pushes markets closer to real-time probability distributions, which can tighten reaction loops across rates, risk assets, and crypto.
3. J&J’s cell therapy facility
This is a reminder that “AI” is not only software. It is also applied manufacturing, workforce development, and domestic capacity expansion in regulated, high-margin verticals.
Net: the portfolio question is not “AI or no AI.” It is “where does the bottleneck form” (compute, power, networking, specialized manufacturing), and who captures pricing power when demand exceeds supply.
Interested to hear your thoughts on the partnerships of NVDA vs. AMD with META.
Feel like AMD is being defensive (going up 10% of business), but META is also buying into their chips+stack and they do inference better than anyone else.
Totally get the sentiment here, this feels like a necessary smack of reality in a world where everyone’s selling the future as something that’s already arrived. The real work and risk is in execution and discipline, not the shiny slogans or rallying cries, and if we ignore that we’re just painting over the tough bits with louder language.
What is your take on this? https://x.com/citrini7/status/2025653614430023864?s=46&t=KmVnI0vA5iaMma61Sg2K0Q
Three items here connect into one macro thesis: we are in a capex-driven regime where compute, power, and industrial capacity become the limiting factors.
1. Meta x NVIDIA
A multi-year infrastructure buildout anchored in Blackwell and Rubin, plus networking, plus a meaningful CPU posture shift, signals that hyperscalers are optimizing systems-level performance per watt, not just buying GPUs. Also, the size of the 2026 capex guidance is the tell. This cycle is not cooling.
2. The Fed, AI, and prediction markets
The minutes read like an acknowledgement that AI is both a productivity tailwind and a forecasting risk. The Kalshi angle matters because it pushes markets closer to real-time probability distributions, which can tighten reaction loops across rates, risk assets, and crypto.
3. J&J’s cell therapy facility
This is a reminder that “AI” is not only software. It is also applied manufacturing, workforce development, and domestic capacity expansion in regulated, high-margin verticals.
Net: the portfolio question is not “AI or no AI.” It is “where does the bottleneck form” (compute, power, networking, specialized manufacturing), and who captures pricing power when demand exceeds supply.
Very interesting read, like it.
Interested to hear your thoughts on the partnerships of NVDA vs. AMD with META.
Feel like AMD is being defensive (going up 10% of business), but META is also buying into their chips+stack and they do inference better than anyone else.
Interesting, circular dynamics at play! lol
Construction workers and Civil engineers eating good after reading that chart