AI is a capital formation cycle, not just a tech cycle (with Caroline Hyde, Ed Ludlow)

Poorvi Agrawal of Altimeter Capital (investor in OpenAI, Anthropic, SpaceX) argues AI has become the largest capital formation cycle. Key distinction: companies receiving CapEx (compute, energy, memory, networking) vs spending CapEx (the labs). She sees strong demand for IPOs and believes more participation is better. Her mental model: avoid businesses hurt by scaling laws (where more compute = more intelligence).

Yields

explicit

RUT

Oil

Metals

USD

Altimeter Capital 8.2
Hedge Fund $3.50B
Poorvi Agrawal 8.8

AI is a capital formation cycle, not just a tech cycle (with Caroline Hyde, Ed Ludlow)

ndx

More than half of speakers were long picks-and-shovels providers (compute, energy). Energy is the ultimate bottleneck at scale.

3 calls

+12

slightly better than random

10/15/2025 10:00:35 PM	long term up	60 days later	+1.66%	+1.66%
10/15/2025 7:44:34 PM	medium term cautious up	20 days later	+3.90%	+1.95%
10/15/2025 3:58:10 PM	medium term up	20 days later	+3.90%	+3.90%

Show all 3 ndx results

Ed Ludlow

Anthropic filed confidentially, Altimeter led that round. Alphabet upsized to $84B equity offering. What does the capital market tell you?
Poorvi Agrawal

AI has gone from a technology product cycle to one of the largest capital formation cycles. On one side: picks and shovels (compute, memory, optics, networking) receiving CapEx. On the other: the labs spending CapEx. This is the biggest mental model—are you on the receiving or spending side?
Caroline Hyde

Is there enough room for all these large language models? Will they be commoditized?
Poorvi Agrawal

Companies aren't binary (consumer vs enterprise vs space AI). OpenAI has ~1B ChatGPT users, 5M Codex enterprise users, plus hardware/robotics bets. SpaceX is the two largest markets: space and AI. XAI is the 'Switzerland' approach—selling compute, models, coding products.
Caroline Hyde

Do you have any pause about the wall of public offerings and whether there's enough demand?
Poorvi Agrawal

IPOs are cyclical. Last big cycle was 2021. There's a lot of demand for these products. In December we crossed a capability level with coding models (Opus 4.5) that pushed to the next frontier. More participation is better for us.
Ed Ludlow

What's one idea you're long, one you're short? Why is that a useful framework?
Poorvi Agrawal

More than half of speakers in my Stanford class were long picks-and-shovels providers (compute, energy). Energy is the ultimate bottleneck at scale. Most common short: incumbents not innovating. My mental model: avoid businesses hurt by scaling laws—if more compute/resources/data makes intelligence go higher, that's a tough place to be.

SignalTube

markets at a glance