← Analysis page  ·  Steve Eisman hub  ·  Research hub

Actionable insights — The AI Companies Are Trying to Manufacture a Crisis

Not what Eisman thinks about AI, but how he reads the industry: repeatable checks you can rerun on any hot sector.
2026-SEP-17 · CNBC Squawk Box · ▶ Watch · full analysis · transcript
How to read this page: each insight is a method Eisman used or described in this clip, written as steps you can rerun later, followed by how it played out here and the signal to watch. A 7-minute TV hit supports only a few.

1:32 1. Read an incumbent's call for regulation as a moat-building move

The repeatable method
  1. When a market leader loudly asks to be regulated (on safety, ethics, "systemic risk"), ask first what its competitive position is doing.
  2. Check for moat erosion: are cheaper or open substitutes taking share? Is pricing power fading?
  3. If share is slipping, assume the regulation push is at least partly an attempt at regulatory capture: licensing, compliance costs or rules only large players can meet.
  4. Judge the incumbent on its economics without the regulation, and treat any capture as upside that may never arrive.
Here
Token maxxing "is over," open-weight models are "taking big market share," the labs have "no moats," so the AI-safety alarm reads as "trying to manufacture a crisis" to get regulation they "can then manipulate to create the moats to create the duopoly." Result: OpenAI Negative.
Watch for

3:06 2. Find the weakest load-bearing customer in a spending chain

The repeatable method
  1. Map the chain of who pays whom (suppliers → intermediaries → end buyers).
  2. Find the few end buyers whose spending funds a large share of the whole chain.
  3. Rank those buyers by financial strength and flag the weakest as the chain's single point of failure.
  4. Size exposure to every upstream name by how much it depends on that weakest link, and trim rather than add while the link is unproven.
Here
Chain: NVDA → hyperscalers → Anthropic / OpenAI, which make up "such a huge percentage of the entire chain." Weakest link: OpenAI ("the weaker company," IPO postponed). Action: "I've taken some off the table" and isn't adding.
Watch for

3:39 3. Treat a postponed IPO as a weakness signal

The repeatable method
  1. When a hot private company delays a listing it has talked up, don't accept "market conditions" at face value.
  2. Compare it with its closest peer: if the peer is still heading to market, the delay points to company-specific weakness.
  3. Test the company's rhetoric against its actions: a firm that says "slow down, it's dangerous" while racing to list for money is telling you its priority.
Here
"OpenAI is the weaker company. I think that's one reason why they postponed their IPO," while Anthropic is still expected to list first. On the safety rhetoric: "really postpone your IPO."
Watch for

Methods distilled from the public CNBC Television YouTube clip (captions end at 6:54). Not investment advice.