In the AI knife fight, Eisman wants you selling the knives

Steve Eisman, the investor best known for spotting the subprime collapse, has a prescription for anyone nervous about picking winners in the AI buildout: stop betting on which chip designer swings the knife, and own the outfits selling it. The line is his: "In this knife fight you want to be the one selling the knives." The picks he highlighted, drawn from a subscriber email, are ASML, KLA, Lam Research, Applied Materials, Micron and SK Hynix. The five with year-to-date figures quoted have all beaten NVIDIA's 17.36% gain. And Eisman attaches a warning sharp enough to cut his own thesis: the AI ecosystem is "dependent, almost completely dependent on the health of Anthropic and OpenAI", and if OpenAI fails within a year "there will be a massive correction in the stock market".

ASML Watch analysis

What happened

The argument is industry structure. Every credible path to frontier AI runs through lithography, etch, deposition, memory and process control, so the suppliers collect revenue whichever designer wins the model race. Eisman highlighted a subscriber email making exactly that case: the safest way to play the buildout is owning upstream suppliers rather than betting on any single chip designer.

The ranking accompanying his remarks, from 24/7 Wall St., puts ASML first on its EUV monopoly and multi-year order coverage, Applied Materials second on record DRAM and HBM exposure, then Lam Research, Tokyo Electron and KLA.

The company numbers doing the lifting, as reported there: ASML's Q2 2026 revenue reached $10.65bn, up 21.3%, with operating margin at 37.1%, full-year guidance raised to $49.11bn to $51.40bn, a $45.06bn backlog and a $12bn buyback running to 2028. CFO Roger Dassen says 2027 low-NA EUV orders are "close to being fully covered", and CEO Christophe Fouquet calls DRAM conditions "the perfect storm for ASML on DRAM this year and most probably the next few years to come". Applied Materials posted fiscal Q3 revenue of $9.12bn, up 24.8%, its 13th consecutive quarter of year-on-year gross-margin expansion, and plans to double quarterly system output by 2028.

The numbers

• Micron, up 241.54% year to date
  The list's runaway leader, riding the memory and HBM buildout.
• Applied Materials, up 78.46%
  Record DRAM and HBM exposure, per the ranking, with Q4 guidance near $10.25bn.
• Lam Research, up 75%
  Etch and deposition steps multiply as 3D memory scaling stacks the layers.
• ASML, up 59.77%
  The EUV gatekeeper, with a $45.06bn backlog behind the thesis.
• KLA, up 49.35%
  Process-control intensity rises with every node shrink.
• NVIDIA, up 17.36%
  The designer the knife-sellers are all beating this year.

Why it matters

Picks and shovels is the market's hedge against the model race: you need not know whether NVIDIA, AMD or a custom chip wins if every contender buys the same lithography, etch and memory. Backlog turns that argument into arithmetic, and ASML's $45.06bn backlog with near-full 2027 low-NA coverage is the strongest version of it.

Eisman's caveat is what stops this being a victory lap. If the spending behind those order books traces back to a handful of frontier labs, and he names Anthropic and OpenAI as the load-bearing walls, the suppliers are less a hedge against the designers than a leveraged bet on two companies' solvency. The past month hints the market is already testing the idea: Applied Materials down 16.71%, KLA down 13.08% and ASML down 8.96% over that stretch, per the same figures.

Our read

The structure argument is coherent and the figures are as reported, but a ranking is one publication's judgement and Eisman's view is opinion, not a measured fact. What earns the eyebrow is that the thesis carries its own refutation: the safest-looking trade in AI rests on the two labs its author says would trigger a "massive correction" if one stumbled. Treat the list as a map of who gets paid during the buildout, and track order books and hyperscaler capex guidance rather than model launches. This is commentary, not investment advice.

What to watch

• Whether ASML's near-full 2027 low-NA order coverage survives the next earnings season intact.
• Hyperscaler and frontier-lab capex guidance, the funding source behind every backlog on the list.
• Whether the one-month pullbacks in Applied Materials, KLA, Lam and ASML deepen or stabilise.
• Any sign of funding stress at OpenAI or Anthropic, the dependency Eisman himself names.

Discussion spark: If the supply chain is, in Eisman's words, almost completely dependent on the health of Anthropic and OpenAI, does owning the knife-sellers actually diversify the risk, or just delay the same bet by a quarter?

Independent WittyWires tracker for public updates about ASML. Not affiliated with or endorsed by ASML; this is not an official account.