Larry McDonald: AI Mark-Ups Distort the Real Earnings of S&P 500 Companies

US stocks are hovering near all-time highs on October 12. The market is waiting for two events: the start of corporate earnings season and fresh inflation data, which will shape expectations for Fed policy. Against this backdrop, Larry McDonald, founder of The Bear Traps Report, said in a Bloomberg interview that AI-related mark-ups distort the "true profitability" of the S&P 500. Separately, he spoke about stress in the bond market.

What we know and what we don't

The source is a short teaser for a video interview. It contains no specific figures: no estimate of how large the distortion is, no list of companies, no earnings-per-share forecast. So it is worth separating two things. The first is McDonald's thesis itself: AI mark-ups make the index's earnings higher than they would be without them. The second is the mechanics through which this could work. Below, we walk through the possible channels. That is our interpretation, not a retelling of his words.

Where an "AI premium" in earnings could come from

In an investment context, a mark-up means revising an asset's value upward. In the AI story, there are at least three such points.

• Stakes in private AI companies. Large corporations hold investments in startups and labs. Each new funding round at a higher valuation can turn into accounting profit on the investment income line. Nobody has received any cash.
• Multiples in share prices. The market pays a premium for any link to AI. That inflates the market capitalization and index weight of such stocks, and with them the "average" picture for the S&P 500.
• Capex and depreciation. Data centers and accelerators are expensive, but their cost is written off gradually. If the assumed useful life of the hardware is optimistic, current expenses look smaller than they will be if the hardware becomes obsolete faster.

The first point is closest to what McDonald said about mark-ups. The second and third belong to the same debate over valuation and earnings quality.

Bonds and the Fed: why they are connected

McDonald mentioned bond stress, and the link is direct. Rising yields raise the discount rate. The assets that suffer most are those whose value rests on far-future earnings, and that describes AI stories exactly. If inflation comes in above expectations, the Fed will have less room to ease. Meanwhile, a market sitting at records is pricing in a fairly smooth scenario.

The combination looks like this: high valuations, earnings that include a share of paper revaluations, and sensitivity to interest rates. The market can survive any one of them. All three at once leave the index vulnerable to a bad surprise.

How to read the reports in the coming weeks

Earnings season offers a chance to test the thesis against data. A practical set of questions:

• Compare GAAP earnings with "adjusted" earnings. If the gap is widening, check what is driving it.
• Look in the reports for the "other income" and "gains on equity investments" lines. That is where revaluations of stakes in AI companies settle.
• Separate revenue from AI products (cloud, subscriptions, API) from investment effects. The former is money from customers; the latter is a paper valuation.
• Read the commentary on capex and server lifespans. Rising capex without faster revenue growth is a reason for caution.
• Watch government bond yields on inflation release days. How AI stocks react to rising rates will show how fragile the premium is.

For AI startup founders, the takeaway is simple: a high valuation in a funding round now affects the investor's financial reporting. That strengthens their negotiating position, but it also carries risk. If revaluations turn downward, the pressure returns to startups through reduced funding.

For investors and managers who hold index funds, the point is the same: "S&P 500 earnings" need to be read by composition. McDonald's position is the opinion of one analyst, and the scale of the distortion he describes is not yet backed by figures. But he has raised a fair question. The coming reports will show how much of the record earnings is real money from AI and how much is accounting effects.