Steroid-era earnings inflation
Steroid-era earnings inflation
One-line summary: Two AI-cycle accounting effects — hyperscalers depreciating AI gear slowly while the same spend books as immediate supplier revenue, plus GAAP write-ups of private AI stakes (Anthropic, OpenAI) held by public companies — are inflating S&P 500 earnings and record net margins, making the market look cheaper than it is; both effects reverse if capex slows or private AI valuations stop rising.
The insight
spencer-jakab (WSJ, "Earnings Forecasts Are on Steroids") argues that the unusually strong 2026 earnings picture is partly artificial. The tell: analysts raised S&P 500 EPS forecasts +3.4% mid-quarter, versus an average cut of 2.7% over the prior 40 quarters. Two mechanisms produce the boost:
- Depreciation asymmetry. Hyperscaler AI capex is immediate revenue and profit for suppliers, but the buyers push the cost through the P&L slowly via extended depreciation schedules — so aggregate S&P earnings are flattered on both sides of the same transaction while the spending accelerates, and penalized later when it slows.
- Private-AI stake write-ups. GAAP requires public holders (Nvidia, Alphabet, others) to revalue stakes in private AI companies each time a funding round marks them up. These non-cash gains equaled ~12% of Q1 2026 S&P net profit (Bao Lan Wang, University of Florida) and are preliminarily estimated at 2–3x that for Q2.
The result: Q1 2026 S&P net margin of 14.8% — the highest ever, ~2x the post-war average (and that operating figure excludes the stake gains). The reflexive loop: write-ups → reported profit optimism → higher next funding round → bigger write-ups.
Evidence
- spencer-jakab in 2026-06-30-podcast-the-compound-and-friends-steroid-era-with-spencer-jakab-the-f-s-going-on: "what's holding up earnings is that you have these hyperscalers spending a lot of money, but it's only going through the P and L pretty slowly because they've extended the depreciation of all these chips and equipment and servers that they're buying. And its revenue, it's immediate revenue and profit for other people in the market… there's going to be hell to pay for that later on when it goes into reverse."
- spencer-jakab in 2026-06-30-podcast-the-compound-and-friends-steroid-era-with-spencer-jakab-the-f-s-going-on: "This professor at the University of Florida, Bao Lan Wang, looked at it and he said… It was equivalent to, let's say, about 12% of the first quarter's net profit. And his preliminary numbers say it's going to be 2 or 3 times as large for the second quarter."
- spencer-jakab in 2026-06-30-podcast-the-compound-and-friends-steroid-era-with-spencer-jakab-the-f-s-going-on: "Net margins were the highest they ever have been 14.8% for the S&P 500… once you strip out all these abnormal things, the stock market's pretty expensive and margins are unlikely to stay this high for that long."
- The valuation math — spencer-jakab in 2026-06-30-podcast-the-compound-and-friends-steroid-era-with-spencer-jakab-the-f-s-going-on: "if you hold the P E ratio constant… then the market should be 25% lower. If you're justifying it based on those earnings." (15% → 11% net margin at a constant ~21x forward P/E.)
- The reflexive loop — spencer-jakab in 2026-06-30-podcast-the-compound-and-friends-steroid-era-with-spencer-jakab-the-f-s-going-on: "people see these big profit numbers for the hyperscalers and Nvidia that own big stakes in these things and then they say, well anthropic must be worth even more… And then the next round of funding is even higher."
- Independent corroboration (2026-07-14) — the reflexive-markup channel now cited across three more sources.
- From 2026-07-03-podcast-forward-guidance-the-ai-trade-is-finally-cracking-weekly-roundup: hyperscalers booked "30% of their income earlier this year… from the markup of the AI labs private valuations on their balance sheets. So it's very reflexive in that if the receivers trade unwinds, that's also not good for the payers." Puts a share (~30% of hyperscaler income) on the mark-up boost.
- michael-cembalest in 2026-07-10-podcast-the-compound-and-friends-the-real-ticking-time-bomb-with-michael-cembalest (JPMAM): hyperscaler "free cash flow projections are heading very low" and funding has shifted from internal cash to borrowing/equity (Alphabet selling stock after a decade of buybacks) — the operating-quality deterioration underneath the flattered earnings. Cross-link mega-issuance-peak-to-ai-capex-derate.
- michael-batnick in 2026-07-07-podcast-the-compound-and-friends-leveraged-etf-casino-apple-breaks-out-sentiment: the semiconductor industry's "aggregate forward profit margin rose to a record 50%" while investors "are paying a forward PE of only 18.4" — the market's visible skepticism of the record margins is itself the check this thesis expects.
The chain
Hyperscaler AI capex + private-AI round mark-ups → S&P earnings and margins inflated by depreciation asymmetry and non-cash write-ups → market valued on inflated E looks reasonable → if capex slows or rounds flatten, both effects reverse and margins roll over. Feeds the down-leg overlay: mega-issuance-peak-to-ai-capex-derate (the earnings-quality channel alongside its issuance/sentiment channel).
Tradeable implications
- Not a standalone position — a valuation-overlay / risk flag on index exposure and the AI-capex cluster, same expression as mega-issuance-peak-to-ai-capex-derate (trim/hedge the highest-multiple AI-capex names; discount "market isn't expensive on forward E" arguments).
- Watchable: whether Q2 stake-gain contribution lands at the predicted 2–3x of Q1's 12%; whether net margins exceed 14.8% again; depreciation-schedule disclosures at the hyperscalers.
Corroboration — the "earnings bubble" framing goes mainstream, and the mechanism is named on air (2026-07-17 ingest)
This concept's thesis — that AI-cycle accounting effects are inflating S&P earnings — is now being discussed as the bear case, which is both confirmation and a crowding warning.
- The framing named — josh-brown in 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up: "The new meme going around now is that it actually it's an earnings bubble. They can't say it's a stock bubble because the biggest, most visible growth stocks in the market have shrinking multiples. This year in a bubble, it goes the other way. Yes, we're paying more for earnings in a bubble. So what the bears have pivoted to is, oh, no, no, we're not saying the stock's in a bubble now. We're actually saying profitability is. And maybe they'll be right. I don't want to mock people just for having that opinion because it sort of does look unsustainable when you look at earnings growth through this prism."
- The mechanism, stated and endorsed — josh-brown in 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up: "that's where the earnings bubble comes from. The companies are over earning relative to what happens when this capex normalizes." jonathan-thomas: "That's exactly right, yeah." Brown: "Which is a very reasonable thing to be worried about."
- ★ The circular-financing leg, described in full — josh-brown in 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up: "I start a NEO cloud... Jensen Wang calls and says... we will financially backstop those purchases... and therefore Nvidia is going to book that chip sale to me and I'm either using money that indirectly comes from them or I'm backstopping that to raise money from someone else. Take that example. Times a thousand and everybody's lending and, or equity investing in everyone else's projects. And the concern is when the music stops, all these people go for the chairs and a lot of people are gonna be left without a chair because some of those funding commitments are gonna vanish as though they were a vapor... I'm saying this is the thing that will make this actually an earnings bubble. Cause these earnings will go away." jonathan-thomas agrees: "No, I think that's exactly right." This is the revenue-quality half of this concept — earnings booked against funding that may not exist — distinct from the depreciation-asymmetry half already on the page. See nvidia-gpu-backstop-to-neocloud-financeability.
- The hyperscaler version of the same circularity — from 2026-07-17-podcast-forward-guidance-the-ai-unwind-is-forcing-a-historic-market, a co-host (diarized "Felix"; source-attributed): the hyperscalers "are receiving income and balance sheet bolstering from their stakes in these privately marked up labs. OpenAI Microsoft anthropic for Google. So it's very circular... you kind of have to chop off one arm to save the other." This is the GAAP-write-up mechanism this concept already names, seen from the decision-maker's seat: cutting capex marks down your own balance sheet.
- Sean Russo's counter-datum on the forecasts — sean-russo in 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up: "generally, outside of recessions, analysts nail it. They nail earnings. Unless it's 2001, 2002, 2009, and then 2020. So for the most part, analysts are actually right." An important base rate against the concept: the estimate-revision anomaly this page rests on is unusual, but analysts are usually right, and they are wrong mainly in recessions — i.e. the estimates break when the cycle breaks, they don't break the cycle. Brown's residual worry is narrower: "The verticality is the thing that bothers me. Of the next year's estimates."
⚠ The dot-com analogue cuts against the concept's timing. jonathan-thomas in 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up on 2001: "that was driven by the Internet... Another capex bubble, a huge capex bubble. And if you remember, the poster child back then for a bubble company was actually pets.com... And really the Internet took a long time to really crystallize its full effect. I'd go as far as to say it wasn't until the pandemic in 2020 where the effect of the Internet that was imagined in 2000 really was fully realized." And his Chewy point — "It's the same exact concept of pets.com, its timing was just better" — is that the idea was right and the timing was wrong. Read against this concept: the earnings may well be inflated and the underlying build may still be vindicated, twenty years later. Inflated earnings are not the same as a wrong thesis, and this concept should not be read as one.
The escape hatch this concept doesn't currently model — josh-brown in 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up: "is there a handoff where we don't have to automatically just have a bear market?... That is the handoff where the S&P493 start to outearn." If adopters deliver productivity as capex moderates, the earnings don't deflate — they relocate. See ai-creators-to-adopters-rotation.
Contradictions / tensions
- michael-batnick in 2026-06-30-podcast-the-compound-and-friends-steroid-era-with-spencer-jakab-the-f-s-going-on: "I'm not really concerned about the one time adjustments for the stocks… The market's not dumb. Okay. It's not putting a multiple on top of that. The same way that companies don't get credit for currency moves one way or the other." (S&P's operating series also strips the stake gains back out.)
- josh-brown's two "get out of jail free cards" (same source): GPUs may actually last (Grace Blackwell chips still running in a decade would validate the slow depreciation), and AI-driven efficiency may make ~15% margins the new normal rather than a mean-reverting anomaly. Jakab's rebuttal: "It's different this time" is the perennial top-of-cycle claim (Irving Fisher's "permanently high plateau").
- Adam Parker (read by michael-batnick, same source): "it is already factually very different this time… This is not just the greatest profit cycle in DRAM history, but it might be the greatest in the history of all cyclicals" — the cycle's profits are real even if a downturn eventually comes.
- The core structural risk both hosts endorse is coordinated capex pullback, not the accounting per se — michael-batnick: "if for whatever reason they all decide together to pull back on spending, well, then that will pull back" (josh-brown: "That'll crush everything in this space… That is the meta risk").