Title: Mark to Myth Goes Public — Private Equity Invented the Playbook. Hyperscalers Now Run It. Show: Paulo Macro (Substack) — paid Guest: Paulo Macro ("Cloudbear") Date: 2026-MAY-15 URL: https://paulomacro.substack.com/p/mark-to-myth-goes-public Length: written post (no timestamps) Note: Back-filled post (processed 2026-JUL-07). Byline reads MAY 15, 2026. Paulo (self-described "Doomer Boomer," tech "not my primary lane") dissects the tech/AI rally since quarter-end and names "the Great Circularity" — multiple layers of circular capital flow propping up the AI complex. Layer 1 (revenue circularity): NVDA sells GPUs to neoclouds (CoreWeave) and AI labs (OpenAI/Anthropic) funded by NVDA equity stakes or hyperscaler-anchored rounds; the game has moved beyond NVDA into the broader semis/memory (Micron's rally so extreme it briefly replaced Berkshire Hathaway as the largest constituent in the Russell 1000 Value). Layer 2 (the new one — via the EARNINGS line): hyperscalers (GOOG, AMZN, MSFT, ORCL, META) invest in private AI companies, mark up those stakes as funding valuations rise, book the gains as "Other Income," and book cloud commitments from the same companies — Alphabet reported $38bn and Amazon ~$16bn of Other Income in 1Q ($53bn combined, ~60% of their combined net income; $49bn from private-stake revaluations), while Anthropic's valuation ran $183bn (Sep-25) → $380bn (1Q26) → a mooted $900bn round to fund $200bn/$100bn compute commitments to Google/Amazon. Paulo's frame: this is the private-equity "mark-to-myth" overmarking playbook, now run through public hyperscalers ("Hold my beer") — a ponzi/flywheel of inflow→higher marks→inflow that is inherently unstable and reverses once OpenAI/Anthropic go public (the "one true mark") or any hyperscaler blinks in the Prisoner's Dilemma. Cites Kevin Muir (earnings-revision surge), Rosenberg Research ("The Curious Case of 'E' in P/E"), Morgan Stanley ("AI: Off Balance Sheet, On the Hook" — >$1.3tn of commitments, META ~1.7x / ORCL >7x fwd operating cash flow, most obligations off balance sheet, contracts underpinning asset-backed private-credit deals), Compound Bros (Micron "fully rational…like IPOs"), Charles Mackay, and Jeremy Grantham's ping-pong-ball top metaphor. Body reproduced for personal study; Substack chrome removed, wording otherwise verbatim.
It is no secret that tech is not my primary lane. I am no Tech Bro. My buddy Le Shrub calls us "Doomer Boomers," and the moniker fits for good reason. Still, if the idea of macro hinges on a certain mental flexibility to drift in and out of lanes while learning just enough to be dangerous, it behooves me to spend a little time on recent developments in broader equities and the nature of the tech-oriented rally since quarter end. Since I don't need to worry about tracking error or benchmarking as someone who trades only my own money, I can certainly afford to have fun staying poor while people max out their token use and jam into the latest new, new thing.
Nevertheless, I would be lying if I didn't feel a regret for not having participated in multi-bagger largecap chip stocks in a month. Upon recent introspection, I find it's not actually the FOMO that disheartens me, but rather that I had this, and was not flexible, agile, and focused enough to do anything about it. I am not saying "I nailed this." I did not nail this. There is no "right call but wrong expression" in this. There's making money, and there's not making money. Making money and being right are always the same thing, and I would rather be lucky and make a ton of money than be smart and sit out a historic move.
So why do I say "I had this?" In early February, I noted that my intense and rather lonely bearishness since late last year came with a catch (Intensely Concerned for Risk — With a Catch). The relevant section toward the end is worth rereading:
But there is a second, more frustrating lesson — that Broadening Out is a very late-stage bull narrative which appears throughout history before one last jam higher by largecaps into the ultimate top. I'm going to zoom in on 1999-2000 to show you what I mean:
Notice how the Nasdaq 100 underperformed the Russell in January 2000 (yellow circle), then stabilized, and went on one last run just to twist everyone's mind inside out — including Druckenmiller who famously bought hours from the high in late March.
This is not unusual. A similar phenomenon occurred during the Nifty Fifty echo-bubble of the early 1970s before the terrible 1973-74 bear market (the analog that led me to create the Rollover Syndrome). To review, this was the S&P 500 over the whole period. The Go-Go Growth mania of the 1960s gave way to a nasty bear market in 1968-70, followed by an epic bull market from May 1970 to January 1973 known as the Nifty Fifty:
Data is harder to come by, but using IBM as a proxy for the Nifty Fifty and comparing it to the NYSE Composite in lieu of the Russell 2000, we can zoom in on the 1971-74 experience into the 1Q 1973 top and the initial bear market:
Notice the 1971-72 run in Nifty… but there was a definite "Broadening Out" moment in 4Q72 as the smaller names did some catching up (see the rip in the orange line). Then there was a final rotation back into the Nifty Fifty just as the smallcaps rolled over.
Again, the first lesson also applies. If you look at the entire bear market into the end of 1974, even though smallcaps outperformed, you had to wait for the bear to run its course (largecaps continued their long slide vs small caps well into 2H75 after the bear market had ended)… relative performance was great, but in absolute terms you still lost money because small caps were not cheap:
This is why the call here is so difficult. On the one hand, we have all the signs we need that retail and institutional investors remain fully committed to risk despite a broader market making little-to-no progress for months, yet we have not seen a proper rupture to shake the tree and reestablish the Wall of Worry the way we did with Liberation Day, the 2022 bear, Covid, or 2018 Volmaggeddon and its Rolling Blowouts, etc.
On the other hand, Broadening Out can fail just as it did in 2021 or 2000, but with the large caps taking one last run at new highs.
As Forrest Gump said at his mom's grave — maybe it's both?
Perhaps we will see a shaking of the tree here in the coming days and weeks, but then one more rally into the spring just to get everyone to wear Druck's 2000 shoes for a couple of weeks [get sucked in], before its lights out for everybody?
I had this. I fricking had this. And I didn't listen to myself as retail jammed into puts in the last week of March and it looked like we were playing out Anatomy of a Crash. I simply couldn't imagine a world where we would actually be doing a Narrowing In rally in the teeth of an energy crisis worse than 1973. Of course, we live in the Upside Down where you have to hold opposing thoughts in your mind simultaneously and believe 6 impossible things before breakfast.
Oh well.
I could spend a lot of time talking about positioning here, but I think readers probably have a good idea of how vol control mechanical funds have more to go before we get back to max long, how CTAs have reversed course in historic fashion, how call buying has been off the charts resulting in historic positioning swings among dealers and their gamma, etc etc. I will skip all that and hone in on what is ultimately serving as justification for the narrative driving the current Narrowing In, because there is something at play here that is really bothering me.
The market makes the news, not the other way around.
My friend Kevin Muir wrote an excellent note Wednesday night on the sudden explosion in upward earnings revisions and analyst expectations right around quarter end. He has done such excellent work and has saved me so much time here that I will just encourage you to head over there and read it directly. He has graciously provided me a link here for my subscribers to get past his paywall. Worth reading that before continuing on here.
Kevin has laid out perfectly what triggered the fundamental justification to drive market prices via upward earnings revisions primarily of chips and AI-related stocks. His conclusion (emphasis mine):
The bulk of the earnings surprise results from energy and AI beats. Although that very well might continue, it's now much more fully baked into the market. I am sticking with my cautious stance. I understand that it's tough to be underweight (or short) when the market is screaming to the upside, but we haven't even talked about the problem with the quality of those earnings. At the very least, I feel better with my negative stance knowing that a lot of optimism is now embedded in the forward earnings curve.
As I read his note it occurred to me that actually the entire upward earnings revisions driving investor expectations in semis/AI is based on multiple layers of circularity that depend on capital flow. This flow can end rather abruptly if investors sense something is amiss and decide their euphoric expectations are misplaced.
Multiple Layers of Circularity
By now most of us are familiar with the revenue circularity at play around the world's largest stock Nvidia, and have seen a version of this chart whereby NVDA sells GPUs to neoclouds like Coreweave or AI labs like OpenAI/Anthropic who were funded either by NVDA's equity stakes to pay for the chips, or more recently the massive funding rounds anchored by hyperscalers among the Mag7:
As investors focused on massive hyperscaler revenue growth, we also know that hyperscaler free cashflow has been sacrificed to pay for datacenter buildouts, many of which are seeing significant delays in addition to skyrocketing construction and energy cost blowouts as Ed Zitron has openly questioned.
Of course, the game has moved beyond NVDA into the broader semiconductor space. My friend Le Shrub wrote Memory vs Tungsten earlier this week about how to value these cyclicals, and featured this chart from Coatue (which ironically they think is a bullish point) but tells you how inextricably linked hyperscaler capex is to chipmaker free cashflow. The revenue circularity has moved beyond NVDA:
The rally in Micron has been so extreme that as I highlighted yesterday, it briefly replaced Berkshire Hathaway as the largest constituent in the Russell 1000 Value index:
But revenue circularity apparently is not enough. We now have a new shooter on the field in what I am calling the Great Circularity™ in US tech today (yes, a play on tech bros' fondness for "Singularity"), and that is via the earnings line. This point matters because even though investors have been growing aware of how hyperscalers (GOOG, AMZN, MSFT, ORCL, META) are torching revenue growth and free cashflow on capex for chips and datacenters, the bullish narrative has noticeably pivoted to earnings growth. In fact, you can hear the concern about heavy capex starting to grow among mainstream commentators like the Compound Bros (see the 34:30min mark here and listen to a few minutes if you can take it — the "I'm open to the idea that maybe it's different this time" at the 39:45 mark is special). But then rewind the same podcast to the 19:30 mark for their exchanges on Micron:
"The buying is rational…their earnings per share quadrupled. It would be weirder if they weren't parabolic because the earnings outlook just literally went up fourfold. It would be weirder if you didn't have a wild reaction to the upside, and you had it. Sometimes these speculative manias come out of nowhere and are based on nothing. This is not that. This is something different. No, this is fully rational. You had an insane change to the fundamental outlook for some very large important companies to that index, and their stock prices — I would argue 99% of the people trading these stocks over the last month have never traded these stocks before in their entire lives. So it breaks anything you thought you know about the long-term average multiple to earnings for these stocks. Like, throw all of that out. It's an entirely new world. It's almost like they are IPOs in the eyes of the people who are trading them.
Think about that for a second. These commentators are saying "throw out everything you think you know" about Micron, but twenty minutes later are wondering if maybe heavy capex by their hyperscaler customers could be a problem.
To this end, Mehmet Beceren and Zackary Young of Rosenberg Research had an excellent note The Curious Case of "E" in P/E on Tuesday May 15th where they very clearly highlighted the problem of earnings quality. Their summary:
Big Tech is investing in high growth private AI companies, marking up those stakes as funding valuations rise, reporting the gains as "other income," and then booking cloud commitments from the same companies, and analysts keep upgrading profit forecasts on the back of headlines.
The FT had an excellent writeup last week that is a must read. The problem is as you delve into the numbers, you will see that Alphabet (GOOG) booked $38bn of Other Income in 1Q, while Amazon reported nearly $16 billion of Other Income. Their combined $53bln in Other Income was nearly 60% of their combined net income, and 34% of the Big 5 hyperscalers' bottom lines. Of that $53 billion, $49 billion was driven by equity stake revaluations higher in private companies. Some charts from the article:
The problem is that these private investment gains are also the customers of the hyperscalers booking the gains (Alphabet is the largest investor in Anthropic, while Amazon is also one of its largest backers):
Now just think: Anthropic's valuation jumped from $183 billion in September 2025 to $380 billion in 1Q26, and yesterday the WSJ reported that they are looking at a $30bn round at a $900 billion valuation! This capital raise, of course, is to pay for enormous cloud-computing commitments to the hyperscalers ($200 billion commitment to Google, $100 billion commitment to Amazon).
Does this seem familiar?
Remember back in the 2022-23 bear market when skeptics (myself included) began to rail about how private equity was grossly overmarking their stakes in private companies, and the proof was in a lack of realizations because Public is the exit for Private?
Hyperscalers to Private Equity: Hold my beer.
The circularity problem is not just that hyperscalers (and Nvidia) are funding their customers through a revenue circularity by funding their sales, but have now gone a step further through the income statement to earnings. GOOG and AMZN invest in Anthropic and OpenAI at successively higher funding rounds seemingly each quarter. The AI players use that capital to book revenue commitments with GOOG and AMZN, so GOOG/AMZN's revenue lines grow. At the same time GOOG/AMZN book earnings growth as the stakes in these private AIs rerack higher from GOOG/AMZN's own investments. GOOG and AMZN's backlog goes up, the AI growth story appears to grow, hyperscalers book "earnings," and the stock market hones in on these largecap techs and rips the stocks higher.
The growth narrative and rising stock prices then propel the next round of private funding, revaluation, and compute commitments, and the AI boom self-validates on beliefs in ever-higher valuations which are marked by the players themselves.
The 'tell' that this Great Circularity hinges on a flywheel of tape-painting flow is that hyperscaler cashflows are lagging miserably as their balance sheets deteriorate and these companies take on debt…and that is the key. Last month Morgan Stanley wrote an excellent note called AI: Off Balance Sheet, On the Hook that really brought home to me the scale of the deterioration. A few key points and charts.
>$1.3tn of commitments ($640bn purchases, $675bn leases) by hyperscalers & Nvidia provide revenue visibility that supports supplier & developer financing.
Contractual commitments span chips, compute, power, AI lab investments, new data center leases, lease guarantees, and more.
Commitments have surged relative to cash flow, with META at ~1.7x fwd operating cash flow and ORCL >7x, reflecting larger volumes and longer contract terms.
Most obligations remain off balance sheet; accounting rules often delay liability recognition until delivery, lease commencement, or a payment becoming probable.
Contracts often underpin infrastructure financing, including asset-backed private credit deals, so they may be difficult to renegotiate if expectations change.
Take another hard look at the first chart. Lease and purchase commitments for GOOG, META, AMZN, and MSFT have run to over 100% of operating cashflow looking out the next twelve months — not free cashflow, but operating! — while in the case of Oracle it's over 700%.
There's a word for this revenue and income feedback loop, one which my friend Kuppy and I used often to describe PE's mismarking game in their flywheel of investor inflow→higher marks→investor inflow:
We called them ponzis. I know that's a loaded, inflammatory word, but what else would you call this? A pyramid scheme maybe? But the thing about ponzis is that they are inherently unstable. If they are not expanding, they are collapsing.
What would cause a reversal? Quite simply, the requirement for reference marks that cannot be set by the players themselves. Which means that once Anthropic and OpenAI become public entities where the stock price at quarter end is the "one true mark," those companies:
A) Better start generating cash as a self-sustaining enterprise; and
B) Never have a serious drawdown in the stock market, because the perpetual motion machine for recent hyperscaler earnings growth reverses via declining earnings through markdowns of stakes in OpenAI and Anthropic.
In the meantime, the hyperscalers are locked in a Prisoner's Dilemma. If any one of them has second thoughts about funding the next up-round and backs away, the music stops on Other Income growth.
Then when you think about how big the weighting of these players are in the US indices, you realize the scale of the problem. Big Tech earnings are doing most of the heavy lifting in the S&P (as Kevin Muir outlined), validating the view espoused by Compound Bros that "the buying is rational"…because 'earnings.' Yet the earnings everyone is celebrating are increasingly a re-marking exercise in the supplier/vendor's valuation.
What happens if investors begin to question not just the revenue circularity but also the bottom line Other Income circularity in all this? This quote from Charles MacKay's Extraordinary Popular Delusions and the Madness of Crowds comes to mind:
"Men, it has been well said, think in herds; it will be seen that they go mad in herds, while they only recover their senses slowly, and one by one." Charles Mackay, 1841
Because investors only recover their senses one by one, this will reflect in the overall market as Jeremy Grantham outlined in this interview during the SPAC/Gamestop mania of February 2021:
"The market tops out when the last bull has put his last money in. There is a moment of maximum enthusiasm, and the next day there's plenty of enthusiasm but less than the previous day. So the buying pressure is released a little bit, like the famous water jets under the ping-pong balls. You turn the faucet down a little bit and the ball is still way up in the air, but it's just dropped a couple of inches. It's that process of slowly lowering the pressure, and the overpriced ping-pong ball slowly descends until it hits the proper level."
—Jeremy Grantham, BusinessInsider, 2021
As I say, I'm not a Tech Bro, and tech is not my primary lane. I know just enough to begin scratching the surface — maybe not even that much.
But I do know that there is something very familiar — and unsettling — about all this…
I hope everyone enjoys a great, relaxing weekend.
As always, kind regards,
Paulo aka Cloudbear