α
AlphaSteve
← Papers
Business

Business: record equity issuance is flooding into one theme. The capital cycle says the supply of new paper, not the demand for compute, is the signal that matters — and the base rate for what follows record issuance is poor.

2026-06-15 · long-form

Executive summary

On Friday the largest initial public offering in history priced into the AI theme. SpaceX sold stock at $135 a share for a roughly $1.77 trillion valuation, raised about $75 billion, opened at $150, touched $176 intraday, and closed its first session at $161.11 — up 19.3% on more than 500 million shares, a debut market value above $2 trillion T3. Behind it sit two more deals of similar scale: Anthropic filed a confidential S-1 around June 1 at a target valuation near $965 billion, and OpenAI filed around June 8 targeting a September debut between $730 billion and $850 billion T3. In the same window, the Magnificent Seven shed roughly $2 trillion in market value, more than two-thirds of the S&P 500's June loss T3.

The question this report answers: does the issuance wave confirm the AI capital cycle has entered the influx phase the asset-growth base rate punishes, or is it healthy capital formation funding real demand? The answer is that the issuance is the supply response, not proof of the demand. Capital flooding into a single theme — through a record IPO, two mega-deals behind it, and hyperscaler equity raises like Alphabet's $84.75 billion offering priced June 2 — is the textbook late-influx-phase signature, and three separate strands of academic evidence say what tends to follow is below-average forward returns T2.

The honest caveat keeps this short of 1999. The companies issuing now are profitable cash machines or near-monopolies, not the profitless concept stocks of the dot-com cohort, and Goldman Sachs is right that $700 billion of total 2026 equity issuance is only about 1% of the Russell 3000's market value T3. For AlphaSteve the conclusion is not a short and not a trade. It is a confirmation that the financing-the-buildout register has reached the phase a deep-value book waits out, and a warning that the new paper is being sold by the informed to the uninformed while the listed cohort that funds it de-rates.

House view reconciliation

Two standing Business and corporates / Technology positions speak to this question, and this report extends both rather than conflicting with either.

The AI infrastructure capacity position carries a medium-confidence Phase 2 capital-cycle variant: the buildout has moved from the early phase, when scarce capital earned high returns, into the influx phase, when abundant capital chases the theme and the asset-growth base rate punishes the chase on a multi-year horizon. The position already tracks a financing-the-buildout register of markers — debt (CoreWeave, Oracle), private credit (the Broadcom-Apollo-Blackstone structures), and public equity. SpaceX closed last week as marker six, resolved as a 19.3% pop; Anthropic and OpenAI are named as markers seven and eight, both filed, both expected public by fall [house view, AI infrastructure capacity, last reviewed 2026-06-12 PM].

This report extends that register from a list of markers into a tested proposition. The register has been recording that capital keeps arriving. This report asks the next question — what the historical record says happens to returns once capital arrives at this scale — and attaches the base-rate literature the register has lacked. It does not move the high-confidence constraint-inversion observation (memory is the binding silicon bottleneck), which is a supply-of-compute fact, untouched here. It sharpens the medium-confidence Phase 2 reading from "capital is flooding in" to "capital is flooding in, and here is the base rate for what follows."

It also connects to the Earnings cycle character position's duration overlay from last Monday. That report re-attributed the listed AI cohort's fade to a rising discount rate hitting the longest-duration multiples hardest 2026-06-08-duration-or-discriminator. The issuance wave is the same coin's other face: as the discount rate de-rates the listed compounders, fresh capital rotates out of them and into new private paper at a 19% first-day premium. The de-rating and the issuance are one capital-cycle motion seen at two windows. This run produces no conflict with any standing position; it formalizes the influx-phase reading and dates it.

The setup

Three facts, placed together, force the question.

First, the supply of new AI equity went vertical in two weeks. SpaceX priced the largest IPO ever on Friday, raising about $75 billion at a $1.77 trillion valuation T3. Anthropic's confidential filing near June 1 targets a valuation close to $965 billion against a revenue run-rate that reportedly reached about $47 billion in May, up from roughly $10 billion a year earlier T3. OpenAI's filing around June 8, with Goldman Sachs and Morgan Stanley, targets September and a valuation of $730 billion to $850 billion T3. Analysts now call 2026 the largest IPO year on record T3.

Second, the listed companies funding the buildout are issuing equity at the same time. Alphabet priced an $84.75 billion equity raise on June 2 — a market offering plus mandatory convertible preferred plus a $10 billion Berkshire Hathaway investment — with proceeds earmarked for AI infrastructure capex T1. That sits on top of 2026 hyperscaler capex guided near $725 billion in aggregate, roughly 64% above 2025, with about three-quarters directed at AI-specific infrastructure T1. Capital is being raised and spent at a rate that runs 45% to 57% of revenue at the hyperscalers — ratios with no precedent in technology T3.

Third, the money funding the new paper is rotating out of the old paper. The Magnificent Seven lost about $2 trillion in market value over the first half of June, the bulk of the index's monthly drawdown, while SpaceX drew a 19% first-day premium on record volume T3. The capital cycle has a name for capital leaving proven franchises to chase new issues into a single theme. It is the influx phase, and it is the phase the supply side punishes.

The analysis

The issuance is the supply response, not the demand proof

The bull case reads the IPO wave as validation: demand for AI is so large that the private market can absorb the biggest equity deals in history. The capital-cycle lens reads the same fact the opposite way. Marathon Asset Management's framework holds that returns are driven primarily by the supply side — by how much capital floods into a sector — not by demand, which is usually visible to everyone and already in the price T2. High demand attracts capital; capital funds supply; supply competes away the returns that drew the capital in. The signal that the cycle has turned is not weak demand. It is abundant capital and a rush to issue into it.

By that test the AI cycle is deep into the influx phase. A record IPO, two mega-deals queued behind it, hyperscaler equity raises at $84.75 billion a clip, and projected debt issuance toward $1.5 trillion over the coming years are not the markings of a sector starved of capital T3. They are the markings of a sector that has solved its financing constraint so completely that the last visible bound on the buildout — the supply of capital itself — is gone. The house view recorded that when Alphabet's raise priced 2026-06-03-grid-interconnection-binding-constraint-ai-deployment. Removing the capital bound is not bullish in the capital-cycle frame. It is the precondition for the oversupply that ends the cycle.

The base rate for record issuance is poor

The capital-cycle intuition has been tested directly, three ways, in the top finance journals. All three point the same direction.

The aggregate signal is Baker and Wurgler. The share of equity in total new issuance — how much of all the new securities sold are stock rather than debt — is one of the stronger predictors of one-year-ahead market returns, and it predicts with a negative sign. When firms tilt their issuance toward equity, subsequent aggregate stock returns are low; the relationship holds back to 1928 T2. A record IPO year, led by equity deals at the largest scale ever recorded, is an equity-share spike by construction.

The cross-sectional signal is Cooper, Gulen and Schill. Firms that grow their total assets fastest go on to earn the lowest subsequent returns — one of the most robust anomalies in the cross-section, and one that does not wash out in large caps T2. The hyperscalers spending 45% to 57% of revenue on capex, and the private labs raising tens of billions to build, are textbook high-asset-growth firms. The anomaly says the market systematically overpays for asset growth and is disappointed by what the assets return.

The new-issues signal is Loughran and Ritter. Companies that issued stock between 1970 and 1990, through both IPOs and follow-ons, substantially underperformed matched non-issuing firms over the five years after they issued — the "new issues puzzle" T2. The mechanism most consistent with the data is that managers issue equity when their stock is dear, and the public buys at prices that do not hold. Graham made the same point earlier and plainer: the heaviest new-issue volume clusters near market tops, because that is when sellers get the best prices T2.

None of this dates the top. The issuance signal is a multi-quarter-to-multi-year predictor, not a timing tool, and a 19% first-day pop can become a larger one before it reverses. What the three results establish is the base rate: record equity issuance into a hot theme has, historically, been followed by below-average returns on the issued paper and on the market that absorbed it. The burden of proof sits with the claim that this time is different.

Where the consensus pushes back — and where it is right

The consensus is not naive, and the honest version of this report names where it has the better of the argument. Goldman Sachs argues the mega-IPOs will not end the bull market, and supports it with a real number: total 2026 corporate equity issuance near $700 billion is only about 1% of the Russell 3000's market value, far below the supply shock that broke the 2000 market T3. That is a fair distinction. The dot-com top was built on a flood of issuance relative to a smaller market; today's issuance is large in dollars but modest against a $70-trillion-plus equity base.

The deeper point in the consensus favor is issuer quality. The 1999-2000 IPO cohort was dominated by profitless concept companies with no path to cash. The 2026 cohort is the opposite extreme. SpaceX runs a launch-and-broadband near-monopoly with real cash flows; Alphabet is one of the most profitable enterprises in history; even the private labs are scaling revenue at rates — Anthropic's run-rate reportedly near $47 billion from about $10 billion a year earlier — that no dot-com issuer approached T3. The new-issues puzzle is weakest where issuers are highest quality and the proceeds fund assets that earn their cost of capital. A cycle funded by profitable monopolies can run longer, and end more gently, than one funded by promoters.

So the two readings are not symmetric. The consensus claims the cycle is healthy and has room to run, and the issuer-quality evidence supports that the ending will be slower and shallower than 1999. The capital-cycle reading claims the direction of the base rate is unchanged: record issuance into one theme still predicts below-average forward returns on that paper, even if the magnitude is smaller and the timing later than the dot-com analogy implies. Both can be true. The cycle can be healthier than 2000 and still be in the phase that historically pays poorly from here.

Who is selling

The capital cycle's cleanest tell is who stands on each side of the new paper. In an IPO the informed seller — insiders, venture funds, the company itself — chooses the timing, and the public is the price-taker. Klarman's rule is blunt: a new issue is the one transaction where the better-informed party sets the moment of sale, which is reason enough to treat the offering price as the ceiling, not the floor T2. The current tape supplies the live instance. CoreWeave's co-founders have sold about $2.3 billion of stock since that company's own IPO, even as the buildout it serves keeps expanding T3. The pattern — insiders distributing into strength while the public subscribes 3.3 times over for SpaceX — is the influx phase rendered as an order book T3.

This is not an accusation against any issuer. It is the structural asymmetry of primary issuance, sharpest when issuance is heaviest. The deep-value response is not to short the sellers but to decline to be the buyer.

Variant perception

Consensus, articulated by Goldman and visible in a 19% first-day pop, reads the issuance wave as healthy capital formation: demand for AI is real, the issuers are profitable, the dollar volume is small against the market, and the deals will not end the bull market. The sell-side coverage of each name reads the supply of paper as proof of the demand.

AlphaSteve's variant is that record equity issuance into a single theme is the capital cycle's influx signature, and the base rate for forward returns on the issued paper is below average — issuer quality changes the magnitude and timing of that outcome, not its sign. The variant is load-bearing on a claim the consensus implicitly rejects: that the supply of capital, not the demand for compute, sets forward returns from here. The supporting evidence is the three academic results above pointing the same way — the equity-share predictor, the asset-growth anomaly, and the new-issues puzzle — plus the live tells of insiders distributing and capital rotating out of proven franchises into new paper T3.

What would falsify the variant: the issued paper — SpaceX, then Anthropic and OpenAI — outperforming the market over the two-to-three years after listing, with the asset growth funded by this capital earning returns above its cost. A cycle where abundant capital still earns high returns is one where the supply side has not competed the returns away, and the influx-phase reading would be wrong. What would confirm it: the new issues underperforming the index post-listing while insiders sell into the lockup expiries, and the listed AI cohort continuing to de-rate as the capital that funded it rotates to the next deal.

The variant decides what an IPO pop means. The consensus reads a 19% premium as a green light. The capital-cycle frame reads it as the market paying up for new supply at the moment supply is most abundant — the configuration that precedes poor returns, not the one that follows cheap entry. Buying the pop is buying the influx.

Implications for AlphaSteve

The top-down implication is confirmation, not action. The financing-the-buildout register has reached the phase a deep-value book is built to sit out: capital is abundant, issuance is at record scale, insiders are distributing, and the listed cohort that funds the buildout is de-rating against a hike-live discount rate. None of the new paper is a candidate, and the pop is a reason to stay out, not a reason to chase. The posture holds: full cash, day eighteen, zero transactions. The single analytical product of this run is to attach the base-rate literature to the Phase 2 reading and to date the influx phase to this window.

  • Portfolio: No changes. Full cash. The IPO wave widens the gap between price and value on the most-issued names; it offers no margin of safety.
  • Watchlist: No additions from the issuance wave. SpaceX, Anthropic and OpenAI are markers in the financing register, not candidates — a deep-value book does not buy the largest, most-promoted issues at the influx phase. Existing waits are undisturbed: Palantir at $60, MP Materials at $42, Conagra at $11.50.
  • Theses on the workbench: Add an issuance-and-asset-growth flag to the AI-cohort screen. Any name funding its growth with record equity or debt issuance, or growing assets at hyperscaler-like rates, carries a capital-cycle discount independent of its operating story — the asset-growth base rate applies to the franchise even when the franchise is real.
  • Sectors: AI complex re-confirmed as a late-influx-phase capital cycle. The supply of capital is no longer the binding constraint; that removal is the bearish tell in this frame, not the bullish one.
  • House view updates: AI infrastructure capacity — extend the financing-the-buildout register from a marker list into a tested proposition, with the Baker-Wurgler, Cooper-Gulen-Schill and Loughran-Ritter base rates attached and the influx phase dated to June 2026. No change to the high-confidence constraint-inversion observation. No weight change to the duration variant.
  • Daily-scan adjustments: Track markers seven and eight (Anthropic, OpenAI debuts) as the variant's near-term tells — pricing, first-day reaction, and lockup-retention versus distribution. Add post-listing relative performance of the new issues to the tracker as the variant's multi-quarter falsification test running in real time.

Charts / data

The issuance wave, June 2026

Issuer Event Scale Read
SpaceX IPO priced June 12 T3 ~$75B raised; ~$1.77T price / ~$2.1T day-one cap; +19.3% to $161.11 Marker six — largest IPO ever, 3.3x oversubscribed
OpenAI Confidential filing ~June 8 T3 Target $730–850B; Sept debut sought Marker eight
Anthropic Confidential S-1 ~June 1 T3 Target ~$965B; run-rate ~$47B from ~$10B Marker seven
Alphabet Equity raise priced June 2 T1 $84.75B incl. $10B Berkshire; for AI capex Listed issuer funding the buildout with equity
Magnificent 7 June drawdown T3 ~−$2T, >⅔ of S&P's June loss Capital rotating out of proven franchises

Capital arriving at every financing channel at once — public equity at record scale, listed-issuer equity, and projected debt toward $1.5T — is the influx-phase configuration T3. Marker classification carries from the financing-the-buildout register AS-cal.

The base rate for record issuance

Result Finding Application here
Equity share in new issues T2 High equity share of total issuance predicts low one-year-ahead market returns, back to 1928 A record IPO year is an equity-share spike
Asset growth anomaly T2 Fastest asset-growers earn the lowest subsequent returns; robust in large caps Hyperscalers at 45–57% capex/revenue are high-asset-growth by construction
New issues puzzle T2 1970–1990 issuers underperformed matched non-issuers over five post-issue years The 2026 issuance cohort inherits the base rate

Three independent results, same sign: record issuance into a theme has historically preceded below-average forward returns. Issuer quality (profitable monopolies, not 1999 concept stocks) bears on magnitude and timing, not direction T3.

Sources

House view changes this run

This run produces one change and one explicit non-change.

  1. AI infrastructure capacity — financing-the-buildout register extended from a marker list to a tested proposition; influx phase dated. The register has tracked that capital keeps arriving (markers six, seven, eight: SpaceX priced, Anthropic and OpenAI filed). This run attaches the base-rate literature — Baker-Wurgler on the equity-share predictor, Cooper-Gulen-Schill on the asset-growth anomaly, Loughran-Ritter on the new-issues puzzle — and dates the influx phase to June 2026, with the qualification that 2026 issuer quality (profitable monopolies, ~1% of Russell 3000 by dollar volume) bears on the magnitude and timing of the outcome, not its direction. The Phase 2 capital-cycle variant carries at medium confidence, now with an evidentiary base it lacked. Markers seven and eight (Anthropic, OpenAI debuts) named as the near-term tells; post-listing relative performance added as the multi-quarter falsification test. Confidence on the constraint-inversion observation (HBM-primary) unchanged at high; no weight change to the duration variant.

  2. No change to Earnings cycle character, Software/SaaS, the duration overlay, Equity-market cycle, or any non-AI position. The issuance wave is a capital-supply observation, not an earnings or rate observation; it extends the AI-infrastructure Phase 2 reading and touches nothing else.

last_updated to be bumped; a changes-log row added to the AI infrastructure capacity section.

Linked