The Largest IPO on Record Priced — and SpaceX Spent It the Next Morning: an ~$86bn Listing, a $60bn All-Stock Grab for Cursor, and Dell Terms Out $3bn

The Lead

Issue 12 ended by putting SpaceX in the lead — not as a rocket story but as an AI-compute one — and noting that the pre-IPO tripwire existed “to catch exactly this.” This week it caught the event itself.

On 15 June, SpaceX completed its IPO: 638,888,888 Class A shares at $135.00, including the underwriters’ full overallotment — an offering of roughly $86bn, which makes it, by a wide margin, the largest IPO on record (Saudi Aramco’s ~$29bn in 2019 was the prior mark). It dual-listed on Nasdaq and the new Nasdaq Texas, with a genuinely global book spanning the US, Australia, Canada, the EEA, Japan, Switzerland and the UK. Read the use-of-proceeds language and the framing is unmistakable: net proceeds fund “the Company’s growth strategy, including the expansion of the Company’s AI compute infrastructure,” and only then launch vehicles, Starlink and general purposes. The AI-compute business the S-1 described — the xAI merger, the COLOSSUS capacity, the ~$1.25bn/month Anthropic compute agreement — is now a public company, and it told the market that the first dollar of its record raise goes to compute.

Then it did something more telling than the listing. The next morning, 16 June, SpaceX agreed to buy Cursor — Anysphere — for an implied equity value of $60.0bn, all in stock, with Cursor’s holders taking SpaceX Class A shares priced off a seven-day VWAP and the deal slated to close in Q3. The compute layer reached straight up the stack and bought the application layer, using equity that had been public for less than 24 hours. It also quietly dissolved the ~$10bn Cursor-related termination liability the S-1 had buried — by acquiring the whole company instead. And on 17 June it added Roelof Botha of Sequoia to the board and audit committee. Three filings, three days: price the biggest IPO in history, spend $60bn of it on the AI-coding layer, and seat a marquee investor. This is the compute-as-an-asset economy that Issues 11–12 traced through Broadcom’s Apollo backstop and Amazon’s delayed-draw — now arriving not as a financing footnote but as the largest equity event the market has ever cleared, immediately recycled into vertical integration.

A note on process, in the spirit of keeping the score honestly: the system’s automated extraction was down for part of this window — an upstream billing error (a 402 from the inference provider) silently failed the run on every filing after 12 June. These three SpaceX filings, the Dell notes and the Credo 10-K were therefore read and extracted by hand for this issue rather than by the nightly pipeline. The tripwire flagged the filings; the analysis was reconstructed directly from the primary documents. The gap is being closed.

Threads

SpaceX prices the largest IPO on record — and spends it the next morning on a $60bn all-stock grab for Cursor

The mechanics matter because they encode the thesis. An ~$86bn raise at $135.00 is not a capital-raising of the ordinary kind; it is a liquidity event for the single largest private holder of AI-compute ambition, and the proceeds statement puts compute infrastructure first in the queue. The global, multi-jurisdiction book (Australia and Japan retail included) tells you the demand was sized to the deal rather than the other way round.

The Cursor acquisition is the part to sit with. SpaceX — which through the xAI merger owns COLOSSUS-class training capacity and an external compute-sales business — is paying $60bn of stock for the AI coding-tool company. Vertically, that is the compute owner buying a high-value consumer of inference, internalising an application that both showcases and absorbs its own capacity. Strategically, it is the clearest statement yet that the winners of the buildout intend to own the stack from silicon-adjacent compute through to the software that runs on it. Financially, paying in freshly listed equity priced on a seven-day VWAP is the most expensive currency a company can spend only if you believe the stock is cheap — which is exactly the bet a newly public compute monopoly would make. The $60bn tag, set one day after a record IPO, is itself a data point about where private AI-application valuations have settled.

Issue 12’s forecast [2026-06-13-001] — that the off-balance-sheet financing of compute would proliferate — is not confirmed by any of this, and discipline requires saying so plainly: an all-stock acquisition and an investment-grade bond deal are vanilla balance-sheet financings, not the vendor-backstop / compute-lease-guarantee / SPV structures that call specified. It remains open, resolving in August. What this week does confirm is the adjacent claim the system has carried since Issue 8 — that the AI buildout’s centre of gravity is consolidating into a handful of names that can self-finance at scale. SpaceX just became the largest of them on its first day.

The credit thread compounds — Dell terms out $3bn of senior notes days after the working-capital balloon

Last issue’s Dell 10-Q was the cleanest tension on the board: +757% AI-server revenue alongside a $20.8bn purchase-obligation balloon ($17.3bn due inside twelve months) and a hiring freeze. This week Dell did the predictable thing — it went to the bond market. On 16 June it priced $3.0bn of senior notes in three tranches: $1.0bn at 4.750% due 2031, $0.75bn at 5.000% due 2034, and $1.25bn at 5.250% due 2037, guaranteed up the Denali/Dell Inc. structure and rated for investment grade.

This is not distress — Dell is funding at tight spreads — but it is confirmation of mechanism. A company growing the top line 88% does not raise term debt because it is short of demand; it raises it because the working capital behind GPU-and-HBM purchase commitments has to be funded somewhere, and the income statement won’t carry it alone. The Issue 11 instruction was “follow the debt.” Across three issues the debt has moved from the hyperscalers’ structured facilities (Amazon’s delayed-draw) to the merchant-silicon backstops (Broadcom–Apollo) to, now, the box-builder’s plain-vanilla senior notes. The credit story isn’t a single dramatic structure; it’s the steady migration of the buildout’s funding onto borrowed money, one filing at a time.

The engine versus the fundamentals — Credo prints +198% and the model still flags it bearish, beside SMCI’s dilutive raise

Credo’s 10-K is this week’s microcosm. Fiscal 2026 revenue of $1.3bn against $436.8m the year prior — roughly +198%, a near-triple — with net income of $472.3m (from $52.2m) and R&D nearly doubled to $279.4m. On the fundamentals, a connectivity winner riding the AI-bandwidth build (active electrical cables, SerDes, retimers, the PILOT diagnostics layer) about as cleanly as exists.

And yet the convergence engine flags CRDO bearish (score 53), and it is not alone: DELL (739), CRWV (315), STX (117), ANET, ALAB and MRVL all read bearish this week, against just TSM (bullish, 283), AVGO (136) and CIEN at the top. Why fade a name compounding at 198%? The 10-K answers its own question: top-10 customers are ≈90% of revenue, with two customers each above 10%; average selling prices “generally decrease over time”; it is fabless on TSMC; and its named competitors are Broadcom, Marvell and Astera — three of which the engine also reads bearish. This is the “great isn’t enough” regime from Issue 11, now operating within the AI-hardware complex rather than across it: the model is separating the foundry/merchant-silicon node it can underwrite (TSM, AVGO) from the crowded, concentrated, ASP-pressured second-derivative names (CRDO, ALAB, MRVL, ANET) that the market already owns at full price.

Super Micro is the same tell at the box-build layer. Its $1.25bn at-the-market programme plus a 45.5m-share underwritten offering and a concurrent preferred is material dilution from a company that, if it had the margin and the balance sheet, would not be issuing equity into its own demand. Strong revenue, thin economics, capital raised to fund the working capital — the pattern that runs from Dell to SMCI to Credo and explains why the engine is bearish on the names whose income statements look the most exciting.

Since Issue 12 (13 June)

• Forecast [2026-06-13-001] — OPEN, tracking. The call (≥1 further structured compute-financing arrangement within 60 days) resolves ~12 August. This week’s Dell notes and SpaceX stock deal are financings, but not of the structured/off-balance-sheet kind the forecast specified, so they do not count toward it. No confirm claimed.
• The pre-IPO watch resolved. Issue 12 carried SpaceX as a tripwire catch on the S-1; the IPO closed this week at $135.00. The thread the system flagged early became the largest equity event on record — the value of being early and on the right name.

Watchlist Updates

• SPCX (SpaceX) moves from pre-IPO sidebar to a live, listed name — but a private-style one: it has declared it will disclose materially through its IR page and X account rather than the wires, so the filing tape (8-Ks, eventually 10-Qs) becomes the cleanest signal source. Worth promoting to active coverage; the Cursor close (Q3) is the next catalyst.
• DELL remains the sharpest demand/margin divergence on the board (convergence bearish, 739) — now with $3bn of fresh term debt against the $17.3bn of purchase obligations due inside a year. The cash-conversion print next quarter is the thing to watch.
• CRDO / ALAB / MRVL are the engine’s clearest intra-complex shorts-against-the-foundry expression and the spine of this issue’s forecast. TSM and AVGO remain the bullish counterweights — the demand that shows up as Credo growth and Broadcom RPO ultimately lands at the foundry.

The Week Ahead

The reporting lull genuinely ends next week. Micron (MU) reports ~24–25 June — the single most important read in the complex: whether the HBM/DRAM cost pressure threaded through every name above is a sector-wide squeeze or a margin gift to the memory makers. Western Digital (22 Jun), FedEx (~23–24 Jun) and Accenture (18 Jun) bracket the storage and macro reads, and CEG’s merger vote and lock-up (30 Jun) sits in the power-for-compute corner. The catalyst worth pre-positioning, though, is the structural one this issue pre-registers:

Forecast [2026-06-18-001]: over the next 60 days, an equal-weight basket of the convergence-bearish AI-hardware names (DELL, CRWV, CRDO, ALAB, MRVL, ANET) underperforms the bullish merchant/foundry pair (AVGO, TSM) on total return. Confidence 0.58, horizon 60 days, price-graded. If it holds, the engine’s intra-complex discrimination is a real, tradeable signal rather than a coincidence of a single week’s flags. If the crowded names outrun the foundry, the “great isn’t enough” read has decayed and the model is fighting the tape.

Methodology and Disclosures

Threads are written from the SEC-filing database the system maintains; this issue covers the 11–18 June window (≈17 filings, the most material of which were extracted by hand after an inference-provider billing outage silently failed the automated run on filings after 12 June — a gap now being closed). All figures — the SpaceX share count and price, the $60bn Cursor consideration, the Dell tranches and coupons, and the Credo fiscal-2026 revenue, net income and customer concentration — are taken directly from the primary 8-Ks and 10-K filed this week and should be verified against those documents before any trading decision. Forecasts are pre-registered, dated and falsifiable, and graded in the following issue. The system also maintains a machine-graded calls ledger (relationship-flip, divergence and radar calls — 42 currently open, awaiting their resolution windows) running the same falsify-and-grade loop beneath the editorial layer. Nothing here is investment advice.