how the largest IPO in history exposed where the AI-and-space physical-economy trade actually still sits, and where it's already too late

On June 12, SpaceX priced the largest IPO in history. It opened at $150, closed near $161,1 and hung a $2.1 trillion valuation on a company2 that eighteen months earlier most ordinary investors couldn't buy at all.3 And the rest of the space sector fell.

Rocket Lab dropped. AST SpaceMobile dropped harder. Intuitive Machines, Planet Labs, Redwire, all down on the day the thing they orbit around finally became buyable.4 The IPO that was supposed to lift every boat in the harbor pulled the water out from under them instead.

That inversion is the whole story, and it travels further than space.

For two years the smart version of the AI trade has been one sentence: don't buy the model, buy what the model physically needs. Don't buy the rocket, buy what reaches orbit. Power, grid, fabrication, the defense-industrial base, the unglamorous physical layer underneath the narrative. It was a good trade. It was right. And by the summer of 2026 it's almost entirely consensus, not only at the headline but one full layer down, in exactly the names that felt clever to own a year ago.

I went looking for the mispricing in that second layer across five fronts: AI power, the vehicles that hold private SpaceX, European space sovereignty, the satellite "balance of plant," and India. I wrote down what I expected to find before I checked a single number. Most of what I expected was already gone.

the categorization gap already closed

Start with the cleanest case. The thesis on Korean grid makers (Hyundai Electric, Hyosung, LS Electric) was that they're stealth AI-power plays priced as sleepy industrials: the number-one transformer supplier into the US grid, once priced at a steep discount to Western peers like GE Vernova. By mid-2026 that discount is gone. Hyundai Electric trades in the high-20s on forward earnings, right on top of Eaton.5 LS Electric trades at a premium to GE Vernova.6 Hyosung more than doubled in a year.7 The discount didn't narrow. It inverted. The market re-categorized the whole sector while the note was still being written.

India was worse. The pitch, an under-covered defense and space supply chain with no Western analyst coverage, is half true and completely useless. MTAR Technologies trades at roughly 250x earnings on flat three-year profit.8 Paras Defence has one or two analysts covering it and a P/E of 116.9 The coverage gap is real. The price gap is not. Indian retail did the re-rating that institutions normally do, and did it harder.

Even the genuinely obscure names weren't cheap. Frequency Electronics built the timing oscillator for a NASA deep-space mission,10 is followed by maybe two analysts, and sits on the first hundred-million-dollar backlog in its history. It trades at 72x EV/EBITDA. Gilat, in the same business and followed by roughly no one, trades in the mid-teens.11 Under-covered turned out to mean overlooked, not cheap. Those are different words, and the market only pays you for one of them.

This is the lesson that survived when the trades didn't: a coverage gap is not a price gap. The absence of analysts tells you who isn't watching. It tells you nothing about what's already in the price.

the one door that stayed locked

There was a single mispricing in the whole survey that held, and it held because it was structural rather than informational, the kind no amount of attention can arbitrage away.

For most of the past year, the only way an ordinary investor could own SpaceX was through a closed-end fund called Destiny Tech100, ticker DXYZ.12 At its May peak it traded more than 50% above the value of what it actually held.13 People paid that premium to own a fund whose single biggest holding was SpaceX.14

A cheaper door stood open beside it the entire time. Scottish Mortgage, a Baillie Gifford investment trust run for the better part of a century, charges about a third of a percent in fees.15 It held an even larger SpaceX position, close to a fifth of the portfolio,16 and traded at a discount to its net asset value.17 More SpaceX, a fraction of the fee, no premium. The crowd queued at the expensive entrance while the cheaper one stood unmarked next to it.

The gap didn't last because nobody noticed. Sophisticated closed-end-fund buyers knew. It lasted because DXYZ has no creation-and-redemption mechanism, so there's no way to manufacture shares and trade the premium away.18 The only thing that closes a gap like that is the underlying becoming buyable directly, which is what happened on June 12. DXYZ is now down about 64% from its peak.19

That's the shape of a mispricing worth your time: locked, not hidden. The information was free. The structure was the moat.

what the IPO actually did

Which brings us back to the inversion at the top. The intuitive read on a marquee listing is that it floods the sector with attention and capital and lifts the comparable names. It does the opposite. CFRA published a sell rating on SpaceX itself within minutes of the debut, and the sell-side's attention stayed on SpaceX, not its rivals.20 A marquee listing doesn't pull the comps up with it. The capital rotates out of them and into the new name.

What the listing did was sort the sector. The names that compete with Starlink, AST SpaceMobile and the low-orbit constellation plays, absorbed the rotation and were left trading down roughly half from their highs.21 Kratos, a defense supplier that competes with no one in that fight, was treated as a beneficiary rather than a casualty.22 The signal had nothing to do with whether space went up or down. It drew a line between who SpaceX eats and who SpaceX feeds.

the screen underneath

Strip the specifics away and the same four questions sit under all of it. Can an outsider actually buy this and get the money back out, is it accessible? Is the demand contracted and executing, or just announced, is it proven? Is it cheap because the informed buyer is genuinely absent, or because the risk is priced correctly, is it really mispriced? And is there a catalyst that closes the gap on a timeframe you can hold?

The third question is where almost everything dies. Finding something that looks cheap and unfollowed is easy. Proving the cheapness is inattention rather than a discount the market is applying on purpose, for governance, for cyclicality, for a controlling shareholder who'll never let a minority win, is the hard part. OHB, the German space group, trades at a discount because the founding family owns more than 60% and has twice shown it would rather keep the discount than share control.23 That's a correctly-priced cage, not a mispricing.

So where does the trade live, now that most of it is owned?

In the two places the crowd structurally can't or won't go. The access-premium unwind is one: wherever a locked structure separates the price of a thing from the value of what it holds, with a catalyst coming to break the lock. The layer beneath the obvious layer is the other. In AI power, the megacap turbine and grid names are fully priced; the genuine inattention has retreated upstream, into grain-oriented electrical steel, transformer bushings, on-load tap-changers, the components whose shortage Wood Mackenzie expects to persist toward 2030.24 In European space, the one name that still works is Avio, and not because it's cheapest. Its float is mechanically opening as Leonardo sells down,25 so attention is arriving rather than already arrived.

I spent fifteen years in markets learning the same thing in different costumes. The positions that worked were almost never about whether a technology was good. They were about a structural gap between what was available and what was priced, held until the gap closed. The boring industrials I back now are the same wager in private markets: not the cleverness, but the part of the market everyone else has quietly agreed not to look at.

The physical-economy trade was the right idea. It just stopped being an edge the moment it became the thing everyone knew. By the time a thesis is clean enough to fit in a headline, the money's already in it. The edge was never the layer. It was being early to the next one down, or finding the door that's locked rather than merely unwatched.

What would change my mind: if DXYZ's premium comes back once the lockups clear and the access-vehicle structure proves stickier than I think, or if the component shortage resolves faster than the forecasters expect and the cyclical names I'm avoiding turn out to have been the trade. Both are testable. I'll be watching both.

References

Live financial figures (stock multiples, prices, fund discounts) are dynamic; each is cited to its source with an access-date of 2026-07-14 and should be refreshed on the publish day.

  1. Roush, Tyler. "SpaceX Opens At $150, Surging 17% After Largest IPO Ever." Forbes, June 12, 2026.
  2. "Initial public offering of SpaceX." Wikipedia (accessed July 14, 2026) — ~$2.1T day-one-close valuation; largest IPO in history; +19% first day.
  3. "How to Invest in Starlink Stock." The Motley Fool — SpaceX was privately held and not directly buyable by retail before the IPO.
  4. "Space stocks dip on SpaceX IPO." Payload, June 15, 2026 (Rocket Lab −10.8%, AST −15.5%, Intuitive Machines −13.1%, Planet −8.8%); and "Space stocks drop as SpaceX soars in debut." Yahoo Finance, June 12, 2026 (adds Redwire −11%+).
  5. Stock multiples: Hyundai Electric (KRX:267260) and Eaton (ETN), stockanalysis.com (accessed July 14, 2026) — both in the high-20s on forward earnings.
  6. LS Electric (KRX:010120) vs GE Vernova (GEV), stockanalysis.com (accessed July 14, 2026) — LS Electric at a premium on trailing earnings.
  7. Hyosung Heavy Industries (KRX:298040), stockanalysis.com (accessed July 14, 2026) — ~1-year return (feeds range 170%–490%; "more than doubled" is the conservative reading).
  8. MTAR Technologies (NSE:MTARTECH), Screener.in (accessed July 14, 2026) — P/E ~200–260 depending on feed; 3-year profit CAGR ≈ −1% (roughly flat).
  9. Paras Defence & Space (NSE:PARAS), Screener.in (accessed July 14, 2026) — P/E ~113–116; thin analyst coverage (~1–2 brokers).
  10. "Frequency Electronics Announces Launch of Ultra-Stable Oscillator as Part of JPL Deep Space Atomic Clock (DSAC) Mission." GlobeNewswire (FEI press release), July 1, 2019.
  11. Frequency Electronics (FEIM) and Gilat (GILT), stockanalysis.com (accessed July 14, 2026) — FEIM ~68–72x EV/EBITDA vs Gilat ~14–15x.
  12. Destiny Tech100 Inc., Form N-CSR (FY2025 annual report), SEC EDGAR, filed March 10, 2026 — a "Closed-End Management Investment Company."
  13. "Destiny Tech100 Stock Jumps as SpaceX IPO Fever Puts Private-Tech Funds Back in Focus." TS2, May 9, 2026 — DXYZ was one of the few retail routes to SpaceX; at its May peak it closed ~$54.60 against a NAV of $19.97 (well over 50% premium).
  14. Destiny Tech100 Inc., Form N-CSR — Schedule of Investments, SEC EDGAR, March 10, 2026 — "SpaceX I LLC" is the fund's largest single position.
  15. Scottish Mortgage Investment Trust — company data, Association of Investment Companies (accessed July 14, 2026) — ongoing charge 0.33%.
  16. "Scottish Mortgage says SpaceX represents 21% of total assets as at June 3, 2026" Reuters (via TradingView), June 2026; corroborated by QuotedData, June 5, 2026.
  17. Scottish Mortgage — company data, Association of Investment Companies (accessed July 14, 2026) — discount to NAV.
  18. Destiny Tech100 Form N-CSR (closed-end management investment company), SEC, March 10, 2026; mechanism explained in "Destiny Tech100 Stock Jumps…" TS2, May 9, 2026 — closed-end shares have no creation/redemption, so a premium can persist.
  19. DXYZ quote, Kraken Stocks (accessed July 14, 2026) — ~$25.78 vs a 52-week high of $72.87 (−64.6%).
  20. "6 Wall Street Analysts Have Issued Price Targets on SpaceX." The Motley Fool, June 18, 2026 — CFRA analyst Keith Snyder assigned SpaceX a sell rating and a $115 target; analyst coverage centered on SPCX itself.
  21. AST SpaceMobile (ASTS) and Intuitive Machines (LUNR), Yahoo Finance (accessed July 14, 2026) — trading roughly half or more below their 52-week highs; IPO-day context in Foreign Policy Journal, June 16, 2026.
  22. "This SpaceX IPO Could Trigger Billions in Spending — 1 Stock to Watch" (Kratos). The Motley Fool, June 11, 2026 — Kratos framed as a picks-and-shovels beneficiary, not a competitor.
  23. OHB SE ad-hoc announcement on the €300/share rights issue, OHB SE Investor Relations, June 2026 (Fuchs family >60%, does not sell); and "Simpson Thacher Advises KKR on €900 Million OHB Re-IPO," July 2, 2026 (2023 KKR stake; 2026 re-IPO at €300 that traded below issue).
  24. "Power transformers and distribution transformers will face supply deficits of 30% and 10% in 2025." Wood Mackenzie, 2025 (constraints "persist well into the 2030s"); trade coverage: "U.S. transformer market faces severe supply constraints," pv magazine USA, May 11, 2026.
  25. "Leonardo: subscription to Avio's capital increase completed." Leonardo press release, November 11, 2025 — Leonardo's Avio stake reduced to 19.30% from 28.75%; free float rising.
◆   ◆   ◆