Mirae Asset Securities is Korea’s largest brokerage by equity capital. It is also, through ten Delaware limited partnerships filed with the SEC, one of the larger foreign holders of what is now Nasdaq-listed SpaceX. Those two facts belong to different disclosure regimes — one Korean, one American — and neither regime shows you the other. Put them side by side and a Korean broker’s quarterly earnings turn out to track a US space company’s share price — and its own board minutes show it went on to underwrite that company’s Korean IPO tranche.
🔑 Key Takeaways
- $499.2M across seven funds, filed with the SEC, all pointing at one man’s companies. Form D filings for Mars I/III/IV, Project X I, and Gaia I/II/III share a promoter (Mirae Asset Ventures II LLC) and a single named executive officer. Korean disclosure names the funds; only the US filings show they are one structure.
- Three tracks became one ticker. SpaceX, Twitter/X and xAI were separate bets made between 2022 and 2026. SpaceX absorbed xAI in an all-stock deal in February 2026 and listed on Nasdaq that June. What looked like diversification across Musk’s empire is now concentration in a single marked-to-market security.
- One consolidated vehicle reported revenue of ₩78M and net income of ₩164.3bn — a ratio of about 2,100×. It is not an operating business. It is a holding structure whose income is the revaluation of what it holds.
- 70.1% of second-quarter net income came from outside the parent company. Consolidated net income attributable to owners was ₩1,895.9bn against parent-only net income of ₩567.2bn.
- Net fair-value gains exceeded operating profit outright. First-quarter net gains on financial assets at fair value were ₩1,553.7bn against operating profit of ₩1,375.0bn — so without them the quarter is an operating loss of about ₩178.7bn. Every figure here is from the filed statement.
- [Estimate] The third quarter carries an unrealised loss of about ₩1.08tn at prices as of 12 August, which would put the quarter at a net loss despite an intact brokerage business. This is our calculation from a disclosed base, not a company figure, and the half-year report can check it.
What Mirae Asset Securities is, and how it got here
Mirae Asset Securities (006800.KS) is a Seoul-headquartered securities firm — brokerage, wealth management, investment banking and principal investment. By equity capital it is the largest securities house in Korea, and its market capitalisation was around ₩19.9tn in early August 2026.
The firm reached that size by merger rather than organic growth alone. Mirae Asset, founded by Park Hyeon-joo as an asset manager, acquired and merged with Daewoo Securities — one of Korea’s oldest brokerages — and operated as Mirae Asset Daewoo before renaming to Mirae Asset Securities. That merger matters to this story for an unglamorous reason: it left the company holding 107,493,380 treasury shares created in the merger, a block whose eventual cancellation or retention is a live question for shareholders and unrelated to anything below.
What distinguishes it from a pure brokerage is the size of its own balance sheet. Disclosed investments in other companies total roughly ₩9.24tn, about 46% of market capitalisation. A brokerage earns fees from other people’s trades. This one also runs a large book of its own.
🔍 The part Korean filings name but do not explain
Korea’s DART system requires a listed company to list its consolidated subsidiaries. Mirae Asset’s list includes entities with names like Project Mars Fund I, LP, Project X Fund I, LP, and Mirae Asset Gaia Fund I. The filings give ownership percentages and summary financials. They do not say what the funds hold.
The US filings do — not the holdings themselves, but the structure. Each of these is a Delaware limited partnership that filed a Form D with the SEC on raising capital. Read together, the seven filings share a promoter and one named executive officer across every fund:
| Fund | First sale | Amount sold | Investors |
|---|---|---|---|
| Project Mars Fund I | 13 Jul 2022 | $107,129,993 | 5 |
| Project X Fund I | 17 Oct 2022 | $56,115,000 | 4 |
| Project Mars Fund III | 12 Jan 2023 | $95,954,443 | 5 |
| Project Mars Fund IV | 1 Jun 2023 | $10,308,000 | 4 |
| Mirae Asset Gaia Fund I | 10 May 2024 | $96,212,259 | 14 |
| Mirae Asset Gaia Fund II | 21 Nov 2024 | $22,264,260 | 8 |
| Mirae Asset Gaia Fund III | 27 Jan 2026 | $111,181,955 | 5 |
| Total | $499,165,910 |
The total is our sum of the seven filings listed. The SEC does not publish a combined figure, and the officer named on these filings also appears on other Mirae Asset funds we have not included here because we could not tie them to a public event.
The dates line up with events. Mars I’s first sale was July 2022, when SpaceX raised at a reported $1bn round. Project X Fund I’s was 17 October 2022 — ten days before the Twitter acquisition closed. Gaia I raised in May 2024, alongside xAI’s Series B; Gaia III in January 2026, alongside a Series E. We are matching dates and amounts to public events, not reading a holdings list. The Form D filings do not name portfolio companies, and we did not find a filing that does.
One detail is worth stating because it answers an obvious question. Gaia I lists 14 investors where the Mars and X funds list four or five. Mirae Asset consolidates Gaia I at 55.50% — so roughly 44.5% of that fund belongs to outside co-investors, and the investor count is consistent with a syndicate rather than a single balance sheet.
📈 Three bets that became one
Between 2022 and 2026 these were separate exposures: a rocket company, a social network, an AI lab. Two corporate events collapsed them.
2022-07 Mars I → SpaceX ββββββββββββββββββββββββββββ 2022-10 Project X → Twitter → merged into xAI (2025) ββ€ 2024-05 Gaia I → xAI βββββββββββββββββββββββββββββββ€→ SpaceX acquires xAI (Feb 2026) 2024-11 Gaia II → xAI βββββββββββββββββββββββββββββββ€ → Nasdaq listing, 12 Jun 2026 2026-01 Gaia III → xAI βββββββββββββββββββββββββββββββ
SpaceX acquired xAI in an all-stock transaction announced in February 2026 and listed on Nasdaq on 12 June 2026. Every one of these positions is now the same listed security, and every one of them is marked to market daily.
This is the sentence an English-language reader is least likely to have encountered: a Korean brokerage’s reported earnings now move with a US-listed share price, and the mechanism is a set of Delaware partnerships disclosed on one side of the Pacific and consolidated on the other.
📊 What the consolidation actually does to the numbers
DART requires summary financials for consolidated subsidiaries. Two lines make the structure legible.
| Consolidated vehicle | Stake | Revenue | Net income |
|---|---|---|---|
| Global Space Investment Association I | 89.57% | ₩78M | ₩164,270M |
| Mirae Asset Gaia Fund I | 55.50% | ₩575,424M | ₩539,768M |
| Project Mars Fund I, LP | 98.50% | ₩203,187M | ₩153,302M |
The first row is the one to sit with. Revenue of ₩78M against net income of ₩164,270M is a ratio of roughly 2,100 times. No operating business produces that. It is what a holding vehicle looks like when the thing it holds is repriced upward: there is almost no revenue because there is almost no operation.
These are principal-investment vehicles the company controls, not client funds. But only the ownership percentage accrues to Mirae Asset’s shareholders — the remainder is non-controlling interest. Gaia I at 55.50% means nearly half of that fund’s result belongs to someone else. That cuts both ways, and it matters below.
💰 The same structure in the quarterly numbers
Second-quarter 2026 net income attributable to owners was ₩1,895.9bn. Parent-company-only net income for the same quarter was ₩567.2bn. The difference, ₩1,328.7bn, is 70.1% of the consolidated figure arriving from outside the parent.
The half-year filing makes the same point more sharply: gross fair-value gains were ₩17.93tn against operating revenue of ₩21.63tn — 82.9% — while operating profit for the half was ₩2.49tn against net fair-value gains of ₩3.19tn. Without them, the half is an operating loss of about ₩0.70tn.
The first quarter shows the same shape in the revenue line, and it is worth separating two numbers that are easy to conflate. Gross gains on financial assets at fair value were ₩10,813.7bn, or 74.9% of the quarter’s ₩14,428.7bn operating revenue. That is a statement about how the revenue line is composed, not about profit — the same assets also produced ₩9,260.0bn of losses.
The profit statement is the net figure. Net fair-value gains were ₩1,553.7bn, against operating profit of ₩1,375.0bn. The net gain is larger than the entire operating profit, which means that without it the quarter is an operating loss of roughly ₩178.7bn.
This is not an argument that the brokerage does not exist. Parent-only net income for the first half was ₩854.3bn, which annualises to roughly ₩1.71tn and puts the operating business on a parent-only P/E near 12×. The business is real. The question is what the market is paying for on top of it — and, this quarter, which direction the top part moves.
📜 What the half-year report added — and what it still does not say
The half-year report filed on 14 August 2026 does not contain a line called “Musk-related holdings.” It does contain the board and risk-committee minutes, and those record something the earlier filings did not.
| Date | Body | Item | Outcome |
|---|---|---|---|
| 29 May 2026 | Risk Management Committee | Approval of firm-commitment underwriting of the SpaceX IPO Korea Tranche; and a project fund contribution for SpaceX IPO investment | Both approved |
| 1 June 2026 | Board of Directors (9th meeting) | Same underwriting approval | Approved, all directors in favour |
| 25 June 2026 | Board of Directors | Report on response measures regarding the SpaceX subscription | Reported, not voted |
This matters because it is a different kind of exposure from the one described above. Everything earlier in this piece concerns positions bought years before the listing. Underwriting an IPO tranche on a firm-commitment basis is a dealer function: the underwriter agrees to take the allocation and is left holding whatever is not placed.
🔲 What we cannot tell you: the size. No amount appears in the minutes, and no separate disclosure of the underwriting commitment was filed that we could find. We also do not know whether any of the tranche went unplaced — that is precisely the number that would matter, and it is not public. The 25 June item is listed only as a report on “response measures,” with no content recorded.
So the honest statement is narrow: the company’s own governance records show it underwrote the Korean tranche of the SpaceX IPO and separately funded a vehicle to invest in it. Anything about how much, or what came back onto the balance sheet, would be invention.
📉 [Estimate] What the third quarter looks like
The mark on these positions is a share price and an exchange rate, both public.
SPCX 30 Jun $164.19 → 12 Aug $133.29 (31 Jul low: $108.40) KRW 30 Jun 1,548.61 → 12 Aug 1,415.58 Position value in won: −25.8%
Applying that to a disclosed-and-derived base gives the following. The base is our build, not a company disclosure: year-end 2025 Musk-related holdings attributable to owners of ₩1.92tn, plus the January 2026 Gaia III commitment, plus first- and second-quarter revaluation gains, giving roughly ₩4.2tn at 30 June.
| Scenario | Move in won | Unrealised loss | Implied Q3 net |
|---|---|---|---|
| Prices hold at 12 Aug | −25.8% | −₩1.08tn | −₩516bn |
| A further −10% | −33.2% | −₩1.39tn | −₩827bn |
| Back to the 31 Jul low | −39.7% | −₩1.67tn | −₩1,100bn |
All three assume parent-only net income holds at the second quarter’s ₩567.2bn. All three are losses.
Three things would move these numbers, and we cannot resolve any of them today. The quarter marks at the 30 September close, not today’s. A lock-up expiry on 12 June’s listing released restricted stock in August, which can press the price independently of the business. And because non-controlling interests take their share of losses as well as gains, the figure attributable to shareholders is smaller than the gross move — by how much depends on the mix across funds we can only partly see.
🧩 What would prove this wrong
- The base is still unverified, and the half-year report did not settle it. We expected the notes to disclose holdings in enough detail to check our ₩4.2tn build. They do not. The filing shows total assets at fair value through profit or loss of ₩89.37tn at 30 June across the whole group — a number far too broad to confirm or refute a ₩4.2tn slice. Our base remains a build, not a disclosure.
- The fund-to-holding link. We matched Form D dates and amounts to public funding events. We did not find, and do not claim, a filing that states what each fund holds. A different mapping would change the concentration argument.
- Disposal. If the company sold into the post-lock-up window, an unrealised mark becomes a realised gain or loss on a different schedule. We have no disclosure either way.
- Hedging. Securities-firm accounting nets fair-value, FX and interest positions that are managed together. Removing one line to show an “operating loss” is a simplification the company itself does not make.
🎯 Why It Matters
The specific numbers here will age. The structural point will not: a company can be transparent in two jurisdictions and still be opaque, because no single reader is required to look at both. Korean filings name the partnerships and consolidate their results. US filings show the partnerships are one structure with one promoter. Neither set is hidden. The gap is in the reading.
For an investor in Korean equities, the practical version is narrower. When a brokerage’s earnings beat comes overwhelmingly from outside the parent company, “record quarterly profit” and “the business had a good quarter” are different statements — and only one of them repeats.
Sources. Company filings via DART (consolidated subsidiary summaries, quarterly and parent-only statements, investments in other companies) and SEC EDGAR Form D filings. Share price and exchange rate as of 12 August 2026. Figures labelled [Estimate] are our calculation from disclosed inputs and are not company guidance.
Disclosure. The author holds a position in Mirae Asset Securities (006800) and reduced it in August 2026. This is analysis of public filings, not investment advice.