VCX's NAV Is $21.70, Not $18.97: The June 30 Schedule of Investments, Position by Position

The correction first, because it is ours

Most premium figures published about VCX since June — on this site and nearly everywhere else — have been divided by $18.97, the NAV in the fund's Form N-CSR for the period ended March 31, 2026.

That figure was correct. It is now a quarter out of date, and as of today there is a newer one straight from the fund: $21.70 at June 30, 2026.

The two percentage figures matter more than they look. Twenty-six points of premium is the difference between "this trades at more than twice what it owns" and "this trades at under twice what it owns." People make decisions on that sentence.

It also kills a lazier reading in both directions. The premium did not compress because the market repriced the fund; it compressed because the fund's own assets were marked up 14.4% in a quarter while the share price moved around. And the NAV rise is not a trick of share issuance: net assets rose 14.4% while the share count did not change (35,797,138 at March 31 and at September 8). This is portfolio appreciation, full stop.

Our VCX review and lockup analysis are being updated to this figure today. If you read either of them before August 28, 2026, every premium percentage in them was computed on the old denominator.

What the fund actually owns, at cost and at mark

This is the part that no summary of VCX contains, because it requires reading a 10-page exhibit rather than a fact sheet.

Read down the multiple column. Outside artificial intelligence, only financial technology (2.00x, mostly Ramp) is marked well above cost. AI carries about 90% of the fund's unrealised gain ($282,319,000 of $314,361,000, our arithmetic), and inside that sleeve it is a handful of lines.

38% of the gain is in one position

The single largest holding in VCX is not a company. It is SaxeCap Advisors VIII, LP — an LP interest acquired on December 6, 2023 for $8,534,000 and carried at $128,169,000 — a vehicle whose underlying company the June exhibit does not name (the fund's March 31 annual report lists a position with the same cost and acquisition date as an Anthropic co-investment vehicle), which is the thread we pull in the composition read of the same portfolio, where access vehicles turn out to be 56.93% of the fund, or 16.5% of the fund's net assets.

That is a 15.0x mark. And because the fund's total unrealised gain is $314,361,000, that one line contributes $119,635,000 — 38.1% of every dollar of appreciation in the entire portfolio.

The top three vehicles carry 65.5% of the fund's entire unrealised gain, and the top five vehicle positions are 44.3% of net assets.

Four of the five are flagged in the filing as Level 3 — valued using significant unobservable inputs; the fifth, AI Growth GRT, is valued at its underlying vehicle's reported NAV as a practical expedient. All five are restricted and non-income-producing. None of their legal names tells you what is inside them.

The fund is admirably direct about that last point. In its own words, "the legal name of an investment vehicle may not identify the underlying company to which the Fund has economic exposure," which is why it publishes a separate table by underlying economic issuer:

A boundary worth stating plainly: the June exhibit maps only one vehicle to its company: its practical-expedient table describes AI Growth GRT, LLC ($72,480,000) as a vehicle investing in the securities of Databricks, Inc. The fund's March 31 annual report goes further and lists positions by underlying company: the lines with the same cost and acquisition date as SaxeCap VIII ($8,534,000, 12/06/23) and Quiet OA Access ($25,890,000, 09/27/24) appear there as Anthropic and OpenAI co-investment vehicles. What the filings establish is that more than a fifth of this fund is economic exposure to a single private company, Anthropic, and that the fund's largest gain sits in a vehicle whose cost and date match an Anthropic position.

The vintage tell

Sort the AI sleeve by purchase date and a pattern appears that no fact sheet shows.

Most positions bought in 2026 are marked at or below cost. Three of the four 2026 lines in this table are marked below it, and Anduril — a company whose valuation has been widely reported as rising — is carried at 0.71x what the fund paid for that particular seed-series stake seven months ago. The exception is large: AI Access 12, LLC, bought February 10, 2026 for $20,800,000, is marked at $45,459,000 (2.19x).

There are innocent explanations, and they are probably most of the story: recent purchases have had no subsequent priced round to mark against, so cost is the honest carrying value; a seed-series instrument is not the same security as the headline round; and structural terms differ between SPV interests. But the pattern is worth holding onto for a simple reason. The fund's marks are overwhelmingly driven by 2023-vintage entries, and 2023 is when this fund was buying private AI exposure before the repricing. Whatever you think VCX is worth, the number is a bet on a small number of 2023 tickets, not on the breadth of the book.

The same discipline applies to the rest of the portfolio. Gaming/Entertainment (Epic Games, $19,180,000) and Biotechnology (Loyal Animal Health, $10,000,000) are each carried at exactly 1.00x cost, and Property Technology nets out at cost only because Inspectify at 1.25x offsets Rhino Labs at 0.51x. Data Infrastructure is at 1.30x, and most of that is one Vanta position bought in 2022.

The part of VCX that is not a venture fund

Buried under the equity book is something almost nobody mentions when they describe VCX: 8.5% of net assets — $65,925,000 — is commercial mortgage-backed securities, and every single tranche is a data centre.

And it is levered. The filing discloses $30,400,000 of reverse repurchase agreements with Barclays Bank PLC, all settled May 29, 2026 at borrowing rates of 4.66% to 4.81%, against $30,532,000 payable including accrued interest — with a footnote confirming that all or part of the QTS tranches have been pledged as collateral for those repos.

That is a real-estate credit carry trade, run inside a fund marketed as access to private technology companies: borrow from Barclays at ~4.7%, hold floating-rate data-centre paper at 5.98%–7.38%, pocket the spread. It is a perfectly conventional trade and the spread is genuinely positive. Two things about it belong in a reader's file anyway. First, it is leverage, and leverage cuts both ways against a NAV that already carries 82.2% restricted securities. Second, the repos carried a maturity date of August 26, 2026 — two days before this filing was published — so what replaced them, and at what rate, is not in this document.

For a site about real estate, this is a nice irony worth naming: the most conventionally "real estate" thing in Fundrise's flagship technology fund is a levered book of data-centre mortgage bonds, and it is the only segment marked below cost.

What this does and does not change

It changes the denominator, and only the denominator. At $39.84 you are paying $1.84 for every dollar of assets the fund says it has, not $2.10. That is materially better than the number in circulation and it is still a large premium for a closed-end fund.

It does not change the composition of that dollar. Of the $21.70 of NAV per share, roughly $17.84 is restricted securities (82.2%) and about $1.17 is money-market fund (5.4%). The assets are marked by the adviser using unobservable inputs, quarterly, on vehicles with no public price. That is not an accusation — it is what a Level 3 private-markets fund is, disclosed exactly as the rules require. It does mean the premium question and the mark question are two separate questions, and only one of them gets a market quote every day.

It does not settle the lockup argument either way. The supply from the accelerated August 13, 2026 expiry has already arrived and the price went up through it, which we wrote up in what happened after the unlock. Nothing in this filing speaks to that.

What it does do is give anyone holding VCX a current, filed, per-position picture of what they own — and the honest summary of that picture is: a fund whose value is 66.5% artificial intelligence, whose gain is 65.5% in three unnamed vehicles, whose 2026 purchases are mostly marked at or below cost (AI Access 12, at 2.19x, is the exception), and whose newest NAV is $21.70.

The short version

VCX's NAV is $21.70, filed today, for June 30, 2026. Almost everything you will read about the fund still divides by $18.97, and that overstated the premium by about 26 percentage points at the August 28 close (about 22 at the September 11 close, our arithmetic).

Underneath that number, the fund is more concentrated than the "Anthropic, OpenAI, SpaceX, Databricks" summary suggests — not just in companies but in marks, with 38% of all its appreciation in one LP interest and 65.5% in three. Most positions it bought in 2026 are carried at or below what it paid; the one big exception, AI Access 12, is at 2.19x. And 8.5% of it is a levered data-centre bond book that has nothing to do with venture capital at all.

None of that makes the fund good or bad. It makes it legible, which for a vehicle trading at 1.54x its own stated assets at the September 11, 2026 close (our arithmetic) is the thing that was missing.