Based on Josh Kushner’s letter to limited partners, as reported by Bloomberg.
Josh Kushner’s LP letter, reported by Bloomberg, shows a $516M fund marked above $3.7B — a genuinely remarkable number governed by one word: paper.
What Happened
According to Bloomberg, which obtained Thrive Capital’s letter to limited partners, Josh Kushner’s $516 million early-stage fund from 2022 — which backed OpenAI, SpaceX, Cursor’s parent Anysphere, Anduril, Ramp, Wiz, and Databricks — is now marked at more than $3.7 billion as of June 30, 2026. That is more than seven times net of fees, a level Bloomberg describes as far above what even top-decile funds of the same vintage have achieved. The firm’s self-reported figures also put total assets under management above $65 billion, with more than half attributable to investment gains, and a net internal rate of return since inception of approximately 33 percent.
Those numbers are striking. They are also, in their current form, marks — not money. Net TVPI counts unrealized paper gains on still-private companies alongside any actual cash returned, and Thrive’s recent early-stage funds have distributed relatively little to their investors so far. The firm is not entirely illiquid: its LP letter reports returning more than $1 billion to investors over the past year, partly through a secondary sale of a growth-fund stake. But that figure is modest set against multibillion-dollar paper marks, and the assets-under-management, gains, and IRR figures are Thrive’s own self-reported numbers from the letter — not audited third-party data.
Notably, Kushner used the same letter to warn investors against AI euphoria in Silicon Valley — a posture that, read as stated, reflects a discipline about valuation risk even as his own fund sits on seven-times paper marks on AI-native companies. The letter is the source for all performance figures cited here.
The key insight: The governing word in every AI-venture returns headline right now is paper. A mark is a hypothesis about what a company is worth; a distribution is a fact about what an investor received. TVPI and DPI are measuring different things, and in the current venture cycle they are being conflated at scale. Thrive’s 2022 fund is the clearest single illustration of that gap — and Kushner’s own warning about euphoria sits inside the same letter that contains the seven-times figure.
The Structural Read
The distinction between TVPI and DPI — between a mark and a distribution — is not a footnote to Thrive’s letter; it is the entire story. A seven-times mark that has returned relatively little in cash to the people who put the money in is a bet that the marks hold all the way to realization. And the events that would convert that bet into actual money are precisely the ones the entire AI venture industry is now lining up to execute: public offerings and secondary sales.
Thrive’s portfolio is effectively a map of that unlock. SpaceX went public earlier this year, providing the fund’s first material realized-path position. The Anysphere acquisition by SpaceX, detailed in our earlier analysis of the SpaceX-Cursor close, converted Thrive’s Cursor stake into public SpaceX shares — but those shares have fallen meaningfully since the June 30 mark and are partly locked up. That means the fund’s single largest realized-path holding is already carried above where it could be sold today, a detail that illustrates precisely how private and thinly-traded marks work: each up-round lifts every prior investor’s carrying value, which makes the asset class look stronger, which makes the next round easier — until the public tape disagrees.
The remaining unlocks — Anthropic’s possible fall offering (explored in our Anthropic Q2 enterprise and IPO crossover analysis), and OpenAI itself, whose scale we’ve mapped in the $40B run-rate and cost-gap breakdown — are the events on which the durability of these marks depends most. OpenAI alone would be the largest single liquidity event in the fund’s life. The concentration matters: a handful of names — OpenAI and SpaceX chief among them — are doing the structural work here. That is how venture power laws operate, but it also means the mark is only as durable as a few outcomes.
This dynamic connects directly to the risk framework laid out in The First AI Financial Meltdown: the self-reinforcing quality of private-market marks is a feature of the system on the way up and a vulnerability on the way down. The paper gains across the AI venture ecosystem are real in the sense that sophisticated investors keep pricing these companies higher in successive rounds. They are not yet real in the sense that matters most to limited partners: cash back in hand, at scale. And as we noted in our analysis of Thrive’s positioning as an AI owner rather than vendor, the firm’s strategic posture is built on that concentration — which makes the IPO window not just a financial event but a structural test of the entire thesis.
FDE Framework — The Founder’s Hedge
“Kushner used his LP letter to chide Silicon Valley for AI euphoria — even as his 2022 fund sits on seven-times paper marks on AI-native companies. Read charitably, that is discipline: a signal that the firm understands the gap between marks and cash and is managing toward it. Read structurally, it is also positioning: a fund that has distributed relatively little to date benefits from being seen as the sober voice in a room full of euphoria. Both readings can be true simultaneously, and neither tells you whether the marks survive contact with a public market.”
Three Implications
IMPLICATION 1 — THE IPO WINDOW IS A LIQUIDITY REFERENDUM
The next 12 months of AI-company public offerings — Anthropic, potentially OpenAI, and others — are not just corporate milestones. They are the mechanism by which the entire private-market mark structure gets stress-tested. If public valuations come in below the last private-round marks, TVPI numbers across the industry compress rapidly. If they hold or exceed them, the paper becomes cash. Thrive’s 2022 fund has more riding on that window than almost any other vehicle of its vintage.
IMPLICATION 2 — STALE MARKS ON PUBLIC POSITIONS ARE AN UNDERAPPRECIATED RISK
SpaceX is now public, but Thrive’s SpaceX position — including shares received through the Anysphere acquisition — is partly locked up and has already fallen since the June 30 valuation date. That means the fund’s most advanced realized-path holding is carried above its current tradeable price. This is a structural feature of how private-fund marks work at the moment of transition to public markets, and it applies to every fund that will receive IPO shares subject to lockup agreements over the coming year. The mark-to-market process does not end when a company goes public; it accelerates.
IMPLICATION 3 — SECONDARIES ARE THE UNDERRATED MECHANISM
Thrive returned more than $1 billion to LPs over the past year partly through a secondary sale of a growth-fund stake — not an IPO. As explored in the context of power-law concentration in AI, secondary markets are becoming a parallel liquidity infrastructure for the venture industry, allowing partial realizations without requiring a public offering. For funds with large paper marks and patient LPs, secondaries extend the runway. But they also cap the upside and introduce their own pricing discipline — secondary buyers discount for illiquidity and concentration risk, which can itself function as a reality check on private-round marks.
