SharkWater Trading | August 2026
In May, a private company you cannot buy raised $65 billion at a $965 billion valuation. Nine days later it confidentially filed for an IPO. Its chief rival, valued at $852 billion, filed its own confidential paperwork the same month and is reportedly holding out for a trillion dollar sticker before it lists.
Anthropic and OpenAI are the two largest IPOs in the pipeline, possibly the two largest in history. Neither trades. But the market has already built a shadow order book around both of them, and if you understand how NAV works, you can trade it today.
This post maps the exposure vehicles, explains the premium and discount math that actually drives returns in these wrappers, and lays out equity and options structures for playing the runway into the listings.
The Scoreboard
Quick state of play as of mid August:
Anthropic: $965B post-money after the Series H in May. Revenue run rate crossed $47B. Confidential S-1 filed June 1. The chatter points to a fall 2026 window, which would put it public before OpenAI.
OpenAI: $852B after the record $122B round in March. Confidential filing in June, but reporting suggests a 2027 listing because leadership will not accept a valuation below $1 trillion. Cash burn is the elephant here, reportedly on the order of $27B this year with a bigger number projected next year.
The race matters for trading purposes. Whichever lists first sets the comp for the second, and every vehicle holding either name gets re-marked off that print.
The Exposure Map
Think of this like a water table. The companies are the aquifer. Every vehicle below is a well drilled at a different depth, with different pipe losses along the way.
Tier 1: Listed closed-end funds (tradeable, premium risk)
DXYZ (Destiny Tech100) is the retail favorite. Per its SEC filings, in January it put $100 million into an SPV holding economic exposure to Anthropic Series B preferred shares. Read that sentence twice. Not common stock, not direct shares. Economic exposure, through a special purpose vehicle, to an early preferred series. That structure means layered fees, stale marks between funding rounds, and liquidation preference mechanics that most holders have never modeled.
DXYZ already ran this playbook once. It spiked hard on the SpaceX IPO filing in May, and now that SpaceX is public, the fund is rotating and the Anthropic stake becomes the marquee private asset. Sell-side estimates put projected NAV somewhere in the $33 to $47 range depending on how aggressively you re-mark Anthropic, and the stock has swung from a 52-week low near $20 to a high near $73. Short interest has run in the mid-teens. This is a NAV re-marking story with a squeeze chaser attached.
VCX (Fundrise Innovation Fund) is the more concentrated AI bet. Roughly 20.7% Anthropic, 9.9% OpenAI, about 5% SpaceX, with AI-related names near 44% of the book. It is the only listed wrapper with meaningful exposure to both horses. The catch: after listing it ripped several hundred percent above a stated NAV around $18. When you pay a large premium to NAV, you are not buying Anthropic. You are buying other people’s excitement about Anthropic, and excitement mean-reverts.
Tier 2: Interval funds (buy at NAV, no listing pop)
ARKVX (ARK Venture Fund) holds both names and prices at NAV with quarterly liquidity windows. No premium to pay, no premium to capture. This is the boring well. It is also the only one where your entry price equals the fund’s stated look-through value. For a long-horizon core position, boring is a feature.
Tier 3: Public equity proxies (liquid, options-rich, diluted)
MSFT is the single largest look-through position in either company. Microsoft owns roughly 27% of OpenAI on an as-converted basis, marked around $135B at the October 2025 recap and worth potentially $270B if the IPO prices at $1 trillion. Here is the interesting part: MSFT is down roughly 19% this year, the worst of the Magnificent Seven, while sitting on a stake worth as much as 9% of its own market cap that the market is arguably assigning little credit. That is a stub trade hiding inside a mega cap.
GOOGL and AMZN are the Anthropic side of the same coin. Amazon has $8B invested plus participation in later rounds, and Google has been a major holder since the early days. The stakes are smaller relative to their market caps than Microsoft’s OpenAI position, so the look-through torque is weaker, but the Anthropic S-1 will force a public mark on both.
The NAV Math That Actually Matters
Premium and discount to NAV is the whole game in the Tier 1 wrappers. Three mechanics to internalize:
1. Marks lag reality. A fund’s Anthropic position gets re-marked at funding rounds and audits, not continuously. Anthropic went from $380B in February to $965B in May. Any fund still carrying the position at the old mark has embedded NAV growth that shows up on the next report. That re-mark is a scheduled catalyst, and the market front-runs it.
2. The wrapper’s scarcity value dies at the IPO. This is the trade everyone gets wrong. The premium on DXYZ or VCX exists because they are the only liquid doors into a locked building. The day Anthropic lists, anyone can buy ANTH or whatever the ticker becomes, and the wrapper’s reason to trade above NAV evaporates. SpaceX just demonstrated the pattern: filing announcement pops the wrapper, the actual listing deflates it. The pre-IPO run is a rental, not a marriage.
3. Preferred is not common. SPV exposure to Series B preferred behaves differently from common stock in a down scenario and converts on specific terms in an IPO. In a strong listing this mostly washes out. In a weak one, the waterfall matters.
Equity Strategies
The premium rental (DXYZ, VCX). Long into re-mark and S-1 catalysts, with a hard rule to be flat or short-biased by the time the underlying actually prices its IPO. You are trading the anticipation, not the event. Position sizing should assume 30% drawdowns are normal here, because they are.
The NAV accumulator (ARKVX). Dollar-cost average at NAV through the quarterly windows. No premium risk, no timing genius required. This is the sleep-well allocation for the thesis that both companies are worth owning through the IPOs and beyond.
The stub trade (MSFT). Long MSFT on the argument that a beaten-down 2026 tape is giving you Azure at a discount and the OpenAI stake nearly free. If you want to isolate the stub, pair it: long MSFT against a partial short in QQQ or a basket of the other mega caps strips out the market beta and leaves you with mostly the OpenAI optionality plus the relative valuation gap.
Options Strategies
Liquidity check first. MSFT, GOOGL, and AMZN have deep, tight chains. DXYZ options exist but trade wide with fat spreads, so use limit orders and small size or skip them entirely. VCX has no practical options market. Structure accordingly.
MSFT long-dated call spreads. The cleanest expression of the OpenAI IPO catalyst. Jan 2027 or Jun 2027 call spreads financed partly by the elevated put skew a down 19% stock carries. The spread caps your cost against the real possibility that the IPO slips or prices soft. Buying the spread rather than naked calls matters because IV will inflate into the IPO date and collapse after pricing, and a spread is largely immunized against that vega crush.
The MSFT wheel. Regular readers know this is home turf. Selling cash-secured puts at technical support on a Mag Seven laggard with a $270B lottery ticket attached is about as good as wheel candidates get. If assigned, you own a quality name at a discount and roll into covered calls above cost basis. The IPO catalyst gives the covered call premium a persistent bid.
Calendar spreads into S-1 events. When the confidential filings go public, the amendment and roadshow dates become known events. Front-month IV inflates around those dates while back-month stays calmer. Selling the inflated front expiry against a longer-dated long leg harvests that event premium. Works on MSFT for OpenAI dates and, with wider tolerances, on DXYZ for Anthropic dates.
DXYZ put protection on core longs. If you are running the premium rental into the Anthropic listing, cheap out-of-the-money puts a few months out are the insurance against the wrapper deflation described above. Spreads are ugly, so buy them on green days when the market maker is happy to sell you downside.
What Kills This Trade
IPO slippage is the obvious one. A 2027 OpenAI listing means a full year of theta on anything short-dated. A broad AI multiple compression re-marks everything down at once and the wrappers fall faster than NAV because premiums compress into discounts. And the burn numbers in the actual S-1s could shock a retail base that has only seen the revenue headlines. $47B of run rate is real. So is $27B of cash burn.
The waiting room is crowded and the door has not opened yet. Trade the line, not the room.
Nothing here is financial advice. I trade some of these names and structures. Do your own homework, size like you can be wrong, and read the actual filings before you touch an SPV wrapper.
No comments:
Post a Comment