TheChek News Flow

I imagine you saw the Faber segment on Friday. The FT and Bloomberg have since added the parts that make it interesting.

CNBC reported that Situational Awareness has been buying options again, several hundred million dollars of premium late last week and early this week. The FT reports that JPMorgan has ended its prime brokerage lending relationship with the fund altogether, and that the new positions are flex options. Bloomberg confirms that the fund has established a prime broking relationship with Clear Street.

The lending relationship is the fact worth sitting with, not the options.

Recall where this started. In July the fund's assets fell 67 per cent on cascading collateral demands against a highly levered AI book, and the bulk of the public portfolio went to Citadel in a single block. The prime brokers who worked the positions down at the time were Bank of America, Goldman Sachs and JPMorgan.

Aschenbrenner then told investors he would "fight another day", that the fund would continue as a hybrid public-private vehicle, and that the public book would henceforth be run on a fully paid-for basis. He also told them the fund was up 80 per cent on the year, which is worth holding alongside the 67 per cent asset decline rather than in place of it. The first is a performance figure, the second is a balance sheet figure, and they measure different things.

Six weeks later he is back. Three things follow.

First, the counterparty verdict. JPMorgan did not trim the relationship, it ended the lending. A G-SIB terminating a prime brokerage relationship after the event is a judgment about the client rather than about the market, and it is made with full sight of the positions and of how the account behaved under margin pressure.

Morgan Stanley, per the FT, held discussions since the fund launched and never opened a relationship at all. Goldman Sachs and Citigroup remain on the list. The tier of institution prepared to face this fund has narrowed, and that is more informative than any single name he is buying.

Second, where the exposure now sits. Clear Street is a prime broking relationship, not merely an execution venue. It was founded in 2018, it is not a bank, and its differentiator is a proprietary unified technology stack rather than the vendor patchwork the bulge bracket runs. So the financing and clearing relationship for a rebuilt concentrated AI book has moved from balance sheets subject to bank capital rules and supervisory examination to one that is not. Whatever view one takes of Clear Street itself, and it is well regarded in the space, that migration is the part of this story with implications beyond a single fund. This is what non-bank intermediation looks like when it happens to a name everyone is watching.

Third, and here I would push back on the comeback narrative: "fully paid" is not the same as "less leveraged". Paying cash for an option caps the loss at the premium and removes the collateral call, which is a genuine improvement in survivability, and prime brokers in discussion with the fund report he has said he will run far less leverage than before. But the FT is explicit that the premiums are small relative to the exposure obtained, which is the entire purpose of structuring a flex option that way. The leverage has moved from the financing to the instrument. Notional exposure per dollar of capital may not be materially lower than it was in June. What has genuinely changed is who bears the gap risk, and the answer is now the dealer rather than the fund.

On sizing, one caution that I think matters. Bloomberg reports, citing Nomura data, that between 4 and 8 September one fund spent at least 315 million dollars on calls in the flex-options market, across AMD, Bloom Energy, CoreWeave, SK Hynix, Intel, Sandisk and the Roundhill Memory ETF. The list matches the reporting.

Bloomberg does not attribute that spend to Situational Awareness, and I would not either. The inference is obvious and probably correct, but it remains an inference, and it is the sort that hardens into fact by repetition.

Two observations on the composition, if the attribution does hold. They are calls, so this is outright long convexity, not a hedge and not a spread. And the presence of a memory sector ETF alongside SK Hynix and Sandisk suggests the memory view is being expressed at both the single name and the basket level, which reads as conviction rather than opportunism.

One question stays open, and CNBC is explicit that it is unresolved: whether new outside capital has come in for the public book. If the premium is funded from what survived July, the concentration is being rebuilt with the residual capital of a fund that nearly failed six weeks ago. If it is new money, someone underwrote that trade remarkably quickly.

My reading: July was widely described as a clearing event for the AI infrastructure trade. The same participant, rebuilding the same exposure in the same names, six weeks later, in calls, with one principal lender having walked away and a non-bank having taken the seat, argues that it cleared considerably less than people assumed. The trade did not deleverage. It changed counterparty.

Happy to compare notes if you read it differently.

Best regards,

Laurent Chekroun
Graham Advisors Sàrl
+41 79 350 71 09