I wanted to add some thoughts after my previous article, now that the week closed and we had more information on what happened. I believe two/three narratives are fighting now, and we see the extreme of each in social medias.
Nothing changed. Buy the dip.
This is another fund blow up. Bounce coming.
This is a macro induced rebalancing and it won’t stop.
There isn’t much to comment on the first. This is most of FinX right now and it gets the most attention as this is what people want to hear. They might be right though. Until one day.
Still, things changed. Maybe not in AI itself but the environment did and ignoring that it impacts the market is like ignoring the weather impacts your life. I am no one to judge but stating that nothing changed is fundamentally wrong.
I’ll take a bit more time to detail the two other points, and conclude with my views.
Leopold v0.2
There is a case to expect a bounce.
Last week’s selling was due to deleveraging. There is no denying it, the violence and pace of the selling confirmed it, and if it wasn’t enough, we had later a Goldman Sach memo from the head of its prime brokerage that confirms it, using their own - extensive, knowledge.
According to Vincent Lin, head of Goldman Sachs’ prime brokerage business, in his latest report, hedge funds net-sold global AI stocks for five consecutive trading days from October 8 to 14, completely reversing all cumulative net purchases since the beginning of 2025 - in other words, every net purchase made over the past nearly two years has been wiped out within a single week.
Funds borrow money to leverage themselves, when the market falls they are forced to sell some of their shares to maintain their borrowing conditions, and when many funds have to sell at the same time, each sale triggers more selling, liquidation event, low and fast.
GS and their data confirms this happened, and confirmed this was mostly due to spot long sales, not short sellers. Funds sold shares.
Massive spot selling = deleveraging.
We also had rumors on a blow up from a fund called VARA AI, with its book circulating on FinX. When there’s smoke shared on FinX, there’s usually a fire - at least has been true for years now outside of clear stupid theories.
Lastly, most of the indexes still look healthy. The S&P is above its W21 and the Nasdaq closed just below it, with an insignificant deviation. Those could be simple retests at least of the indexes.
A portion of the AI growth names I follow look more or less the same, close to or bouncing on their W21/W50. Those usually act as demand regions, many funds/retails look at them as buying zones - me included.
Those three arguments make a case for just another liquidation event, even if we do not know the victim yet. Someone blew up, had to deleverage to save itself, pressed a big fat button which triggered others to do the same, etc… Until eventually the pressure stops, which should be soon as those events usually resolve behind closed doors when they get too big, and that seems to be where we’re at today.
When the fund is sold, selling pressure stops as no more deleveraging is needed. Buying pressure eventually comes back if only for optimism and “ah a fund blew up so it’s ok let’s go all in again”. Just like in July.
That’s case one. But it comes with counter arguments from case two.
Same, But Different
First, we had no news of funds blowing up, but that will be known only once deals are done behind closed door, so I not the final argument. But still waiting.
The best argument is the difference in environment and the selling conditions.
Rotation
Back in June, nothing had changed, the market just fell. No reasons, no warnings. The selling was confined mostly to AI names, just like today, but didn’t rotate. Today, if you look at defensive indexes, you’d see a strong weekly performance.
This isn’t just about selling tech, it is about selling tech and buying something else, defensives. GS confirms it once again as this massive tech deleveraging hasn’t much deleveraged the funds.
Goldman Sachs’ Prime book saw its overall gross leverage decline only slightly, by 0.4 percentage points to 310.3%, placing it at the 72nd, 91st, and 94th percentiles over one-year, three-year, and five-year horizons - near five-year highs.
Sell $1 left buy $1 right. This is what rotation looks like.
Triggers
The reason of this sell off matter even more. Let’s rewind back in June, what triggered the sell-off? I meant, what did we think triggered the sell-off then?
I don’t know.
I remember all FinX helpless - myself including, as stocks kept falling on no significant news, except maybe some strikes on Iran, a Fed pause or the usual AI ROI worries. There are always news but I’d frame June as “nothing new”. Nothing triggered the sell-off itself, it just happen on a stable environment. That’s liquidation.
This week was different.
The sell-off started after news. Important news, like Fed rate-hikes. I’ve detailed them at length already, but the TL;DR is simple: AI isn’t a question anymore, the question is who pays for it, with wars and inflation on top? Today, I can make a list of what happened these last weeks, each news triggering a bit more selling.
Mid-September - Fed rate hike and US10Y new highs. This was the setup, back in September, which actually triggered a push as the new was priced in and the market focused on fundamentals again, expecting relief from here.
Late-September - Trump refuses a deal with Iran and oil flow disruption increases while tankers that still pass increase oil prices due to insurance/logistic costs.
Early October - Ukrainian attacks on Russian refineries and Fed speakers talking about higher rates for longer, maybe even new hikes. Rumors on more U.S. strikes on Iran and oil starting to push higher again.
October 10 & so - OpenAI ARR blunder by the FT, more Ukrainian attacks on oil infrastructures, CXMT memory “FUD”, CoreWeave yields start to go nuts, and not just CoreWeave, even heavy cash flow companies’ yields go higher, like Meta.
October 13 & so - Broadcom CEO interview is misunderstood but he clearly states that access to capital will be a bottleneck - while environment toughen. Iran sank a ship in Hormuz and vowed to keep oil as high as possible. CPI & PPI come hot.
Last days - More attacks on oil infrastructures in Ukraine and week end fears with rumors of wars intensifications.
There was a clear roadmap and many trigger points that pushed the market lower, following a clear narrative. The market is asking for a single question: AI works. Yes. But who will pay for it if funding conditions get tighter?
We are far from “nothing’s changed but stocks go down”.
Price Action
Many indexes and some stocks are still on strong supports, but many leaders have lost them, on massive volume.
Nebius as an example, the healthiest neoclouds out there, supply constrained with enough source of financing to finance expansions - compared to CoreWeave & co. The market hasn’t been easy on, down -24% week on large volume - meaning no buyers stepped up at higher price while sellers sold massively, breaking down a trend that held since its first breakout, a year and a half ago.
Why aren’t buyers stepping in?
I believe that as long as the market has no visibility on financing conditions - meaning inflation and wars, it doesn’t want to jump on risk assets which depend on external funding - and therefore debt, to continue accelerating.
Frame it differently. Would Nebius stock continue overperforming if it cannot borrow to increase compute capacities?
The market prices cash generation acceleration potential. That’s what multiples are. That future cash generation is tied to compute providers’ capacity to expand their compute capacity, rapidly. If they can’t… Their future cash generation potential reached a ceiling. Why pay a premium for them?
My Views
TL;DR - I expect a bounce short/medium term, but nothing but a bounce. Conditions remain tight and as long as Trump doesn't stop the Iran war, I’ll be neutral to bearish AI, mostly hardware. I’d love to be wrong and look like a fool, very much.
I have been selling a lot these past few days, some broken supports and some names which didn’t as I need my book to remain afloat, above the S&P500 YTD.
I don’t do it because I’m bearish, but because preserving my capital is more important anything else. I do not need to be involved 24/7 in a volatile and dangerous environment to make money. I need to find two to three great trade every year. And keep the money they generated. Nothing else.
I do expect a bounce. Sooner rather than later. The only question is when. It could be Monday, it could be later, but eventually short sellers will take profits, kind of what we saw on Thursday but lasting longer. I do not expect a V shape recovery nor a rapid recovery, except if Trump was to stop the war, complete taco. This is the only situation in which I see AI hardware bounce hard and long.
Meanwhile, liquidity is rotating, into defensives but we can also identify some strength in other sectors. Software cooled off and didn’t break supports, cryptos have been volatile but saw stable prices through the sell off, many other assets are depreciated and could attract liquidity, sectors left behind for years… Lots of potential rotations.
This isn’t the end of AI nor hardware. This is a breather, waiting for better conditions. I believe it’ll last a few weeks to months before AI hardware leads again, but I struggle to se funds buy back in volume short-term - expect taco. Liquidity will come back to hardware, if only because wars should end - soon hopefully, and borrowing conditions improve.
Will it be in two weeks, post U.S. mid terms? Before, with a Trump taco? In a few months, with healthier inflation reports?
You can gamble on which it’ll be. It could honestly be any and I’d love to be wrong and have a strong bounce next week. But I believe this isn’t investing in companies’ fundamentals, but putting money on a binary outcome in a prediction market.
This isn’t what I do. But to each their own.
On another note, I know sharing a bearish take while writing an investment Substack is a bad look - and a bad move financially. This isn’t what most of you are looking for. I think this is a mistake. Being a perma-bull is great and all but one day, the music will stop. Maybe in a month, maybe in a decade. But if you only buy the dips, you will end up losing everything one day.
The hardest job of an investor isn’t to make money, but to keep it.
