Flood——史上最大AI基金爆仓内幕
Flood 的核心判断是，Situational Awareness 对 AGI 的判断方向正确，但组合构建把一场历史级胜利变成了被迫卖出。 据称，Leopold Aschenbrenner 将约5亿美元变成了50亿–100亿美元，AUM升至接近200亿美元，并在7月1日前后实现了超过450亿美元、包含杠杆的NAV。但部分仓位据称已达到某家公司市值的20%：“你可以判断正确、赚到一大笔钱，但最后仍然以错误告终。”
这场爆仓源于流动性失灵，并被杠杆、价格下跌和波动率上升加速，并不一定意味着底层AI判断被否定。 Flood 复盘称，经纪商要求在当天补缴数十亿美元抵押品；由于现金不足，Aschenbrenner 据称寻求紧急融资，最终组合被出售给 Citadel。一旦专业机构发现市场上存在被迫卖家，“市场在闻到血腥味后，惩罚人的方式会令人难以置信。”
据报道，Citadel 的买入是典型的困境组合交易：以折价接下巨额仓位，再从反弹中获利，或将其并入更广泛的组合。 Flood 假设折价10%，并指出数个标的随后上涨约20–30%。他更激进的猜测是，Citadel 可能利用这笔收购回补空头，或提前压低市场；但这明确只是推测：“那会是有史以来最疯狂、最像鲨鱼的一招。”
Jane Street 没有成为买方，是本期最关键、也最悬而未决的问题之一。 市场传言它曾向 Situational Awareness 投资约10亿美元，但据报道最终收购困境资产的是 Citadel，而不是 Jane Street。Flood 无从得知原因，但认为知情的LP通常会是过桥融资、补充抵押品，或竞购自己已经熟悉的资产的天然来源。
Flood 做空 memory 的依据，更多来自持仓和市场结构，而不是某个精确的基本面催化剂。 硬件过去一直是很难做的生意，相关市场涨幅最高达2,000%，而 Aschenbrenner 的买入还吸引了散户和对冲基金跟单。当 Micron 成为市场上交易最活跃的股票之一时，Flood 只问了一句：“还有谁来买？”
Flood 仍刻意不在“AI牛市已经完成出清”和“更大规模调整刚刚开始”之间做决定。 他的基金目前约50%持有现金，远高于通常不到10%的水平，因为被迫平仓可能先带来剧烈反弹，随后才继续下跌。“我不担心错过 Nasdaq 上涨7%的行情。我担心的是在-30%下跌中的-10%位置买入。”
更大的AI空头逻辑是，超大规模云厂商约1万亿美元的投入，除了云收入之外，仍未带来清晰可识别的回报，其中部分收入在 Flood 看来是自我循环的。 多头认为，成熟运营者把这笔投资视为关乎生存；空头的检验标准则是“拿出钱来”——可量化的收入、利润率或劳动效率提升。在这些结果出现之前，他预计市场可能继续惩罚资本开支更高的公司。
1. 一场天才般的AGI交易，最终膨胀到市场无法承受
Flood 将 Aschenbrenner 的崛起还原为交易史上最伟大的战绩之一：约5亿美元变成估计50亿–100亿美元，AUM升至接近200亿美元，杠杆后的NAV据称在7月1日前后超过450亿美元。在反转发生前，Flood 会把他排进“在世最顶尖的5名交易员”。
这套逻辑本身是自洽的：AGI即将到来，需要的算力、能源和基础设施远超市场预期，而相关股票被错误定价。相关标的随后上涨数千%，证明了方向判断正确，却也在杠杆作用下证明了执行方式不可持续。
据报道，基金LP资金或持仓中，超过25%配置于 Anthropic 股权。Flood 猜测，这一缺乏流动性的持仓可能以10–20%的贷款价值比支持进一步借款，实际上把私募股权敞口变成抵押品，继续押注更多公开市场股票。
Flood 粗略的风险基准是，单一仓位不应超过日均成交量的5%；但据他说，Situational Awareness 在部分标的上的持股接近市值的20%。这让整个组合暴露在聚光灯下、变得拥挤，也很难在不惊动市场的情况下退出。
2. 追加保证金把市值损失变成了最后期限
Flood 对经纪业务的解释从杠杆开始：基金提交现金、美国国债或其他抵押品，再从 Goldman Sachs 等机构获得融资、交易执行、掉期和定制化敞口。他可以想象 Aschenbrenner 在100亿–200亿美元规模上获得10–20倍杠杆；而他之后举的假设，是买入单个标的时使用4–5倍杠杆，并非在描述基金整体杠杆。
随着 memory 和基础设施标的下跌，抵押品价值下降、波动率上升。这两项变化都会推高经纪商要求的安全垫，于是出现 Flood 戏剧化描述的场景：“我们需要紧急谈一谈”——随后要求补缴数十亿美元，而且可能必须在当天收市前完成。
主持人难以置信地追问，截止时间真的可能是“今天”吗？答案毫不含糊：“当然。” Flood 表示，下跌期间紧急募资通常意味着流动性危机，因为主动在弱势市场卖出等于公开承认困境，还会引来其他交易台做空、扩大报价价差，或等待更差的价格。
如果始终无法补充新的抵押品，Goldman 可能会向少数有能力承接整个组合的机构询价。Flood 知道据报道 Citadel 和 Millennium 都参与其中；他的假设是，胜出者以约10%的折价向 Goldman 电汇可能高达100亿美元，随后证券被释放并转入其账户。
3. Citadel 买下了它最擅长定价的困境资产
市场传言 Jane Street 曾向 Situational Awareness 投资约10亿美元，因此它没有成为买方，令 Flood 尤为不解。一个已经熟悉该组合的LP，理应是天然的过桥贷款方或竞标者，但据报道最终出手救场的是 Citadel。
Flood 将这笔收购描述为 Ken Griffin 的标志性交易，并提到 Citadel 在2001年收购 Enron 能源交易部门、参与 Amaranth 天然气交易公司崩盘，以及 Griffin 在一次量化交易失序期间接触 Cliff Asness。他记得 Asness 当时说：“我听到死神的镰刀在敲我的门”（“I heard the Grim Reaper’s scythe knocking at my door.”）。
保守的解释已经足够：Citadel 判断这些资产具有正的预期价值，并因速度和确定性获得补偿；同时，它可以把仓位纳入数千亿美元规模的更广泛股票敞口。几个标的反弹约20–30%后，Flood 说 Citadel 在账面上已经赚了大笔钱。
更具挑衅性的情景仍然只是明确的假设：Citadel 可能做空了部分标的，了解 Aschenbrenner 的风险上限，并利用被迫出售的大宗仓位回补空头或退出仓位。Flood 承认“我只是在猜测”，但认为组合规模巨大，令其持仓和潜在崩溃点异常容易被反向推演。
4. 杠杆造就了传奇，也决定了清算价格
Flood 认为，如果 Aschenbrenner 在最后一轮 memory 挤压行情中降低风险、转为做空，或持有100亿–300亿美元现金去买入残局，他可能会成为有史以来最好的交易员。相反，在可能是自己第一次真正遭遇回撤时，“他在顶部有点贪心了”。
与 Citadel 的比较，不是低杠杆对高杠杆。Flood 猜测，Citadel 可能在约800亿美元AUM上运行约20倍杠杆，但同时拥有严格的分散、仓位和风险控制，避免所有仓位同时朝不利方向移动。
他的教训是个人化的，而不是胜利者叙事：“杠杆往往是杀死你的东西。它让你成功，也杀死你。”即便面对看似千载难逢的机会，他自己的基金通常也不会超过2%的方向性总杠杆。
Flood 也拒绝进行幸灾乐祸式的围观。Aschenbrenner 对股票上涨的判断“完全正确”，即使 carry 回拨可能让结果从约10亿美元降至2亿美元，他大概率仍然赚到了惊人的金额。失败在于仓位和生存能力，而不是缺乏洞察力。
5. 当每一个边际买家都已经做多，memory 顶部便出现了
Flood 将更多交易架构归功于同事 Kyle：memory 和硬件过去一直是难做的行业，如今却正在经历“史上最大的逼空”（“the mother of all squeezes”），部分市场涨幅约达2,000%。这本身足以让人考虑逆势做空，但不足以提供精确的入场时点。
更强的信号来自反身性持仓。Aschenbrenner 的买入推高了小型公司的股价；散户跟随他，其他基金则抢跑或尾随一个被认为在结构上不太可能卖出的买家。最终，Micron 主导了交易，crypto 交易员庆祝公开盈亏，于是问题变成：“还有谁来买？”
Flood 不认同 crypto 参与者那种简单化的股票分析：把财报喂给 ChatGPT，再拿 Micron 的 P/E 去对比专业半导体投资者。他并不是说基本面无关紧要，而是认为，在一场“超高效的游戏”里，拥挤且滞后的共识几乎没有 alpha。
这笔空头大约痛苦了6周，相关标的有时单日上涨20%。Flood 的基金将仓位控制在10%，没有使用有意义的杠杆；他说，通过 HIV-3 做空这些标的，平均还能获得约40%的补偿，而不是支付股票借贷费或期权 theta——代价是承担逼空、跳空、黑客攻击和交易对手风险。
6. 清算之后，决定性问题是AI能否拿出真金白银
Flood 认为目前有两条路径：这次清算只是洗掉了杠杆，“AGI只会上涨”随后恢复；或者，更广泛的调整才刚刚开始，因为 S&P 500 仍接近高点。此前的清算，包括 Three Arrows 和“10/10”，都曾先带来30–40%的反弹，随后再遭到市场惩罚。
这种不确定性解释了基金为何约50%持有现金，而通常不到10%。抄底和周度期权过去反复奏效，但成功的均值回归交易者如今手上的资金更多，一旦市场状态改变，潜在损失也更大；Flood 预计，“耐心会在这里得到回报。”
在情绪顶峰，即便财报也可能变成一场无法取胜的游戏：Micron 的爆炸性业绩被视为不可能重现的高点，而 SK Hynix 仅仅因为没能达到被抬高的预期就遭到重创。“真的不可能再好了”——如果 Micron 继续以当前速度增长，它会成为全球市值最大的公司。
Flood 的基本面担忧是，超大规模云厂商已经投入约1万亿美元，却看不到明显的云业务之外收入或成本节省。Meta 增长约20%，同时资本开支正把自由现金流推向负值；Google 的情况也被他以类似方式描述。他承认多头逻辑：这些公司具备成熟的资本配置能力，并把AI视为生死攸关的事情；但他始终回到一个要求：“拿出钱来。”
他的 crypto 持仓比外界看到的 Hyperliquid 空头更复杂。他仍持有 Bitcoin 和 HYPE；针对已经升值的 HYPE，他使用衍生品对冲美元 delta，同时收取质押收益——他称这在美国无需纳税——以及做空资金费。“我可能在做空 HYPE，但这并不意味着我看空 HYPE。”
完整逐字稿
All right. Flood, wow, what a day. Thank you for being here. I don't know if you remember this, but this is your first time on the stream since you were shilling Hyperliquid at around $5 in 2024. Welcome back, man.
Yeah, that was a pretty legendary podcast where I said Hyperliquid was the only crypto asset that you needed to own and pay attention to, and that you could just buy it and log off. That was obviously true, but I victory-lapped that one enough. I think more interesting things are happening right now. What's on your mind, man? What do you want to discuss?
Yeah.
What do you want to discuss?
I also want to give you credit as a friend of the stream.
Yeah.
We'll get over the glaze session really quickly, but I have two of your tweets on my TradingView. The first one is “Oil top” at the top, and the second one is “Memory top” at the top. Well done on that. We follow your Hyperdash; we know what it is.
Anyway, today's a crazy day. I was covering it live: Situational Awareness. Boy wonder Leopold Aschenbrenner is out. He sold completely to Citadel. You understand the intricacies of this better than I do. Can you first set the stage and break down what has just happened today? Then we can talk about the implications.
100%. The way trading works when you're at a hedge fund is that you have a pool of assets, typically cash or Treasuries, and you go to a broker like Morgan Stanley or Goldman Sachs and say, “I have this basket of assets, and I would like to trade other assets. I would like exposure to these equities.”
You can buy them through your brokerage, you can do total return swaps, you can do options, and you can do a number of different things. The reason you would do this rather than using Interactive Brokers or something that we normies use is that you get a tremendous amount of leverage. You can get much higher margin rates, better financing, bespoke products, and better execution.
There's this entire business at banks and brokerages called their ATS, or alternative trading systems. This is where they block trades for you, trying to give you tight execution. It acts as a leverage supplier, risk manager, and execution service.
How much leverage are we talking?
Which is perfect. You can get tremendous amounts of leverage on equities, as equity volatility is typically quite a bit lower than crypto. We get tremendous amounts of leverage on crypto—we get 100-to-1. I don't know what someone like Leopold would get, but you could imagine him getting 10–20× leverage on $10–20 billion.
Whoa.
The reason you use leverage is to juice returns. You want to use as little capital as possible while keeping it in T-bills for the risk-free rate of return, and then be able to own equities or have equity exposure that essentially allows you to put on these positions.
If you're a winning trader, you're going to have more and more capital, you're going to get more and more cash, and you're going to continue compounding. Essentially, this guy's thesis was genius. It was kind of infuriating for a long time, and I don't want to dance on his grave too much. My heart goes out to anybody experiencing a large drawdown. I've been there in the past.
But this guy's strategy was genius. He said, “AGI is coming. AGI is going to need a tremendous amount of compute and energy, there's going to be a ton of infrastructure build, and these equities aren't priced properly. We can accumulate them very cheaply, and they'll reprice significantly.”
He was right. He essentially turned $500 million into, I want to say, $10 billion—or, you know, $5–10 billion. Then he raised more and got his AUM up to about $20 billion.
What was reported was that, on July 1st, which was close to the top of the market, it looked like he had a NAV in excess of $45 billion. NAV includes leverage, so it's different from AUM or actual LP capital.
Got it.
That was the exposure of his book. The thing to remember is that he wasn't just trading public equities; he also had a large private-equity book. More than 25% of the LP capital, or the positions at the fund, was in Anthropic equity, which is obviously illiquid.
But if he was sharp—and I'm sure he was—he probably went to the bank and said, “I would also like to use this as collateral.” The bank said, “Okay, we can give you 10–20% LTV on something like that.”
This is a good way of juicing returns. If I can use locked HYPE as collateral to trade, and I'm a profitable trader, I'm essentially using free money to some extent. This is how their portfolio is constructed.
Modern portfolio theory dictates that you never really want to be more than 5% of the average daily volume as a position size in a single-name equity. SanDisk trades $10 billion of volume a day, so you don't really want to have a position larger than $500 million. This is the rough framework, although it's far more complex than that.
Uh-huh.
If you were starting a fund and asked what a reasonable risk metric was, I would say, “Your position can't exceed 5% of the average daily volume traded.”
Okay.
Right? When you look at his book, including leverage, it was clear that he was 20% of the market cap—not the average daily volume—of a lot of these single-name equities.
Dude—
That is way outside the bands. The reason you have these constraints around position sizing is that when the market moves against you very quickly, people may know that you have a large position. Then you're going to get adversely selected: liquidity thins because they realize you may be a forced seller, and the market punishes you.
Markets have an unbelievable way of punishing people when they smell blood. If I know that you have a large position in an asset, you're leveraged, and you need to sell, why would I quote any reasonable price? I'm just going to keep walking the book down.
This happens every time, I feel like. They hunt these positions out, and then you're stuck.
Yeah. Think about what finance is. Finance is a big poker game. There's money in the middle. Everybody contributes money by buying equities, stocks, and assets, and then you try to take more money out than you put in by taking your chips off the table.
Imagine there's a pile of money and we're all sitting around a table. Suddenly, everyone starts rapidly grabbing all of the dollars off it. It's going to be much harder for you to grab your initial balance. This is how liquidity and markets function: when people know that you're a seller or that you're really concentrated, it's easy to figure out your positioning and for people to press the market down on you.
What was interesting, and probably the most interesting thing about Situational Awareness, was that Jane Street was an investor. Jane Street is probably the top trading firm in the world. Arguably, it could be Citadel Securities, but Jane Street certainly dominated last year, as it had the best performance.
I was surprised to see Jane Street invest in a stock-picking fund. It's clear this Leopold kid had a tremendous amount of alpha that was increased by leverage. I was really interested to see Jane Street investing a rumored $1 billion into Situational Awareness.
Whoa.
What's even more interesting is that when Situational Awareness was forced to sell its assets, it wasn't Jane Street that bought them. It was actually Citadel.
If I'm winding down my fund or I'm under duress—if I have a capital call, or maybe I get stuck on the wrong side of duration—I would go to my LPs and say, “Hey, guys, I need capital. You're an investor in this fund, so you clearly understand it well. I need a bridge loan or some sort of financing, or I need you to buy these assets from me.”
Presumably, if you're invested in a fund that holds these assets, you must think they have some value. But Jane Street wasn't the buyer. It was Citadel.
This is kind of Ken's trade. Ken has done this a few times in the past. He bought Enron's energy trading division in 2001, and then he blew up this natural-gas trading firm called Amaranth for $6 billion. He basically twisted the knife and gutted that firm as well.
Ken does this, and this is the Citadel trade.
This is their signature sharky trade. He's done this also to Cliff Asness at AQR. There was a big quant quake in the 2010s where a lot of people were running very similar statistical arbitrage strategies, and they all sort of correlated. The market had this sort of flash crash.
Ken called Cliff Asness and was like, “Hey, I'll basically buy your fund from you.” Obviously, Cliff said, “No, fuck off.” But Ken has a history of doing this, where he's known as the Grim Reaper, basically.
Whoa.
There's a very famous interview with Cliff Asness where he was asked, “How did you feel when Ken called you?” And he said, “I heard the Grim Reaper's scythe knocking at my door, and I just kept walking towards the light, and I tried to walk away.”
Ken is known for doing this. The reason why is, back to risk management, modern portfolio theory dictates that you have these rules, best practices, and guidelines. But the real reason why all of these firms have really strict risk management is because they are running so leveraged.
People look at Situational Awareness's leverage and say, “This is crazy. This guy was 4x leveraged long.” Well, Citadel probably has, I don't know, $80 billion of AUM, and they're probably 20x leveraged.
Wow.
The difference is that they have extremely tight risk controls and extremely tight constraints that prevent them from, ideally, having an event like this where all of their positions are moving against them at once.
Wow. Okay, before we talk about the downfall, I'm curious. I just love the lore. Where does Leopold Aschenbrenner's ascent rank amongst the best run-ups ever, before today? Before June 28th?
Definitely top 5 traders alive pre-blowup, just in terms of going from $500 million in AUM to $45 billion in NAV.
Top 5?
That is a tremendous increase. He absolutely could have been the best if he had flipped short or something. If he had the ability to de-risk his book at the top during this last memory squeeze, when we saw the Micron blood earnings, this guy probably would've been the best trader of all time.
He would've been sitting on billions and billions and billions and billions, maybe $10 billion, $20 billion, or $30 billion worth of cash. Then he would've been able to buy the blood, and maybe because they knew they were shorting, they could've put on a ton of out-of-the-money put options and even made money on the short.
Now, the craziest thing about all of this is, why did Citadel buy this? One, they thought it was positive expected value. They probably got some sort of discount. They were like, “Okay, we look at these assets. It's a momentum basket. They've sold off 30%, 40%. We probably get a 5%—I don't know, 10%—mark-to-market discount just for filling everything and making it easy for Leopold and stuff.” You're going to pay for simplicity and convenience.
Okay.
The S-tier execution, if they're really the GOATs, would be if this was a way of them covering their shorts. Imagine if Citadel was actually net short these names, and then they bought this book.
Because it was very public, in traditional finance, all of these firms are able to decode, with a high degree of probability, what almost every single trade in the market is.
Mm.
Because Leopold was so concentrated and only holding specific names, and he was not a high-frequency trading firm—he was a long-only or long-short fund—it was very easy to understand what this guy owned.
Presumably, if you got access to his investor documents, you would know. It was probably widely circulated. Leopold shopped his fund around quite a bit and asked a lot of different firms to invest. I know someone who's an LP in Situational Awareness.
You would know what stipulations he had around, “If there's a 30% drawdown, we'll de-risk the book. If there's a certain level of volatility, we'll de-risk the book.” I don't know. I'm just speculating here.
Wow.
But it's not hard to reverse-engineer because of the sheer size. For someone trading under a billion dollars, even with some leverage, it doesn't really matter. Citadel and all of the other firms would be able to see your flow, but they'd be like, “This is sort of not uninteresting.” They'd try to match it to a 13F and say, “Okay, this is just a small fund.”
But because he was so large in small-cap equities—companies with $100 billion and less in market capitalization—SanDisk was one of his larger positions.
Bloom. NBS.
Yeah, Bloom. But these are really small companies. Google is a $4 trillion company. Apple is a $5 trillion company. That's very different from a $100 billion or $200 billion equity.
Because Leopold's positions were so large, presumably Ken and the bandits could have known, “We know when this kid's going to cry uncle, and that's the point where we're going to be able to have a forced liquidation to exit into.”
I don't know. That would be the craziest, sharkiest move ever, but that's what finance is. Finance is a blood sport. It's a combat sport. For me to make money, someone else loses.
There was a wealth transfer from people who were short memory to Leopold, people who were Leopold copy traders, and people who were long memory. Then there was a rapid wealth transfer from people who were long memory to people who were short memory. That's how markets function.
Holy fuck. Okay, how did this actually work? The last 4 days, we're sitting here watching memory tick down and tick down and tick down. Is Leopold Aschenbrenner at his desk with an FTT $22 line, waiting, and the banks are calling him? How does this process actually happen over the last 3 or 4 days?
How does this work? There are traders at the Goldman broker desk, at the trade desk.
Uh-huh, uh-huh.
Leopold comes to you and says, “Hey, I just got another $10 billion of cash from Jane Street, or $5 billion of cash from Jane Street and all these guys. I want that 4x, 5x leverage you give me, and I want to buy a billion dollars of Bloom, a billion dollars of SanDisk,” whatever he bought. He held Bloom, SanDisk, NBS, and all of the names that are bouncing today.
Goldman is like, “Okay, this is pretty fucking crazy. That's fine. We'll take your money. We'll place this bet for you.”
The thesis from Leopold, and the thesis in his Situational Awareness essay, was, “I think AGI is coming. It's inevitable. Digital God is coming, and these assets are going to be up only in a straight line. The magnitude of investment and scale needed is way beyond what people are currently forecasting and calculating.”
Goldman is sitting there and probably thinking, “These guys are fucking crazy. This is the stock market. Markets go up and down.”
But it worked, and so Goldman is like, “Okay, this kid is making billions and billions of dollars.” This has happened a lot. A lot of funds have had crazy, crazy run-ups.
Then the market starts unwinding, and Goldman is like, “All right, Goldman Sachs doesn't lose money.” So they say, “Now you get a margin call.” Now it's like, “We need you to post another $5 billion of collateral,” not just because markets went down, but also because volatility went up. The expected move is no longer—
So, like, when they give him the margin call, they're saying, “The collateral—
Yeah.
—you've given us is no longer sufficient based on how large your position is. You have—
Yeah.
—to give us more money, or volatility has gone up. Price has gone down, volatility has gone up, or both have happened. We don't feel comfortable with how much collateral you've posted.”
Your ops guy gets a call and probably a few emails saying, “We need to have an urgent conversation.” You look at the market and say, “Okay, here's your exposure, here's your portfolio.”
Hopefully it's not double-pledged like Archegos was. That's why Bill Hwang went to jail, by the way. It wasn't because he lost money—
Oh.
—it was because he borrowed against the same collateral from multiple banks, juicing his leverage 2 to 3 times what other people thought.
Essentially, Goldman comes to you and says, “We're really worried about the exposure here. We need you to post additional collateral.” Leopold is like, “Fuck, I don't have the collateral.”
That's why the news came out that he was raising funds. You're not raising funds when the market's going down because you're having a good time.
You're raising funds because you have a liquidity crunch or a call, and you presumably don't want to sell your assets. Because when you sell your assets and the market's moving against you, it exacerbates the move.
Yeah.
Because now all of these very sophisticated trading firms see you selling and they're like, "Oh, this guy is liquidating. Let me quote even wider. Let me try and front-run this guy and short him because this guy has a huge position to unwind." You get stuck.
And so Goldman then goes, "All right, we need collateral by market close." And then Leopold doesn't have it.
Oh, it's that aggressive, like today?
Oh, yeah. We need it today. We need billions of dollars today.
And he's like, "I don't have any. I have nothing. I have no dollars."
Yeah. So you call your LPs, like Jane Street and all this stuff, and you go, "Hey, Jane, I need $5 billion." And Jane goes, "Fuck off. We're probably on the wrong side of this too." Because they were investors in Situational Awareness, so they at least somewhat understood what was happening.
And it was very profitable to trade these momentum names. Not only are they making money directionally, long-short market making, but then also market making the options because the implied volatility was so high. There was so much retail interest. These names were moving a tremendous amount, and so Jane Street probably said, "Fuck off."
And then Goldman goes, "Okay, markets are closed," which is sort of nice in this instance, and then they run a process. They call Millennium, they call Citadel, they call D. E. Shaw, they call Hudson River Trading—
Because there's only a couple of guys that can fill this—10 guys. It's not that many people that can pull this money up, right?
Well, there's a bunch of different wealthy people in the world, but Bezos isn't going to pick up the phone for this, right? That's not in his wheelhouse. Berkshire Hathaway could have been a call.
Yeah, you call all the big guys with big pools of capital who you know always have cash. Citadel always has cash. Jane Street always has cash. Millennium always has cash. These guys have almost more cash than they know what to do with.
And so they go to them and they're like, "Hey, here's the deal." Citadel and Millennium were the only two who I know were in the process. Citadel and Millennium then race and try to forecast and say, "Okay, what's the price? What price should we offer?"
And then basically, best bid wins, right? So let's just say that was 10% under market. Citadel wires $10 billion to Goldman on behalf of Leopold, and then Goldman releases the securities to their account.
To Goldman, yes.
Wires $10 billion to Goldman, and then Goldman releases the securities to their account. This is hard to do because the markets are closed. You don't know what the move is going to be tomorrow, but you have some expected-value calculation.
And then it's even harder if this liquidation happened while the market was going on. That is entirely possible because we don't know where Leopold actually tripped his risk limits or the guidelines that Goldman set for him. Imagine you're trying to compute the value of this portfolio while it's swinging 5% to 10%—
It's just drilling.
—during an FOMC meeting. So, best bid.
And it's also like—you have to sign an NDA to do this, right?
Okay.
And you're not allowed to insider trade, but who's refereeing this? Listen, if Ken is calling you, you're in a really bad spot, and word gets out fast in finance. People talk.
Rumors are fine to trade on. You're not supposed to trade on direct information. That's illegal because it's non-public. But you can hear a rumor about something, and it's secondhand information, and that's totally fine.
The art of this is that Citadel is so good at systematically going into something and pricing it, and having the balls to say, "All right, I'll bid this. It seems good. I'll bid it." And they've clearly made a tremendous amount of money because all of those names are up 30%.
Okay, I know you probably don't want to put a tinfoil hat on this, so go as far as you want, but I'm curious. This is Citadel's trade, as you say. How long, hypothetically, would something like this be in the works?
Citadel has this bizarre publication that comes out on Tuesday saying, "We think Warshaw is going to hike." And everyone's like, "Wait, what the fuck?" It's also at this precarious time where forward guidance is gone. It's his second FOMC, the first one where he could presumably do something, and then the markets start to panic.
In particular, the Leopold names, which maybe that was him selling—I don't entirely know—start to scream. Bloom is crushed, SanDisk gets crushed, Nebius is down 20%. What are the machinations in the background that could hypothetically be going on here, and how long would that have been in the works for someone like Citadel?
Well, look, if I'm buying a house, I'm not going to talk about any of the positives of the house. I'm going to try and argue a case for why my price is fair. "I heard gunshots in the neighborhood. The fence is a little messed up, and there might be some structural issues." You're going to say everything possible to drive the market down.
This is the same thing Ackman did. He went on CNBC and—
About COVID, was crying.
Yeah, and he was talking about how disastrous COVID was going to be, and in the meantime, he was actually covering his shorts and flipping long.
Wow.
There's nothing against it. Is it a little bit shady? Sure. But if I know that a fund is stuck, I know that someone's leveraged, and I have the size to do it, why wouldn't you press the market down?
Wow.
Because they're going to be far more agreeable when Goldman is about to say, "Listen, we can dump this into the market." And they go, "No, no, no, let's find a bidder and get this cleanly off our books."
Goldman would never really dump it out of the market. That would incur quite a bit of losses. But, yeah, listen, Ken is not a nice guy. You don't make $50 billion by having everyone's best interests in mind. Probably more—$80 billion, $100 billion. You don't become the GOAT by not sharking people.
So, yeah, I would do everything possible to get as much information as possible and understand every single thing that this guy owned. It's like you're playing poker and the fish has all of the chips, and you're like, "Okay, I'm going to study every single thing that this guy does, all of his strategies, and then I'm going to play a style of poker to try and take all of his money." That's exactly what finance is.
This guy was the big fish at the table. But, listen, it's a good lesson for everybody that you can be right, and you can make a tremendous amount of money, and still end up being wrong.
Leopold's going to be fine. The question is whether he'll have clawbacks on his carry. But, oh no, boo-hoo—instead of making $1 billion, he makes $200 million. This kid made a tremendous amount of money. The people laughing at him—that's why it's not really fair to dance on his grave. He's done exceptionally well.
Of course.
But it's an important thing: leverage is often the thing that kills you. It's the thing that makes you, and it's the thing that kills you.
We almost never lever a fund. It would have to be a once-in-a-lifetime event for us to even exceed 2× leverage on a directional position for our gross book. There's just no need.
Markets always go up more than you expect, and they go down way more than you expect. Think about Bitcoin. When you first got into Bitcoin, did you think it would go to $100,000?
No, that's nuts.
No way.
Yeah.
And then, from $100,000, it's gone all the way down to $60,000. More importantly, from $70,000 or whatever it was in 2021, it went all the way down to $18,000.
Yeah.
I would have never expected it to retrace almost 80% because of FTX and stuff. So, listen, leverage is the killer. He was right.
He was completely right. These names have gone up thousands of percent, but he got a little greedy at the top, and that happens. You feel for him—this kid has never had a drawdown. This is probably the first drawdown he's ever experienced.
He didn't have a tremendous amount of trading and finance knowledge. I don't really know who works at Situational Awareness. I assume they're not top-tier quants. They were probably people who really believed in AGI, really believed that the thesis was right, and they were right for a long period of time.
They were right. So what's the game theory if you're Ken Griffin at Citadel? They presumably bought everything 10%, give or take, under market, and Bloom Energy is up 25%, SanDisk is up 25%, Micron—all these names.
Yeah.
SK Hynix is up 20%. You're up a fucking lot.
Yeah.
How long are you in these positions for, and how do you think about that?
It's impossible to know without understanding the composition of their entire portfolio.
Got it.
Right? They may have said, “You know what? We're going to sell a lot of our Nasdaq position,” or, “We're going to sell some of these hyperscaler positions, and then we're going to buy these momentum names because they've come down so much, and eventually we'll rebalance our portfolio to whatever we think is optimal.” They're trading all the time, 24/7, 365.
Got it.
A $10 billion position is large because of the market cap of these names. These are not massive companies, so owning $1 billion of Bloom Energy is actually a lot. That would be a single-digit percentage of the market cap.
It's—
Right? But Citadel has hundreds of billions of dollars of equities, so this slots right in. They probably made some movements, and then it's kind of fucking business as usual.
They'll rebalance out of the positions and go back to their optimal position sizing. They just got a cheap price.
How is Goldman Sachs not just sliming everybody? They know what everybody needs at all times.
Yeah.
Or are they?
I mean, Goldman is—
They just are.
Goldman has had around 20 losing trading days on their desk this year. They're going to make $8 billion. They are tremendously informed about the market, and they would never front-run or back-run a client. There are a lot of regulations. You would never go to the broker that's sliming you.
But do they know what's going to happen, and do they have a good understanding of the composition of people's portfolios—the net position of retail versus institutional? They have a lot of information that other people don't, and whenever you have asymmetric information, you're able to potentially make trades that are far more plus-EV than the market thinks. Goldman makes a tremendous amount of money. Here's a key example.
Okay.
If you place an order to buy QQQ, and I'm a market-making firm using payment for order flow, I'm not allowed to then go buy QQQ and sell it back to you. You're not allowed to get MEV'd. You're not allowed to get front-run.
Oh, okay. I see.
That is illegal. But I can trade Nasdaq futures because that's an entirely different product.
Got it.
So that is—
Got it.
That's sort of the—
Again, this isn't my world. I'm not claiming to be an expert. This is above my pay grade. But this is sort of the way modern finance functions: there are people who have strategies, they come to the market, they run it up, and then the market has the ability to punish them.
There were a lot of firms on the losing side, as well as retail traders, who shorted memory. We were underwater on memory shorts for probably 6 weeks this year.
Wow.
We averaged up. We gave ourselves—
That was a brutal 6 weeks as well, by the way. Painful 6 weeks short.
It was. Yeah, yeah, yeah—it was ripping 20% every day. But we just had this thesis that these things were a bit overheated.
This is also where game selection and venue selection make things a lot more comfortable. We were being paid 40%, on average, to be short these names on HIV-3. If you went and shorted on IBKR, you would pay a borrow rate to short a stock. It costs you money.
Got it.
If you buy a put option, you're paying theta, which is the time decay of the option. We were short on HIV-3, and we were being paid funding. Now, you have squeeze risk, weird price-jump risk—maybe Hyperliquid could get hacked—and counterparty risk.
Yep.
That's what you're being compensated for. We were being paid to be short these names. We gave ourselves a large margin of error, and we didn't use any leverage. You're always levered a little bit when you're short because of the way a short functions, but it was not a large percentage of our overall fund. It was a 10% position.
Got it. Okay, I have 2 questions for you, and then I'll let you go. This is generational.
Sure.
First question is, I've always been curious about this idea that the news or the outcome was written in the charts. Things start to roll over, then they go, and then the news comes out—whether it happened because of the charts or the charts happened because of the news, whatever it is.
I have 2 lines on my TradingView. One is the oil top, which is less relevant—
Mm.
—but you sniped it. The second one was memory, and you were pretty vocal about this.
Mm. Mm.
At the time, fine, things were euphoric and overextended, but it didn't just go down. You were vocal about this for multiple weeks: “I'm short memory. I'm short memory. Memory top, memory top, memory top.”
Mm.
Do you think it was topping because guys like Leopold were being hunted? Do you think they were being hunted because it was topping? How and why did this play out the way it did?
The thesis was, one, memory. Kyle was far more involved in this trade than I was, but if I had to architect his thesis, he was like, “Hey, look. Memory and hardware have typically been a terrible business. Historically, it's been a terrible business. It's experiencing the mother of all squeezes.”
Part of it is gut feeling as well, where you're just like, “Hey, these markets are up 2,000%.” Surely it's not a crazy trade to think about shorting them.
The other thing for us—and this was the component of why I was bought into this thesis—was that I looked at Leopold's portfolio, the size of the equities he was trading, the size of his fund, and how much of the market he represented.
There was also the follow-on effect of people copy-trading him because he had made so much money: retail traders copy-trading him, and other hedge funds copy-trading him because they knew he was going to buy and push the market up further. They tried to front-run or tail that, because presumably they knew he wasn't really a seller.
Got it.
He had a long-term thesis on this. The market was getting overextended.
This is going to sound shitty, but crypto people don't have any alpha in equities. Equities are an extremely difficult and hyper-efficient game, with the smartest people in the world playing. A lot of people on crypto Twitter were like, “Look at Micron's P/E ratio.” I'm like, “Guys, this is fucking crazy.”
There's no way you think that because you put the 13F—or, excuse me, the 10-Q—into ChatGPT and asked, “What is the P/E ratio?” you have more alpha than people doing semiconductor research, Jane Street, and all these firms.
It was just like, okay, crypto people are unfortunately quite late. You saw this with gold, you saw this with oil, and you're seeing it again with memory. When I see people posting a lot of P&L, posting about how this is the craziest run ever, and you're seeing everybody long memory, you're kind of like, “Well, who's left to buy?”
Who's left to buy?
Right? Who's coming in and who's going to buy $100 billion of Micron over the next few months? It's nobody. When everybody is long and Micron is the most traded stock in the market for weeks on end, at a certain point, the smart money has gotten out of its position and sold to the people who are really late.
Yeah.
More often than not, that tends to be crypto people, unfortunately.
I'll give you credit before my last question.
Because the OG stream viewers know you're in here a lot, we were watching Micron earnings, which was the greatest earnings of all time. You were saying something along the lines of, “Watch the lack of volatility. This is the top.” It wasn't at first, and we were kind of like, “Fuck you, Flood, fucking bear.”
Yeah.
And then it played out the way it played out. My last question for you is about the collateral damage of what happens from here. There have been a couple of posts, I think, from TradFi talking about how some other funds are potentially in trouble.
Mm-hmm.
This is a little bit of a weird day. I wasn't that in tune with the markets when the whole 3AC-FTX situation happened, but it was pretty dark because there was extreme collateral damage and everything was limit down. Now everything is limit up.
Mm-hmm.
Which makes it ironic.
Mm-hmm.
But I'm curious about the collateral damage on other funds, how you think that will play out, how they're allocated, and then your views on the market short to medium term from here. I know you closed a lot of your memory shorts.
Yeah. We closed too early. I didn't know Leopold was getting liquidated. If I knew that he was going to be liquidated and I had heard some rumor—I mean, I knew he was in trouble, just because it's easy to reverse what his holdings were—but I didn't know how levered he was. That's just information you're not privy to.
Yeah.
Sorry, give me one second.
No, you're good.
What I think happens from here is that there are 2 frames of reference. One is maybe AGI is up only and this is a temporary pit stop. We just had to kill some retards, and then it's going to resume up only. We washed out a lot of the leverage, and now the smart money is rebuying. The other frame of reference is that equities aren't really down that much. Pull up the S&P 500 chart. It looks like—
3% up from ATH.
It looks like it's a tiny blip. The correction hasn't really started. Actually, what happened after Three Arrows got liquidated was that we had a massive bounce. After 10/10, we had a massive bounce. Things were up 30% or 40%, and then they got hammered later. So I don't know. I don't really have a strong bias. We're about 50% cash.
Okay.
We're pretty risk-off for us, as we're typically close to fully deployed, running less than 10% cash. But I think patience is going to be rewarded here. There are a lot of people who have made a tremendous amount of money buying dips and doing mean reversion and saying, “Oh, if equities are down 2%, I buy weekly options and then I make 500% returns.” Or, “Oh, Micron is red, I buy. The next day it's up 8%.” That has been the correct thing.
Yeah.
These people have made a lot of money, but that means they also have a lot of money to lose.
Yep.
What I think is that you're going to be rewarded for patience, because there are 2 things that happen at peaks. One, when earnings reports come out and they are booming—when they are blockbuster—the market, when sentiment has turned, says, “Oh, this is a high-water mark. It's down only from here.”
Yeah.
So that was Micron, right?
That was Micron.
This was a one-of-one.
How could it get better?
It can't. It literally can't get better. Micron would be the biggest company in the world if it continued growing at the pace that it's grown. Second is SK Hynix, where people say, “Oh, we're expecting blockbuster earnings,” and then you miss. So it's almost like you can't win once market sentiment has really turned. You beat too big and the market actually punishes you because they say, “This is the top,” or you miss and you get hammered like SK Hynix did. It's no wonder that SK Hynix jammed that $27 billion offering in.
Yeah.
The number-one thing that we are looking at and thinking about is that there has been $1 trillion of spend by hyperscalers. I don't know where you can point to me in the financials and show massive revenue increases other than cloud, which is recursive to some extent.
Yep.
I don't see where there is some line item that says, “Hey, we were able to have a 70% cost reduction,” or, “Instagram Reels are now 30% more profitable this quarter.” There is very little to show for the spend. People are asking, “Why did Meta get hammered?” Meta is growing 20%, and they're spending to the point where they're now free-cash-flow negative.
Yeah.
Google is the same.
Google, same thing.
The bull case is that all of these hyperscalers are smarter than we are, they've compounded capital at very high rates, and they all feel like this is existential, so maybe this is it. But the bear case, or the rational case, is: show me the money. You said there are going to be crazy gains from AI and massive improvements to the business. We'll be able to have massive efficiency gains by firing employees, or one employee being able to do the work of 10. But that just hasn't happened yet.
It hasn't happened at all. Best-kept secret, right?
At all. Show me the money. Show me where any company other than Anthropic and OpenAI has had massive revenue increases that weren't cloud. Cloud is a little bit scary because it's recursive. They invest in cloud, then they sign a deal with Anthropic or OpenAI, book it as revenue, and it's like, okay, yes, this is sort of revenue.
Yeah.
That is what we're paying attention to. As time progresses and there isn't much to show for it in the short term, markets will continue to punish companies for raising CapEx. That's something where you can just be patient. I'm not worried about missing a 7% move in the Nasdaq up.
Yep.
I'm worried about buying the -10% of the -30% dip. That's what I'm worried about.
I think that was really good perspective. My last one for you is the obvious elephant in the room: your views on crypto. You're still short Hyperliquid on the HyperDash, which is now labeled Flood. It's been an insane trade. When does that close? Do you think Bitcoin lows are in? Are you still optimistic? For a while, you were buying a ton of Bitcoin and tweeting about it. I don't know where you're at right now.
We still own Bitcoin. We only own Bitcoin, HYPE, and some stocks, but I shouldn't really talk about stocks because of regulations and stuff. There are other assets we're looking at that are interesting. The way we think about HYPE is the way we would think about any asset: how fast is it growing?
Yep.
If HYPE starts growing really fast, then we will buy more of it. If growth stalls, then we'll continue managing our risk. When you think about HYPE, I'm obviously up quite a bit on it. If I sell it, I have to pay quite a bit of tax. So I can hedge it with a derivative, protect my delta, and then I get paid on both sides because you get paid the HYPE staking yield, which is not taxable in the US, and you also get paid the funding payments because HYPE basis is typically paying shorts at mostly the crypto baseline funding.
Cool.
I get paid to hedge, and if I'm bearish, that feels like a pretty nice trade. It's very tax-efficient. I may be short HYPE, but that doesn't mean I'm bearish on HYPE. It might just be an okay use of my cash.
I own HYPE, and I'm short some against it. I have other wallets as well. It's not like I'm not psyched about making money. I didn't make $1 million shorting. It's just me protecting the US-dollar value of my portfolio and being paid to do so.
I'll tell you what, it's scary to see. Flood, you're the GOAT, man. This was an important emergency stream. You're one of the best, dude. I appreciate you coming on. I think it was a proper number 2 on a whim. Chat loves you, I love you. You're the GOAT, man. Thanks for coming on.
Thanks for having me.
All right, dude. See you soon, man. Have a good one. Peace.