--- title: "Leopold Aschenbrenner sells entire public book to Citadel in global markets; US$16bn deal to meet margin call" sdDatePublished: "2026-08-10T11:29:00Z" source: "https://www.schroders.com/en-au/au/individual/insights/srrf-commentary-when-the-chips-are-down/" topics: - name: "financial service" identifier: "medtop:20001370" locations: - "China" - "Japan" - "Germany" - "Taiwan" - "Australia" - "Iran" - "United States" --- Leopold Aschenbrenner sells entire public book to Citadel in global markets; US$16bn deal to meet margin call SRRF Commentary: When the chips are down SRRF Commentary: When the chips are down Higher oil prices and higher rate uncertainty led to high anxiety as equity market darlings started crashing lower. Is this just a technical wash out or the signs of something more sinister? The leveraged play on the AI buildout took a brutal hit in July, with the NYSE Semiconductor index down over 21% in US dollar terms, pushing AI heavy indices like Korea down 22%, the Nikkei down over 8% and Taiwan and the Nasdaq down over 6.5% over the month in local currency terms. Investors who had recently invested in levered ETFs, or even took out margin debt to participate in the parabolic rally, felt the brunt once this trend reversed sharply. According to Goldman Sachs, over 1.2 million retail investors in Korea, or roughly 3.5% of the adult population, received a margin call. But perhaps most publicly, the 25-year-old founder of hedge fund Situational Awareness, Leopold Aschenbrenner, was forced to sell the entirety of his public book to Ken Griffin’s Citadel (approximately US$16bn worth) to meet a margin call after his fund lost 78% in July. Given the selloff was predominantly isolated to the chips trade, emerging markets suffered the most, down over 3% for the month. Global equities and the overall US S&P 500 were mostly unchanged, thanks to a 2% rally in the Magnificent Seven, driven by strong earnings from Microsoft, Amazon and Apple, which stemmed the rout by the end of the month. Most of this was based on the belief that the hyperscalers may have to pull back on capital expenditure due to questions over AI profitability, given a fall in token usage and cheaper Chinese open-weight competitors like Kimi showing impressive performance relative to the US AI heavyweights. These fears were somewhat alleviated after the hyperscalers produced bumper profits and even higher-than-expected capital expenditure plans. But perhaps a simpler and overlooked cause of the crash was the rise in government bond yields. US 30-year bond yields increased over 32 basis points this month, with overall yields close to 5.3%. The US Federal Reserve (Fed) kept rates on hold, but three members dissented in favour of an immediate hike, the most one-directional dissent the Fed has seen since 2016. Fed Chair Kevin Warsh offered no forward guidance after this month’s meeting and said he “welcomes” higher long-term Treasury yields. This uncertainty caused bonds to sell-off, not just in the US, but around the world, with most 10-year treasuries yields moving between 20 and30 basis points higher, other than Japan and Australia, which were more insulated. Credit remained fairly subdued, with spreads mostly unchanged. The US dollar fell, predominantly through a strong Japanese yen as the Bank of Japan intervened in the currency at the end of the month. Watch the latest portfolio update video For a brief moment, the market stopped focusing on the war in the Middle East and instead shifted its attention to the war of tokens. The last few months have felt euphoric as the rolling AI bubble moved into new and untouched sections of the AI supply chain. The AI trade started with the hyperscalers, with capital pouring into cloud monopolies and those most likely to see future revenues from the foundational model or application layer of AI. This then rotated into high-end semiconductor designers like Nvidia, who saw immediate revenue from their GPUs which compute the complex AI algorithms. Most recently, this expanded into memory chips, with companies like Samsung benefiting from core infrastructure needs, with specialised memory chips being used to keep the datacentres working. Extreme capital expenditure from the hyperscalers has seen free cash flow plummet, but this flows directly to the hardware manufacturers required in the AI supply chain, boosting their profits to levels never seen before. This led to the semiconductor index and countries linked to chips, like Korea, rallying around 100% this year to June. But in July, this whole premise came under question. July had another ‘DeepSeek moment’, when Kimi K3, a Chinese open-weight AI from Moonshot, scored exceptionally high on benchmark tests, suggesting that low-cost Chinese competitors were quickly closing the gap with the American AI heavyweights OpenAI and Anthropic. What most surprised markets was how Kimi could perform as well as the premier US AI models, but for a tenth of the cost. Even worse, Kimi is “open-weight”, which in AI terms is not fully open source, but essentially gives away the most important part for free, allowing users to use the AI locally on their machines and customise as needed. This is a direct threat to Silicon Valley’s closed-source proprietary business models. This made many wonder, if AI can become commoditised, will all these trillion-dollar capital expenditure plans from the hyperscalers be rewarded? And if not, any pullback in spending would have immediate direct downstream implications for chip providers and the economy as a whole. Chart 1: Artificial Intelligence Index Score vs Cost Per Task Source: Artificial Analysis, JPMAM, July 2026 As the market worked through these consequences, a US $1.3 trillion sell-off ensued. What exacerbated the sell-off was positioning. We’ve mentioned for a while that the market is overly stretched. Margin debt is through the roof, newly issued three-times levered ETFs hit the market almost daily, and positioning in technology equities hit almost 6 standard deviations overbought. Coupled with this was a resumption in hostilities in Iran, sending the oil price higher, right when the Fed shifted more hawkish and removed forward guidance. The spike in bond yields undoubtedly accelerated the unwind. But, once the panic subsided and positioning washed out (US technology positioning now a more ‘normal’ 2.4 standard deviations over-owned), the market found its footing. Hyperscaler earnings saw AI spend beat expectations, reviving the chip trade, but perhaps more importantly, earnings for Microsoft, Amazon and Apple smashed expectations. Despite significantly higher capex announcements, earnings continued to beat expectations, with Google’s Cloud revenues beating expectations by 17% and delivering 82% growth. When the chips were down, the hyperscalers didn’t fold. The AI trade lives to fight another day. Chart 2: Quarterly Cloud Growth Revenue Surprise Looking more broadly, we’ve seen US growth disappoint, with GDP coming in at 1.5%. However, this was mostly due to trade and inventories. Under the surface, personal consumption beat expectations, coming in at a very healthy 3.2%, showing resilience in the underlying economy. The Atlanta Fed’s GDPNow is estimating over 6% GDP growth next quarter once trade and inventory distortions revert. While this is too optimistic, it shows a clear recovery in consumption. This rebound in consumption is happening with a backdrop of weaker than expected inflation. June core PCE eased to 3.3% year-on-year, confirming weaker than expected producer price index (PPI). The Federal Reserve Bank of Dallas Fed’s Trimmed Mean PCE (Warsh’s preferred inflation metric) was 2.2% on a 12-month basis and services inflation was 2.3% year-on-year. Could we be shifting back towards a more ‘goldilocks’ scenario as opposed to a stagflationary one? Chart 3: US GDP growth supported by private consumption, while inflation starts to soften Source: Schroders, LSEG, Bloomberg, August 2026 Currently our global wave indicator remains in expansion, while our recession model shows no signs of concern. Earnings continue to beat expectations, with forward expectations in line with our top-down earnings model, and earnings momentum is turning higher across the globe. Growth remains robust. The job market is not collapsing and nor hot enough to justify higher wages. Inflation appears to be subsiding and the market has already priced in a number of hikes across the world. On the less positive side, tensions in the Middle East have reignited, policy uncertainty from the US Federal Reserve demands higher term premia, and it will be near impossible to predict when hyperscalers will pull back on their capital expenditure. Given AI spend is now approximately 1.5% of US Gross Domestic Product (GDP), any pullback will not only affect the chip makers but could also see economic growth slow as a consequence. We started the month moving more cautious, looking to defend the downside, but as the month came to a close, we started buying the dip. This saw our overall equity allocation end the month 1% higher on a 35% delta-adjusted basis. We started the month by selling 1% from emerging markets given the region’s strong performance the prior month and rotated this into the US S&P equal weight, which has a far lower exposure to technology. We then started reducing equities overall by 2% across the board as the market started to roll over. Our S&P put options started to kick in, reducing our equity exposure while volatility started to rise. By the end of the month, we started rebuilding our position back but pivoting towards some of the more beaten-up areas of the market. This included buying back emerging markets, shifting our Japan exposure from Topix to the more technology heavy Nikkei, and switching 1% from the S&P 500 towards the Nasdaq. While we may not be out of the woods, we believe the fundamental story for risk assets is still intact and the positioning washout in technology companies has mostly played out. We maintain our 5% notional 7300 S&P put option to protect the downside in case we are early in our re-risking. The fund increased its credit allocation by 2% over the month, which came from increasing our high yield exposure. Credit spreads are tight across the board. While both high yield and investment grade are extremely expensive, we prefer high yield to investment grade purely from a carry perspective. We added 2% to global high yield by taking profit on our credit default swap protection as spreads widened by 20 basis points in the US and 30 basis points in Europe. We also took profit on our European investment grade credit default swaps after spreads moved out 5 basis points, funding this 2% increase by reducing our allocation to Australian investment grade credit by the same amount. This does not change our overall view but reflects a modest diversification of our allocation after some modest spread volatility. Australian corporates and US securitised credit remain relatively attractive. Hyperscaler issuance has seen a drop in demand from investors, pushing spreads higher on these names. Off-balance sheet debt continues to raise eyebrows. Within high yield, we prefer Australian higher yielding corporates, which are typically either floating rate subordinated issues or BBB infrastructure and utility names. We prefer US high yield to European high yield given the more positive outlook on the US cycle and the potential for higher rates from the ECB at affect European corporates. While spreads are also uninspiring, we do continue to like emerging market corporate bonds, where we also see some relative value, lower duration and better technicals versus developed markets. The recent repricing of rate expectations combined with weaker than expected inflation prints has led us to increase our duration position. We used the volatility in July to increase duration by 0.4 years to end the month with 2.5 years of duration at the fund level. Softer-than-expected US core PCE combined with weaker-than-expected producer price index (PPI) data, makes us want to be long US front-end bonds, although this is being offset by higher oil prices due to the resumption of hostilities in Iran. Front-end yields have been closely tied to oil, particularly in Europe given gas-price sensitivity and the ECB’s inflation-focused mandate. In the US, the Fed’s reaction function is uncertain, but inf