I wasn’t planning to write today but the headlines on the screens in Spark Towers are certainly in ‘triggering’ territory. Where to start? Count Binface’s rival for the seat of Clacton-on-Sea looks like he’s going to have to explain another £2m donation from Posh George’s mum. Who needs parody scriptwriters these days, unless you want to be America’s top legal official. Apparently, Todd Blanche, as US Attorney General in waiting, can’t get in writing from his boss that he won’t deploy an illegal slush fund for January 6th felons or avail of a self-dealing tax waiver with the IRS. Thankfully, federal judge, Kathleen Williams, seemed to be more in tune with the law of the land when recently ruling the settlement between the Trump family and itself (ahem the US Government) “had no viable basis in law or fact”. In Todd we do trust. Then again, we thought 25 year old hedge fund superstar, Leopold Aschenbrenner, could do no wrong either. At the beginning of July his fund was worth almost $50 billion. Now, not so much, in a spectacular Wall Street wipeout and a dramatic reminder of the rocky road ahead for AI-related investments. And, it’s not just the young ones.
Only months ago any big tech name announcing hundreds of billions of AI infrastructure spending could be guaranteed its market value would rocket. Now, Mr. Market has become more discerning. Ask Google. It recently delivered a phenomenal set of quarterly results with cloud/AI revenues humming at 80% growth rates. However, a small increase of its 2026 AI capital expenditure plans from $195 billion to $205 billion seemed to spook investors and wipe $300 billion from its market value. In contrast, Microsoft decided to stick to its plans (with no increase) and benefitted from the largest daily increase in market value by any individual company in history. Yep, just the $450 billion increase in value for showing capital discipline. Confused? You should be. Amazon announced a $20 billion increase (to $220 billion)in capital expenditure on its AI infrastructure (including its own chips) and has been rewarded today with a 15% spike in its share price. For this writer, the huge volatility of market reactions in a week of multiple (probably) hedge fund implosions, indicates we are into increasingly risky AI investment territory. Here are a few contributory risk factors…
China: As we know, memory chip manufacturing is hot with massive AI demand potentially outstripping supply and generating staggering profits for the likes of Micron, Sandisk, Samsung and SK Hynix. Anyway, this week a small Chinese memory chip player, CXMT, listed as an IPO in Shanghai. The CXMT share price rocketed 466% on its first day of trading to surpass the mighty Intel Corp in market value ($480 billion) and become China’s most valuable company. Then later in the week, news stories emerged of China state backed companies developing cheaper versions of lithography machines, a sector dominated by Europe’s most valuable company, ASML. It’s early days yet but I sense we are going to have more “Deepseek” type shocks which will upset assumptions about industry dominance, and costs. One can’t ignore the lessons of the global solar, battery and electrical vehicle markets. All are utterly dominated now by enormous Chinese investment and economies of scale bringing production costs down by well over 90%.
Leverage: Charlie Munger used to say liquor, ladies or leverage is the usual undoing of smart people. Warren would say it’s just leverage really. Indeed, that was the 1998 lesson of the LTCM collapse brilliantly narrated by Roger Lowenstein in “When Genius Failed”. This week’s collapse of the ironically named, Situational Awareness fund, steered by AI wunderkind Leopold Aschenbrenner is just the latest chapter in leverage lunacy. Leopold had huge concentrated bets on AI stocks but 80% of the money invested in these stocks was borrowed money from Wall Street banks and prime brokers. All it took was a few South Korean butterfly wing flaps to set in motion a catastrophic series of events for the fund. Seoul’s stock market index, the KOSPI, lost 20% of its value in a few days thanks to huge exposure to national chip champions Samsung and SK Hynix. Some of Leopold’s fund stocks were down 25-30% which meant the banks and brokers’ funding was at risk of loss. Recall my perennial warning about ‘other people’s money’, and the risk of those people wanting their money back at exactly the wrong time(for the borrower). Of course, Wall Street got their money back and Leo’s portfolio holdings were sold at a firesale discount to hedge fund giant, Citadel. Timing is everything. Today the KOSPI rocketed 18% in one session with Samsung, Hynix and other AI wonder-stocks all flying upwards by 25-30%. Too late for Leopold. He should have an interesting groom’s speech at his wedding tomorrow. On a more serious note, leverage is now a growing factor in the AI capital/infrastructure spend story. Indeed, financial media publisher, Nikkei, highlighted in a recent article $1.65 trillion of AI infrastructure debt which does NOT appear on company balance sheets. That number is likely to be now over $2 trillion. But remember, it’s “other people’s money”.
The more money in play on or off balance sheet, the more volatility is likely to ensue. There will be other weeks like this one and it’s entirely likely it will be one of the following rocks where individual AI hopes will perish: technology competition, capital overspend, leverage, concentrated speculation or costs. We are still learning the same lessons it seems….
“For men who prided themselves on being disciples of reason, their drive to live on the edge seemed inexplicable, unless they believed that becoming the richest would certify them as also being the smartest.”
― Roger Lowenstein, When Genius Failed; The Rise and Fall Of Long-Term Capital Management