France's bond market had investors on edge all week. Cristina Roca's note in Monday's Money & Investing says the thing they fear most for next year is something else. George Saravelos, who runs foreign-exchange research at Deutsche Bank, had just finished a week of client talks. France wasn't the answer. The answer was AI.
Saravelos summed up his clients' view in one line. The biggest systemic risk next year is something going wrong in the AI ecosystem. He named three ways it could happen. A safety event. A failed IPO. Disappointing revenues. Notice what's on that list. Nobody needs a crash in chips or a bubble popping on its own. One flop at the wrong moment could do it.
The size of the bill
He backed it with a number from a Brookings paper. AI labs need to earn upward of $3.5 trillion a year by 2032 to get a return on what's being spent on data centers. We wrote on Oct. 3 about a similar estimate that framed the bill as a share of GDP. The point is the same. The spending is real now, and the revenue is still a plan.
The week offered a small preview. On Tuesday, Oct. 13, Asian chip shares sold off. The trigger was a remark by the president of CXMT, one of China's biggest memory-chip makers. He told a Shanghai conference that it plans a new kind of DRAM, the working memory in computers and phones. South Korea's KOSPI fell 5.46%. Micron (MU) dropped $35.85 to $953.19 in New York. The week before, the PHLX Semiconductor index had already lost 4.30%. None of that was a safety event or a failed IPO. It was one comment at a conference.
Our read
This is Investments and Risk (IN04), and the risk is concentration. For most people with an index fund, the AI build-out is the biggest bet they own, whether they chose it or not, because it's the biggest piece of the index. Our books hold it at weight. Nvidia (NVDA) in 31 books, Microsoft (MSFT) in 29, Broadcom (AVGO) in 29, Alphabet (GOOGL) in 28, Taiwan Semiconductor (TSM) in 30. Nothing added this week. The September letter's three conditions for new money aren't met either. Through Tuesday, Oct. 13, only one of the three had passed. Saravelos's list is useful because it's specific. A safety event, a flop IPO, a revenue miss. Those are things you can watch for, and none of them announces itself in advance.
The check for your own plan is inside your biggest fund. Open your 401(k) or IRA statement, find the largest holding, and look up its top 10 stocks. In an S&P 500 fund, a handful of AI names now make up a big share of the whole, and most target-date funds hold the same index underneath. Add up that share and ask whether you'd be fine seeing it marked down on one bad IPO. If the number surprises you, that's a conversation for a calm week, and this is one. An umbrella is cheap on a sunny day.