Skip to content
← Back to Active Thesis Feed
⚪ New Oct 5, 2026

This Looks More Like 1999 Than 2008

The 10 year just hit a 24 year high and tech rallied anyway. A strategist explained it this way: if AI's benefits are essentially infinite, so what if borrowing costs more? That argument has been made before, almost word for word, in 1999. The comparison that fits this market isn't the 2008 crash. It's the year before the 2000 top.

Oct 5: the US 10 year yield hit a fresh 24 year high of 5.349% and the 30 year reached 5.696%, its highest in 24 years. Tech stocks rallied the same session. Interactive Brokers strategist Steve Sosnick: "If you think the benefits of AI are essentially infinite, so what if you have to pay more to borrow money?" The setup now: long yields have risen for months, the Fed hiked in September, roughly three quarters of S&P 500 stocks fell in September while a handful of AI leaders held the index near its record, and chipmakers are financing their own customers, with Broadcom agreeing to lend Anthropic up to $42 billion to lease Broadcom's chips. The setup then, from the historical record: the 10 year rose from about 4.7% in early 1999 to about 6.7% by January 2000. The Fed hiked six times between June 1999 and May 2000. The Nasdaq nearly doubled in 1999 while more S&P 500 stocks fell than rose. Telecom equipment makers including Lucent, Nortel and Cisco lent billions to their own customers so those customers could keep buying equipment. The Nasdaq peaked on March 10, 2000 and fell about 78% over the next two and a half years.
View source ↗ 2026-10-05
The 2008 comparison gets made because credit is cracking underneath a market near its highs, and that part does rhyme. But in 2007 and 2008, bonds rallied as the crisis built. The 10 year fell from about 5.3% to the mid 3s. That is not what is happening now. What 2008 cannot explain is stocks at records while long bonds collapse for months. 1999 can. The pieces line up closely. Rising yields ignored by a narrow group of tech leaders. A Fed tightening into a boom. A new technology treated as a reason that interest rates no longer matter. Suppliers lending to customers so the orders keep coming, which in 1999 was called vendor financing and ended with Lucent and Nortel writing off billions when the customers couldn't pay. How 2000 actually broke matters for what to watch. There was no single crash day like Lehman. Rates kept rising, the Fed drained the extra liquidity it had added for Y2K, and customers who had been financed to buy equipment stopped buying. The top came from the cost of money finally outweighing the story, and the earnings that had been pulled forward by financing running out. There are two big differences, and both cut toward more risk, not less. In 1999 the US ran a budget surplus. Today the deficit is near 6% of GDP and the President has said inflation could pay down the debt. And in 2000 and 2001 inflation was falling, which let the Fed cut aggressively once things broke. Today input prices are accelerating in both ISM surveys and an energy shock is still working through the system, so the rescue would be slower and more constrained.
  • Late October megacap earnings: the 2000 break came when financed demand stopped. Any capex guidance cut or slowing AI revenue from the leaders is the equivalent signal.
  • Vendor financing disclosures: whether more chipmakers or cloud providers lend to their own customers, and on what terms.
  • Oct 7 and Oct 8: 10 year and 30 year reopenings. Oct 14: CPI. Oct 27 and 28: FOMC.
  • Nov 4: Treasury refunding, the one lever that could change the rate backdrop.
  • Falsifier: AI revenue grows into the spending, breadth broadens, and long yields stabilize without a break. That would make this 1995 to 1997, a boom with years left, not 1999.
⚡ The market's defining feature, AI equities at records while long bonds collapse, fits the 1999 to 2000 setup better than 2008, with a weaker fiscal position and less room for a rescue.