One Tenth of One Percent

On Wednesday morning the government reported that core inflation rose 0.3% in September. Economists expected 0.2%. The difference is one tenth of one percent, a rounding error inside a survey with a margin of error wider than the number itself.

The market treated it like a regime change. The Nasdaq 100 fell 2%. The S&P 500 fell 1.1%. Small caps fell almost 2%. The 10-year Treasury yield spiked to 5.33%, and the odds of a rate hike at the Federal Reserve meeting later this month jumped from 22% to 39% before lunch. By the closing bell, roughly $1 trillion of US market value was gone. Erased by one tenth of one percent. And the market that gets paid to price inflation barely moved. The 10-year TIPS breakeven rate, the bond market's expectation for inflation over the next decade, closed at 2.35%, two basis points higher on the day and almost exactly where it stood a year ago. To size that number: $1 trillion is the entire stock market of Brazil. It is a fifth of India's, and a third of France's and a third of Germany's.

Then Asia opened on Thursday. South Korea's Kospi fell 7.6%, and trading was halted when the drop hit 8%. Samsung fell 9.5%. SK Hynix fell 11%. Japan's Nikkei fell 4.9%. Taiwan fell 3.4%. Wednesday's losses had crossed the Pacific.

New York opened in stress. Micron hit a low of $858 in the first ten minutes. Then the 10-year yield collapsed toward 5.03%, down from 5.33% at Wednesday's peak. The fear had flipped. Investors were buying the bonds they had sold the day before. By mid morning the market realized the fallacy and rallied, taking back most of the day's losses. The broad market came back. Dell did not. It fell 11% on Wednesday and around 9% on Thursday.

And a hike would aim at the wrong target anyway. The 10-year Treasury already tightened for the Fed, spiking to 5.33% on the print. What is left of this inflation is supply: tariffs at the border and diesel at the pump. Raising the fed funds rate will not cure supply shock inflation. It does not drill wells, build factories, or move freight. It only makes demand pay for a supply problem.

Now consider what the economy actually did this week.

Goldman Sachs reported earnings of $17.01 a share. Analysts expected about $13. It was the second-best quarter in the firm's history, with equities revenue up 72%, and the result came after a $450 million charitable contribution that reduced earnings by about a dollar a share. JPMorgan earned $15.3 billion, up 6%, and its credit portfolios show borrowers paying on time. Citigroup posted its best third-quarter revenue in a decade. The American consumer, whose collapse has been forecast for three straight years, keeps paying his bills.

Or take Dell. The stock fell 11% on Wednesday, its worst day in more than a year, after an analyst downgraded it on valuation. A day earlier, Michael Dell was on television saying that demand for AI infrastructure exceeds supply and will through 2030, and that memory stays tight until 2030 because a new chip factory takes four years to build. JPMorgan reported this week that three quarters of Micron's fiscal 2027 memory output is already sold. Micron closed at $914 on Wednesday, down 16% in five sessions, in the $900s. The companies that buy memory are warning about scarcity. The market spent Wednesday selling the companies that sell it, on a theory that prices are about to normalize.

This was never about value. The market did not reprice earnings on Wednesday. It repriced positioning. Money was crowded on one side of the boat, volatility was cheap, and the index sat at records, so a tenth of a percent started a rush for the same exit. None of this tells you the economy weakened. It tells you a lot of people were leaning the same way when the print landed one tenth higher than expected, and leaning is what moves a trillion dollars in a day.

I have invested through enough of these days to know the rule that survives them. Watch what people do, not what they say. Companies are still building. Consumers are still paying. The builders of the AI economy cannot manufacture memory fast enough to meet demand. Earnings season, which just opened with the strongest bank numbers in years, will spend the next six weeks reporting what actually happened. Inflation surveys will keep printing tenths.

And all the while, the market is getting cheaper. S&P 500 earnings are expected to grow 29.5% this quarter, the third straight quarter above 25%. The index itself is up 16% over the past year. When profits grow faster than prices, the multiple shrinks. The forward price-to-earnings ratio has fallen from 22.9 a year ago to 19.0, below both its five-year and ten-year averages. The market sits at records and costs less per dollar of earnings than it did last year.

And earnings season is just starting. Six weeks of reports will now test every story the market told itself on Wednesday. A selloff into that, on a tenth of a percent, is a gift.

Benjamin Graham said the market is a voting machine in the short run and a weighing machine in the long run. Wednesday was the vote. The next six weeks are the weighing.

Buffett's version is shorter. Liked it at $100, love it at $90.

The analysts are voting the other way. In a typical quarter they cut estimates, by 2.2% on average over the past five years. This quarter they raised them: the S&P 500 earnings estimate rose 1.4% from June to September, a record 72 companies guided higher, and full year estimates rose 6.2%. The tenth did not change that either.

After the close, Anthropic reported third quarter revenue of $16.8B, up from $11.5B in Q2 and $4.8B in Q1, with adjusted profitability, and plans to spend more than $500B in capex over the next decade. One more quarter like that puts the run rate near $100B. The percentage slowed because the base got huge. The dollars did not.

Then Friday extended it. The chip names fell again, and the semiconductor selloff has now erased about $2 trillion. The selling stopped sorting. CoreWeave, the most leveraged of the AI builders, fell about 12% on Friday and about a quarter in five sessions, and its bonds now yield over 15%. This is a company carrying about $50 billion of debt, with a contracted backlog above $100 billion and customers still waiting for compute it cannot deliver fast enough. The market is throwing the baby out with the bathwater.

None of it changed the earnings. ASML still cannot meet demand for two years. The capital spending continues, and as long as it converts into earnings, the framework holds. A tenth of a percent started the rush for the exit. For long term investors, weeks like this are where the opportunities are.