“I’ve got information man! New sh*t has come to light!” - The Big Lebowski

“When the facts change, I change my mind - what do you do, sir?” – John Maynard Keynes

One of the most confusing claims in finance and economics is that markets are efficient. In our minds, the word “efficient” produces an image of a system achieving maximum productivity with minimum wasted effort. Efficient systems don’t suddenly drop 10-30%, so if you went through the dotcom crash, 2008, or the Covid flash crash, “efficient” doesn’t sound like the stock market you know.

Instead, in economics, market efficiency has a two-part explanation:

Prices at any given moment reflect a reasonable understanding of all past, present, and future expectations of a company’s earnings.
Because prices reflect the consensus opinion of thousands of different analysts and investors, it is difficult to outperform the market’s opinion over long periods of time.

Market efficiency is not a claim that prices accurately reflect a company’s value from today until the end of time. It only claims prices reflect all available information today at this very moment. What changes prices is new information. And while market declines and volatility are uncomfortable, an even scarier world is when prices do not adjust to new information, because that world leads to shortages and the misallocation of capital. You can ask the remnants of the Soviet Union how good an idea that is. Thankfully, the first quarter of 2025 gives us a great lesson on the way prices change with new information and why a bias towards value stocks can provide some insulation during market volatility,

DeepSeek and NVIDIA

NVIDIA is a perfect example of how new information impacts the value of a company. In 2015, someone could have written a check for $18 billion to buy NVIDIA. Between 2022 and 2023, NVIDIA rose from a valuation of $364 billion to $1.2 trillion. At the end of last year, the company had a market cap of $3.4 trillion. What propelled this success? New information. In November of 2022 Open AI launched Chat GPT-3, the most advanced large language model AI at the time, and everyone learned that it was NVIDIA’s graphics processing units (GPUs) that were key to training the product. From 2022 until January 27, 2025, it was assumed that NVIDIA’s most expensive GPUs were the key to training even more advanced AI models.

On January 27 of this year, NVIDIA suffered the greatest one-day value wipeout in history. It lost $589 billion in market cap, going from a $3.5 trillion valuation to $2.9 trillion (NVIDIA sits at $2.5 trillion on April 3). What caused this incredible decline? New information. In late December, a Chinese company named DeepSeek had released their V3 AI model and people were discovering that it performed at nearly the same level as the most advanced American models (GPT-4, Llama, Claude, and Gemini). While OpenAI has hinted it cost more than $100 million to train GPT-4 on the most advanced NVIDIA GPUs, DeepSeek claimed V3 only cost $6 million to train and used less-advanced (and less expensive) NVIDIA chips. NVIDIA’s price adjusted to new information, as it should. What kind of world would we live in if facts changed, but prices didn’t?

Prices and Earnings

In my job, I’ve had the opportunity to provide financial due diligence for several small businesses that want to be bought or sold. When looking at small businesses, I look at the health of their earnings, future growth expectations, and how much a buyer would have to pay for each dollar of earnings. We call that comparison a price to earnings or P/E ratio. A P/E ratio of 6 means you are paying $6 dollars to the seller for every $1 of earnings. A P/E of 50 means you are paying $50 for every $1 of earnings. For publicly traded companies a P/E of 16-20 is considered normal. In summary, to justify paying a higher-than-average P/E ratio, you are saying (with your dollars) that you expect earnings to grow at a faster rate to justify the price you paid.

When examining different segments of the global markets, it is interesting to see which part of their investment return comes from fundamental returns (earnings and dividend growth) vs multiple expansion (paying more each dollar of earnings).

The above chart is from a paper released in February from the Dutch asset management company Robeco. The X axis shows the contribution of fundamental returns while the Y axis shows multiple expansion from 2015 through 2024. What you’ll immediately notice is that the best performing asset class of the last decade, large growth companies (think NVIDIA, Apple, Tesla, etc) have had both high fundamental returns and high multiple expansion. These are not your money losing dotcom era companies.

However, you’ll also notice that small and low volatility US stocks have had nearly as good or better earnings growth compared to their US large growth peers. Investors have simply decided they are willing to pay more per dollar of earnings in large US growth companies. In the case of small companies in Europe, not only have they had the highest earnings growth, but investors are paying less per dollar of earnings than they were in 2015!

This isn’t to say we don’t own large growth companies. We believe in owning the entire market because we don’t know which segments are going to outperform or for how long, regardless of if we think Tesla is going to sell a car to every man, woman, and child (probably the only way to justify its valuation). We just own a smaller percentage of companies subject to excess multiple expansion because future growth expectations are baked into their prices, and the future is always subject to the whims of new information. And we see what new information can do to prices.

The Magnificent Seven companies have been some of the greatest beneficiaries of multiple expansion. For the first quarter of 2025, these companies were down -14.8%. The other 493 companies in the S&P 500 are collectively +0.5%. And these numbers are not set in stone. They will change as new information (tariffs, anyone?) is brought to light.