Disruptive new uses cases pop up every day for Artificial intelligence, making it the topic of conversation in the stock market. This has had a dramatic impact on $NVDA, which specializes in processors that power new AI applications.

In the past, I have speculated that other companies like $TSLA couldn't sustain growth, and I was wrong. So my opinion that $NVDA will disappoint probably isn't worth much.

Instead, I want to talk about how paying a premium multiple can be consistent with a value-oriented investment approach.

Many value investors focus exclusively on companies that are trading at a low multiple of sales or earnings. But you could look at it a different way;

Instead, simply look to buy businesses for less than they're worth.

Often, these businesses will be unknown or unloved and will therefore trade at low multiples. But sometimes, a stock can be a bargain, even if it trades at a lofty price relative to its current performance.

Consider a company like Costco $COST. Investors could have paid a healthy 40x trailing earnings for Costco at any point since 2002 and still earned at least a 10% compounded return. Or look at Amazon $AMZN and Booking Holdings $BKNG, where early investors could have paid 40 times sales and still achieved attractive long-term returns.

It's tempting to look at those success stories and conclude that quality is the only factor that matters for long-term investors. But these companies are the rare exception— not the rule. Amazon and Booking both endured brutal periods when their share prices plummeted more than 90%. And for every Amazon and Booking, there are dozens of examples in which paying a premium price produced poor performance.

One potential analog for Nvidia is Cisco Systems $CSCO, a company whose hardware enabled the creation of the internet. Like Nvidia, Cisco's valuation surged in a short period of time, ultimately reaching a peak of 36.8x trailing sales in early 2000. But demand for Cisco's networking products declined during the dot-com bust, and the company faced more competition than investors anticipated. Cisco has proven to be a strong and durable business that produces plenty of free cash flow, but its share price remains well below those dot-com-era levels even more than two decades later.

It's not sufficient for a business to have strong competitive advantages, sound financials, and a shareholder-friendly management team. Stocks must also be attractively priced relative to a conservative forecast of their future performance. You must invest with a margin of safety in order to limit losses when your investment thesis doesn't play out as you'd expect.

It's entirely possible that Nvidia will prove to be a market-beating investment from current prices. But buying a company at 37x sales requires quite a lot to go right for quite a while. Any slip-up in execution, increase in competition, or shift in geopolitical conditions, and the stock will likely get slammed. I'd rather be conservative in terms of both the modeling assumptions and the price.