•••

Shares in NVIDIA (NASDAQ: $NVDA — $955bn), a computer chip maker, have boomed. They're up 170% this year and have crushed almost every other stock. Investors think artificial intelligence (AI) will change the world and want a slice. They reckon NVIDIA will make the hardware that powers AI, and they've bid hand-over-fist to get in. However, some punters wonder whether the rally is overdone. Has the price risen too high? The answer is yes. Investors have overvalued the company and will get lacklustre returns from here.

First, despite ambitious growth and profitability expectations, the shares are expensive. Fifty analysts cover NVIDIA and expect the top line to grow five-fold in ten years. They also think margins will expand to 55% from their already-juicy 35% level. If they're right, NVIDIA would be the world's fastest-growing billion-dollar-plus public semiconductor company. It would also have the widest profit margins. It's plain to see that analysts are confident about the firm's future.

Using those assumptions, a discounted cash flow model suggests each share is worth $110-130. That is far below the $390 level the shares trade at, meaning they could drop by 70%. Even if this already rosy vision for the future turns out to be conservative, the shares are still overvalued.

Second, equity investors have baked a 5-6% annual return into the current price. That is far too low to compensate them for the risks they face. The yearly expected return is so small it's about half the 11% minimum return investors should demand of a company like NVIDIA. Since NVIDIA's cash flows are riskier than the market, equity investors should expect a higher return. By way of comparison, they can get almost 9% by buying the S&P 500 at current prices.

Moreover, imagine you bought ten-year US government bonds. In that case, you would get almost 4% per year. The semiconductor business is cyclical, discretionary, and has high fixed costs. Given these risks, an extra two percentage points per year above the risk-free rate is insufficient.

Third, the shares have a price-to-sales (PS) ratio comparable to dotcom tech stocks. In 2000, speculative tech IPOs traded at about 40x sales. But those companies were small and had quick-growing sales. NVIDIA doesn't fit this profile as it's a big company that already has $26bn in sales.

Moreover, NVIDIA's ratio is the highest of its peers and 17 times higher than the median semiconductor company. If the company paid all its profits as dividends, it would take three lifetimes to get your money back. But it would only take 18 years for the average semiconductor stock.

Yet, artificial intelligence, and the industries that serve it, have huge potential. The history of technology is littered with the bodies of naysayers and sceptics. However, it is unclear how big the market will get or who the winners will be. While some bold predictors will be correct, most will be wrong. Fewer still will get adequate compensation for the risk they take. But that doesn't even matter in this case. Even if you assume NVIDIA takes over, investors will get poor returns at the current valuation.