Image source: Amazon.

According to legendary investor Warren Buffett, a company should only repurchase its own shares when it "finds its stock selling in the market below its intrinsic value, conservatively calculated." (AMZN -0.68%), the e-commerce leader that has seen its stock tumble 25% so far this year, recently announced a brand-new $5 billion share buyback program, stating in an SEC filing that it would purchase shares "when it believes that doing so would enhance long-term shareholder value."

I would argue that the majority of share buyback programs are a mistake, and the reason is that pesky "conservatively calculated" qualifier in Buffett's statement. A company overpaying for its own shares has a net negative effect on long-term shareholder value, despite any subsequent boost in per-share earnings.

Show me the value
The most bullish Amazon investors believe the stock is undervalued based on assumptions about future growth. The most optimistic scenario involves Amazon dominating e-commerce, cloud computing, logistics, and any other businesses the company eventually enters. Under that scenario, it's not difficult to come up with a lofty price target for Amazon stock.

Using more conservative assumptions, however, it's clear that Amazon's stock is extremely expensive. Amazon currently trades for about 400 times GAAP earnings, and while the company generated $7.3 billion of free cash flow in 2015, putting the stock at "just" 33 times this number, Amazon's reported free cash flow is greatly inflated by its use of capital leases to finance its expansion. In a nutshell, capital leases allow Amazon to depreciate assets, which boosts operating cash flow, without counting the purchase of those assets as a capital expenditure. In 2015, Amazon's adjusted free cash flow was about $2.5 billion, an improvement compared to 2014, but still good for a P/FCF ratio of about 100.

There's another wrinkle in all of this: stock-based compensation. Amazon, like many technology companies, issues employees and executives stock as a form of compensation. Stock-based compensation expense is deducted from GAAP earnings, but because it's a non-cash expense, it gets added right back when calculating the free cash flow.

It won't even offset dilution
The problem with Amazon's share buyback program, beyond the high price that the company is paying for its own shares, is that all of that money is going toward simply counteracting the dilutive effects of stock-based compensation. Amazon would need to spend all $5 billion very quickly, over the course of a couple of years, for the total share count to actually decrease. Any slower than that, and Amazon's share count will continue to rise.

This buyback program is not an effort to return capital to shareholders; it's an effort to manage dilution caused by stock-based compensation. Here, another quirk in how free cash flow is calculated becomes important: If Amazon is issuing stock to its employees, and then turning around and buying that stock back, shouldn't this stock-based compensation be treated as a cash expense? Amazon's buyback program effectively turns a non-cash expense into a cash expense, but since spending on buybacks gets filed under the financing section of the cash flow statement, it has no effect whatsoever on the free cash flow.

I've always been of the opinion that stock-based compensation should be universally treated as a real expense, and when I evaluate a company, I generally exclude it from my calculation of free cash flow. In many cases, this item is so small relative to the free cash flow that it doesn't have much of an effect. But in Amazon's case, stock-based compensation represents the vast majority of its adjusted free cash flow. With cash being used to undo the dilutive effects of stock-based compensation, any argument for treating it like a non-cash expense seems to go out the window.

During 2015, Amazon booked a $2.12 billion stock-based compensation charge. Treating this as a cash expense and backing it out, Amazon's adjusted free cash flow falls to just $374 million, for a P/FCF ratio of 650. I'd like to hear the scenario that justifies that valuation.