It was somewhat surprising when Warren Buffett's conglomerate Berkshire Hathaway (BRK.A -1.12%) (BRK.B -0.86%) purchased its first semiconductor stock, Taiwan Semiconductor Manufacturing (TSM -0.37%), last summer. Yet it was even more surprising to learn that Berkshire has already cut its TSMC stake significantly, lowering its holding by 86% in the fourth quarter -- just one quarter after purchasing the stock.

TSMC seemed like a solid bet for Berkshire. As the largest outsourced foundry in the world, TSMC makes high profit margins and seems set to benefit from the growth of semiconductors over the long term. So, why might the Oracle of Omaha have already decided to sell the stock?

Taiwan Semi is up 22% year to date

It's a bit confusing that Berkshire sold its stock so soon. While there is a large amount of macroeconomic uncertainty, which would affect all semiconductor stocks, Buffett is notable for taking a longer-term view. With sentiment around the economy slightly improving since Q4, Berkshire has already missed out on TSMC's 22% year-to-date gains, including dividends.

Of course, the stock market can move in mysterious ways, and Buffett has long said it's impossible to predict the movements of stocks in the near term. But that makes the sale even harder to understand. After all, TSMC has a competitive lead in advanced node semiconductor manufacturing. And while the current semiconductor downturn has caused most stocks in the sector to fall in value, the semiconductor industry should grow above the level of gross domestic product over the long term.

A high-margin leader in a growth industry -- and one trading at just 14 times earnings -- seems right up Warren Buffett's alley: a wonderful company selling for a fair price.

So, why might Berkshire have sold its stake?

Buffett may have been a reluctant buyer in the first place

Since Berkshire had gone more than 50 years without buying a semiconductor stock, it's a pretty safe bet that Warren Buffett and Vice Chairman Charlie Munger don't like, or at least aren't totally comfortable with, the semiconductor industry in general. Nevertheless, TSMC stock's decline to a low-teens valuation, along with its strong market position and ability to pass along price increases to customers -- even customers as strong as Apple, Berkshire's largest holding -- was likely what piqued Buffett's interest last summer.

So what might have changed? I think there are two possibilities.

Warren Buffett looking to the right without smiling.

Image source: The Motley Fool.

Risks to the TSMC thesis

First, Buffett might not have been totally assured of TSMC's strong competitive advantages over the longer term. After all, both Samsung and Intel (INTC 1.11%) have announced ambitious plans to build leading-edge fab capabilities and catch up to TSMC in the coming years. For instance, Intel has outlined a plan to go through five chip nodes in four years -- an incredibly fast pace -- and overtake TSMC in that time.

Given the severe headwinds Intel is experiencing due to its loss of leading-edge superiority and exposure to the PC market, I think it's somewhat doubtful that Intel will be able to overtake TSMC in just a few years. However, it's still possible, especially under Intel's energetic new CEO, Pat Gelsinger. Since Buffett generally likes near-sure things, uncertainty over the competitive prowesses of Samsung and Intel may have given him pause.

Second -- and perhaps most important -- TSMC's business is fairly capital-intensive. Although the company has generated high profit margins and has been able to raise prices to customers, TSMC must spend a lot on capital expenditures to expand capacity and keep its technological lead.

Charlie Munger raised this point at the recent Daily Journal Corporation annual meeting on February 15. When asked about the semiconductor industry, he replied:

The semiconductor industry is a very peculiar industry. In [the] semiconductor industry, you have to take all the money you've made, and with each new generation of chips, you throw in all the money you previously made. So it's compulsory investment of everything you want to stay in the game. Naturally, I hate a business like that. ... [N]ow if you're now ahead of it, like Taiwan Semiconductor is, that may be a good buy at these prices. It's not at all clear to me that they're not going to succeed mightily. ... But it's a difficult business and requir[es] everybody to keep increasing the bets on and on with all the money.

This may be getting to the heart of the matter, and it's why, back in November, I wrote that Buffett might prefer the semiconductor capital equipment stocks to TSMC. While semiconductor equipment stocks don't have high capital spending needs, they still have exposure to the growth of the chip industry and generally trade at reasonable earnings multiples like TSMC.

TSMC should still do fine

Despite these legitimate concerns, I still think TSMC will do well. The company has a lot of ingrained semiconductor manufacturing expertise, especially on the leading edge, which is very difficult to achieve.

That being said, it may not be a total cinch, so TSMC investors should monitor the progress of competitors Intel and Samsung. If these competitors experience technological breakthroughs and begin landing more large, leading-edge customers, it may be time to get more concerned.

However, that likely wouldn't occur for a few years, if at all, as each competitor has multiyear plans to catch up, and TSMC itself isn't standing still in this race.