Taiwan Semiconductor Manufacturing (TSM) , or TSMC, is the world's largest independent semiconductor foundry and is the lifeblood of much of the mobile semiconductor industry. Big names like Qualcomm (QCOM) , NVIDIA (NVDA) , and Broadcom all rely on some degree -- often to a large degree -- on TSMC to build their chips. The company is doing extremely well, but for investors looking out to 2015 and beyond, the question of whether Intel is a threat needs to be carefully considered.
The broader stage
The long case for TSMC is actually pretty simple: The mobile computing and network/cloud infrastructure booms are driving an insatiable demand for more complex semiconductor devices. Whether you have feature phones being replaced with low-end smartphones packed with modems, apps processors, and increasingly sophisticated connectivity, or whether there's growth in large, high-ASP, application-specific integrated circuits, TSMC is well-positioned to benefit from this growth trend.
One could argue that there is foundry competition -- indeed, there always has been. But the capital spending needed to build just one leading-edge factory -- and the research and development required for the manufacturing processes to stay at the cutting edge -- is quite massive. To put things into perspective, TSMC allocated $10 billion to capital expenditures during 2013 and spent about $1.65 billion in R&D -- and that latter number is growing.
Putting this into perspective
According to IC Insights, Global Foundries -- the world's second largest foundry -- did $4.26 billion in semiconductor sales during 2013. Samsung and United Microelectronics (UMC) each did approximately $3.95 billion in sales.
TSMC simply dwarfs these other players as far as sales go, which in turn allows it to spend more on advanced R&D while remaining profitable. If Samsung, for example, were to spend $1.5 billion a year in semiconductor process R&D, it would need $7 billion in sales just to break even. Samsung is backed by an extremely profitable memory business and a consumer-devices business, but it's not going to waste its money on unprofitable ventures.
The more interesting question is where Intel (INTC) fits into all of this. Intel does about $50 billion in semiconductor sales per year, making it the only company on the planet that sells more semiconductor logic, by dollars, than Taiwan Semiconductor. Intel also spends more than $2 billion per year on manufacturing technology development and has indicated that it intends to continue to push it. Intel has also expressed its interest in being a foundry, and at its investor meeting, Intel was unashamedly bullish about its prospects and very firm in its commitment to this business long-term.
Intel is probably an indirect threat -- but it's not to be ignored
While Intel has been rather vocal about pursuing a broader set of customers, it is unlikely that the company will build chips for direct competitors that can use the leading edge, i.e. Qualcomm. It also doesn't seem likely that Intel would be too keen to build GPUs for someone like NVIDIA, whose GPUs double as competitors to Intel's HPC accelerators known as the Xeon Phi. A company like Broadcom, which does heavy-duty network processors and switches, is also in direct competition with Intel's most profitable and prized business -- the data-center group.
So, these customers -- the ones that really do make up the bulk of TSMC's business -- are probably not going to be a good fit for Intel's foundry. However, the kinds of companies that would be a good fit for Intel are the following:
- Companies like Cisco, Juniper, and Brocade that do custom ASICs for routers/switches
- FPGAs -- Intel is already working with one of the leading players in this space, Altera (ALTR)
- Game-console chips -- although as long as AMD has this contract, this is unlikely
Basically, companies that sell big, expensive chips that wouldn't mind a slight cost increase if it translates into pricing power are perfect candidates for Intel -- not TSMC's main bread-and-butter chips.
However, Intel is an indirect threat to TSMC. In a situation where Intel gains meaningful modem share at 14nm and meaningful apps processor share from Qualcomm, TSMC's revenue on the apps processor and modem side of things will be negatively affected. Potentially countering that is the secular growth of the underlying markets, a potential addition of Apple to the top-line story, and the fact that TSMC builds many of the other, less sexy components inside of these handsets, even if Intel grabs the more lucrative apps processor/modem spots.
While it was easy to be unequivocally bullish on TSMC at $10 and even at $16, it's a lot harder to feel comfortable with the shares now that they're in the $19-20 range. It's not likely that the Intel threat will really kick in until 2016 -- during 2015, Intel will still be building its highest-volume mobile chips at TSMC -- but this is something to be mindful of now that the stock has seen very real multiple expansion since the smartphone boom took off during 2011 and 2012.