Trump and Taiwan Semiconductor Manufacturing (TSMC)

Note:  the post just before this has a brief description of TSMC.

new restrictions on TSMC

Last week the Trump administration announced that foreign users of American-made semiconductor design software or production equipment will need permission from Washington to use these to fashion chips for Chinese telecom company Huawei, the world leader in 5-G wireless technology.

At the same time, TSMC announced that it will be opening a new $12 billion fab in Arizona in 2024, its second in the US.  No details yet on why, although presumably Washington is footing the bill.

my thoughts

Huawei is TSMC’s second-largest customer, after Apple.  60% of TSMC’s output goes to the US, 20% to China.

I’m a fan of TSMC as a company but not of TSM as a stock.  This is because I don’t have any edge in evaluating TSM.  I find Taiwanese accounting opaque and I believe a ton of local knowledge is needed to be successful in sizing up that market.  While the latter is true just about everywhere, Taiwan is, for me, an extreme case.

I wonder how this new Trump rule can/will be enforced.

What would I do if I were TSMC?  I’d see if I could rearrange assembly lines to avoid making Huawei chips using US-sourced machines.  My ultimate goal, however, would have to be to minimize the threat to my business by transitioning away from US equipment suppliers.  This might mean giving extra assistance to Japanese or EU companies, or encouraging technology transfer to develop Chinese alternatives.  It could mean moving advanced production equipment to foundries on the mainland to supply Huawei from there–making clear this output is not coming from Taiwan.  I’d probably be figuring I’d shed current generation US-made equipment I already own by moving it to the new US foundry.

If I were a current US supplier to TSMC?  If I wanted to keep TSMC business, I’d be starting to figure how to shift at the very least that part of my operations out of the US.  The same if I were a US-based maker of semiconductor design tools.

 

I think this will end up being another aspect of the “chaotic disaster” that is the Trump economic policy.  In this case, though, the purpose of the move appears solely to be to deflect attention from Trump’s worst-in-the-world response to COVID-19, in support of his lie that somehow not his bungling but Beijing and/or the Obama administration are responsible for the unnecessary deaths of tens of thousands of Americans.

I continue to think that Trump and his enablers in government and the media are doing enormous damage to the long-term economic prospects of the US.  What strikes me most about the developing TSMC situation, however, doesn’t have much directly to do with the stock market.  It’s that Trump et al are concerned only about covering up what they’ve done; their cynical strategy is to lie and to distract by doing more harm elsewhere.  There isn’t the slightest hint of remorse for what they’ve done nor sympathy for the relatives of the death.

 

 

 

 

 

 

 

 

Taiwan Semiconductor Manufacturing (TSMC): background

what TSMC is

In the early days of semiconductors, chip-making firms tended to be vertically integrated, meaning the companies that designed semiconductors also manufactured them in their own plants.

That changed as the semiconductor industry began to expand rapidly in the early 1990s, for several related reasons:

–chip designs became progressively more specialized and complex, putting increased focus on the design process

–the cost of building chip fabrication plants to manufacture newer, higher-specification, designs rose exponentially, putting them out of reach for all but the biggest firms, and

–TSMC opened in 1987 as a third-party manufacturer, allowing dedicated design shops to set up on their own and still be able to have their designs fabricated.  The design business, something at which Americans have excelled, has flowered since.

Today, TSMC is the most advanced chip manufacturer in the world, and by far the best third-party fabricator, matched only by Samsung, an integrated firm, and maybe Intel.

 

semiconductor equipment makers

Today’s semiconductor fabs are extremely expensive.  TSMC has just agreed to build a new fab in Arizona, for example.  The cost:  $12 billion.  (More on this in the accompanying post)  The equipment inside, the most advanced pieces of which can cost hundreds of millions of dollars, comes from a small number of specialized machinery firms, which are located mostly in the US, Japan or the EU.  Because of the complexity of semiconductor manufacturing and the expense and long lead times involved in developing and testing new equipment, there tend to be very close cooperative research and development relationships between the fabs and their equipment manufacturers.

 

foundries are the future…

…absent some revolutionary change in computer technology.  A decade ago, when I was more up-to-date on semiconductors, a state-of-the-art fab cost about $4 billion.  Operated efficiently, it would churn out, say, $7 billion worth of output.  Both figures are out of reach for most firms.  Hiring a trusted third party to manufacture your designs is the easiest way to go.  Although the ratio of sales to assets has shrunk since I was better informed, the absolute numbers have risen a lot.

 

 

 

 

 

Intel (INTC) and ARM Holdings (ARMH)

chipmaking rivalry

The big division in the chip-making industry over the past 15-20 years has been between giant vertically integrated makers like INTC, Texas Instruments … which manufacture chips designed in-house and smaller digitally-oriented design firms who rent structural intellectual property from ARMH, modify it and have chips made in third-party contract fabrication factories like those run by TSMC.

INTC’s advantages have been the raw power of its chips and its manufacturing superiority.  Users of the ARMH framework tout the elegance of their designs that enables output to be smaller, use less electricity and generate less heat.

disruption by iPhone

The balance of power began to shift away from INTC and toward the ARMH camp when INTC decided not to make chips for the iPhone.  It may be that INTC management thought smartphones were a flash in the pan, as urban legend has it, or it may simply have been that INTC knew its chips ran too hot and used too much power for Apple to be satisfied with them.  In any event, INTC has been trying to reinvent itself since then, by improving its chip design while maintaining its manufacturing edge.

On the latter front, INTC continues do well; on the former, not so much.  Despite a lot of design effort, its low-power, low-heat solutions for the smartphone world haven’t been good enough to gain much traction.

This itself threatens the manufacturing operation.  As INTC steadily shrinks the size of its chips, each silicon wafer processed becomes capable of yielding more output.  At some point, INTC’s factories are potentially going to be capable of churning out more chips than the company can reasonably expect to sell to its PC and server customers.  The capital equipment used in chip making is so expensive–$3 billion+ today, maybe $10 billion+ for the fabs of a few years from now–that the factories have to run at high utilization rates to be profitable.  INTC has already said that next-generation (extreme ultraviolet lithography) technology is too expensive for even INTC to invest in by itself.

Hence the deal with ARMH.

three other points:

–presumably working with ARMH-based firms will help INTC fine-tune its manufacturing processes for mobile and the Internet of Things

–this may be the first step in closer cooperation between the two companies

–the arrangement has been announced very quickly after Softbank agreed to acquire ARMH.  Are the two connected?  If so, Masayoshi Son may have plans for much greater integration of the two rival firms.

 

 

 

 

Intel (INTC) and ARM Holdings (ARMH)

At its Developer Forum yesterday, INTC announced that it is opening its cutting-edge fabs to manufacture chips that employ ARMH designs created by third parties.  So, as at least part of its business, INTC intends to become a foundry like TSMC.

(An aside: despite its glitzy style, it’s much harder to find information about the move on INTC’s website than on ARMH’s.  I don’t know whether this has any significance, but it’s the sort of odd fact that rattles around in a security analyst’s head until an answer can be found.  Is it me?  Is INTC more interested in sizzle than steak?  Is INTC’s IR effort still mired in the mindset of the former regime?…)

I’m not sure what the total significance of this move is, but at the very least:

–TSMC, the premier foundry, a Taiwanese company, trades at about a 17x price earnings multiple.  INTC now trades at about the same PE, although it has typically traded at a lower rating than TSMC in the past.  In contrast, ARMH trades at about 70x, a PE that I think must be unsustainably high, even though ARMH has managed to do so for years.

For my money, INTC’s fabs are better than TSMC’s.  Making loads of ARM chips for others will likely not lower INTC’s pe ratio.  Arguably, as the foundry business expands, INTC’s pe will rise.

–in every generation, the size of chips shrinks while the cost of a next generation fab rises. As a result, the amount of output that a fab must have to be able to operate profitably increases, while the penalty for having too little output goes up as well.

The ARMH partnership signals, I think, that INTC believes that to maintain its manufacturing edge, it must accept manufacturing orders from outside parties.

 

More tomorrow.

 

 

 

 

Intel (INTC) and next-generation semiconductor plants

We’re living in a time of immense structural change.  For investors, the internet-led waning of established brand names and bricks-and-mortar distribution networks and the loss in value of existing manufacturing plant and equipment as new factories spring up in developing countries are among the most important.

One exception to this trend has been in semiconductor production.  There the engineering knowhow required to run the factories is very high and the pace of change has been very swift.  In addition, the cost of building a new fab is prohibitive for most:  a current-generation plant costs about $3 billion.  If a firm wants to fill the plant exclusively with its own chips, that requires annual sales of $7 billion or so.  In other words, only INTC and Samsung are big enough and rich enough to afford their own plants and the technological edge that brings.  Everyone else has to use third-party contract foundries like TSMC.

INTC currently has maybe a two-year lead over other semiconductor manufacturers in process technology.  It can make smaller, faster, less power-hungry chips than anyone else.  To preserve this advantage, the company has been making preparations–including funding research by equipment manufacturer ASML–to be the first out of the blocks with plants running new “extreme ultraviolet” technology and processing much larger silicon wafers to boot.

What has come to light very recently, however, is that these new factories are going to be mega-expensive, at $10 billion apiece or more.  As things stand now, that makes building one of them a bet-the-company move for anybody, even an industry giant like INTC.

Where to from here?

It’s not 100% clear, to me anyway, but:

–extreme ultraviolet lithography is probably at least a half-decade away, not a 2017 event as previously thought.  This means R&D and capital spending will be spread over a much longer period of time.  All other things being equal, this will mean higher cash generation by INTC.

–absent a significant cost-reducing breakthrough in EUL, INTC will likely partner with at least one other party to fund a next-generation fab.  Partnering could either be with other chipmakers or with one or two large deep-pocketed users of chips.   …AAPL?  …MSFT?

–it’s conceivable that INTC will itself end up with a substantial foundry business manufacturing chips designed–at least in part–by third parties.  One might argue that this would be a come-down for the premiere manufacturer of proprietary microprocessors.  However, the best foundry, TSMC, trades at a substantial PE premium to INTC even though its technology is inferior.  So morphing into a foundry could easily add a quarter or a third to INTC’s stock price.