Chips: The Crude Oil of the 2020s?

I heard on the news this morning that Ford was slowing down production of their F-150 line.  This is the crown jewel of the company's offerings, and has been the best selling model of any make of car or light truck in America.

For almost four decades. 

(Note that last year, GM's Silverado and Sierra models combined just barely outsold F-150 for the first time.)

Why would Ford slow production of their most popular offering, especially when they are neck and neck with a fierce competitor?  Was it slackening demand?  Production problems?  Reliability issues?  No, none of the above.

They were missing a chip.

A tiny, little, piece of silicon, perhaps the size of your fingernail, brought the Ford Motor Company to their knees.  They can't produce an F-150 truck without it.

There is an average of 3000 chips in each and every car which is manufactured these days.  And these little pieces of silicon are needed not only for cars and trucks, but for refrigerators, cell phones, computers, TVs, dishwashers, pacemakers, and washing machines.  

The machines that make vaccines and test for Covid have chips in them.  Blood pressure measuring devices, stoves, ovens, instapots, light bulbs, thermostats, watches, headphones, hearing aids, hotel room keys, coffee pots, servers in Amazon's AWS data center (serving the apps on your smartphone), the machines that make the chips themselves, spaceships, satellites... they ALL have chips in them. Heck, even my brother's new bicycle has a chip in it.

You would have trouble finding non-food items that don't have chips in them (and soon, food items will feature RFID tracking devices which... yup, have chips in them).  

(And just in case you are wondering, no, the Covid vaccine doesn't contain a 5G tracking chip.)

If there is a shortage of a chip in any of the aforementioned household products, that production line will come to a screeching halt.

Chips are like "the crude oil of the 2020s" and, unfortunately, America is no longer the "Saudi Arabia" of chips (even though we invented them).

And, as we will see, America will have a very hard time taking back the lead in manufacturing of semiconductors.

Making New Chips from the Old Block

Semiconductor manufacturing is not like other industries.  Each and every generation of chips requires a brand new facility filled to the brim with unbelievably expensive, extremely high tech machines.  Historically, it cost about $4B to build a new fab (the plants that make chips are called fabs).  However, to make the latest, deep-deep-deep submicron (e.g. very advanced) chips, these days you'll need a fab that costs on the order of $20B.  For a single factory.  Even more critically, you need the expertise to run your new $20B fab at reasonable yields.

For chips, smaller is better (and more advanced) - and the newest generation of chips have very small feature sizes, indeed.  In fact, the features that are in the chip are smaller with each and every new generation of devices.  The actual chip need not get smaller, you typically just fit more stuff on your chip - i.e. more capability with the same sized device.  

And having smaller features with each new generation also enables higher performance because they are closer together and can run faster (a big caveat is that technical issues with power consumption / dissipation and current leakage significantly limit all this... perhaps a topic for another post). 

Ever since the integrated circuit was invented in 1959, chip feature sizes have shrunk for each and every successive generation.  This has caused a typical chip's capability to double and redouble for new chips roughly every 18 months.  This is Gordon Moore's eponymous law. Mr. Moore was a founder of the Intel Corporation and Moore's Law has proven true for over half a century.

Over the years, this doubling and redoubling of semiconductor capability has given us many of the incredible things that we now take for granted.  A good example is the smartphone that you are holding in your hand (perhaps you're reading this blog on it). Apple's latest iPhone has far more computing capability than an entire building-filling datacenter had just a few decades ago.

Moore's Law is what mathematicians call an exponential progression, and the implications are mind-bending.  Very few things in our natural world behave exponentially (an example is the growth of, say a Coronavirus in a human population, which, like Moore's law, also behaves exponentially).  One of my favorite illustrations of Moore's law is hypothetically applying it to the internal combustion engine.  If Moore's law did apply, you'd be able to drive a typical 2021 model year automobile at a velocity that approaches the speed of light and travel over 90 million miles on a gallon of gasoline.  

The features on these chips have shrunk so much over the years that they are becoming much smaller than the wavelength of light. This wreaks havoc with the lithography processes that create these features.  One of the crucial steps in the making of chips is kind of like taking a photograph, and when the things you want to resolve in the picture are smaller than the wavelength of the light you are using, you have to become a very clever engineer to get a sharp picture.  Other countless things also become exponentially more difficult for each and every new generation of chips.

In the past, this was not an issue.  There were plenty of smart process engineers who were employed in fabs throughout the world, and these talented designers along with the semiconductor equipment makers (the firms that make the machines that make the chips) were keeping pace with Moore's law.  As time went on, however, fewer and fewer chip-making companies were keeping up with Moore's Law, as it became very challenging to yield the latest generation of chips.  Recently, these technical challenges have narrowed the field of chip makers to only three companies who can do it:  Intel, Samsung, and TSMC (Taiwan Semiconductor Manufacturing Corporation).

I have personally toured the fabs in all three of these companies and they are impressive, indeed.  They have exclusively amassed and retained the engineering expertise required to make the most advanced chips - all the other chip makers have fallen behind.  In addition, the three leaders have kept close relationship with the chip-making-machine fabricators, and worked exclusively in lockstep, advancing each generation.  These companies have acquired a significant competitive advantage as a result.  

The three companies also have acquired what Warren Buffett calls a deep "moat"... knowledge, expertise, equipment, corporate relationships, exclusive contracts with vendors, infrastructure and general knowhow that is almost impossible for other companies to replicate.  This is a great place to be in if you are in one of these three companies.

Until Intel missed a step.  And now there are two.

A Broken Clock

When I worked at Intel back in the early '90s, the company ran like a clock.  Literally.  We had Intel's famous tick-tock production scheme, where each year a new generation of chips came out, right on schedule.  Tick-tock. Tick was a major change, and then tock was a more minor update the following year.  Merely thinking about slipping even a "tock" schedule got you fired.  

In 2019, Intel missed the introduction of the 10nm "Tick".  ("10nm" signifies that the feature size of the chip making process is 10 nanometers or 10 billionths of a meter.  Needless to say, a smaller number is better.) 

That put Intel well behind Samsung and TSMC in producing the most advanced chips.  They also had to slip out the following generation as well which had features sizes of 7nm. It might come out in 2022. (Intel recently fired their CEO Bob Swan and hired back Pat Gelsinger from VMWare as the new CEO - Pat was Intel's CTO when I worked there.)

TSMC is now shipping 5nm finFETs (BTW Intel invented the finFET - which is a geometric feature in the chip that significantly reduces current leakage and power consumption).

Samsung is now shipping 5nm mobile microprocessors.

Intel is still making 10nm chips and now releasing their latest more-advanced designs into TSMC. No wonder they fired the CEO.

What Now?

If you need to source a chip that needs the latest and greatest process to be built, you can't get it in our country anymore.  You'll need to go to Korea or Taiwan.

Korea

Samsung is a very interesting company and probably a poor choice for a supplier of chips in most cases.  They are one of the vertically-integrated Korean Chaebols and I can tell you from personal experience, they are very difficult to do business with.  

Part of the reason is that if you are seeking to make a chip in their fab, or buy one of their subcomponents, you are probably competing with them.  Samsung probably also makes whatever final product you are producing.  They are in all kinds of markets - household appliances, TVs, nuclear reactors, shipbuilding, and many more businesses.  It is a very bad business practice to source a critical component from a competitor - and a foreign one at that.  They can and will hold you hostage once you are dependent on them for continuity of supply.  For that reason, I would predict that American companies will only use Samsung as a supplier of last resort.

Taiwan

That leaves TSMC.  They are a wonderful supplier.  Their only business is making chips for other firms, so you will not be competing with them (you'll be competing with their customers instead).  Apple sources their chips there, as do many other leading companies around the world.  TSMC can afford to make massive R and D investments and they have access to probably the world's best IP (intellectual property) resources to draw from for your designs.  

Taiwan is a wonderful country, with a young, highly educated and motivated workforce.  Their culture is amazing - I have traveled there over 50 times, and each and every time, I learn something new about Taiwan.  They have made investments over the last five or six decades in technology design and manufacturing and are now reaping the benefits of it - they have world class capabilities, especially in chip manufacturing.

But there is another country watching all this.

The Dragon In The Room

110 miles to the west, across the Taiwan Strait, lies an existential threat to Taiwan, and as we will soon see, also to the United States.  

In 1949, Chiang Kai-shek crossed these 110 miles of open water, escaping from Mao Zedong's armies and founded the Republic of China (ROC).  Accompanying Mr. Chiang was over 2 million refugees along with THE national treasure of China - boatloads of irreplaceable artifacts that represent over 4000 years of Chinese history - these treasures previously resided in the Forbidden City of Beijing and now are on display in the National Museum in Taipei.  If you happen to visit the Forbidden City, which I did about 10 years ago, you'll find all the rooms in the city absolutely empty. It is a constant reminder to the citizens of PRC of the missing treasure.


The Forbidden City in Beijing, China


Note that Taiwan is now known as the ROC or the Republic of China.  And ever since 1949, the Republic of China has done a dangerous and very calculated dance with the PRC or People's Republic of China (known to most Americans as Mainland China).  Every mainland Chinese administration since Mao has claimed that the ROC is part of PRC.  And every Taiwanese administration since Chiang Kai-shek has not denied that claim - ROC administrations have been very careful to even avoid the slightest notion indicating independence from the PRC in all communications (recent administrations in Taiwan have taken somewhat more risk in this regard though).  

So Taiwan has been a tinderbox for more than seven decades.  And given the very recent repression of protests in the city of Hong Kong (fully acquired from the British in 1997), it wouldn't surprise me to see escalating tensions across the Strait of Taiwan.  The precedence has been set.

There are cultural and historical reasons for the tension (the mainlanders are still mightily pissed about the national treasure).  Now there are economic and strategic reasons.

And, importantly, China's semiconductor fab technology lags significantly behind the rest of the world.

It would not surprise me to see an escalation of hostilities between the PRC and the ROC (possibly even an outright hot conflict breaking out):  TSMC has become a mighty juicy target.

Just my opinion.  What is yours?  And what do you think the Biden administration should do about this?


Comments

  1. far more concerned about China's control of world Cobalt supply ( electric car batteries)

    solution? a more targeted national industrial policy that identifies strategic industries and subsidizes diverse sourcing

    ReplyDelete
    Replies
    1. Fair enough, strategic metals and rare earths could indeed become extra scarce in the next decade. NMC cathodes for batteries will be in demand for sure. Some are working on solutions to get away from cobalt, it is also a difficult metal to dispose of. We shall see!

      Delete

Post a Comment

Popular posts from this blog

Will Tariffs Jack Up Your Bills and Push the World to the Brink?

America’s Ideological Dumpster Fire