Ben Scott

On September 12, 2018 Goldman Sachs downgraded its view of the semiconductor capital equipment sector to “neutral” from attractive. In this article they said,

“We are lowering our coverage view and making select rating changes as we account for the deterioration in memory pricing (and thus margins) and potential implications for future memory.”

This marks a recent skid in the semiconductor sector as Morgan Stanley recently warned that demand for DRAM memory chips was weakening, pricing pressure was building, and vendors were struggling to move their wares.

Now, the question is:

Why would Goldman downgrade the heart of the electronics world?

The answer is a lack of confidence in management to deliver future cash flows.

Whether you are experienced in the world of semiconductors, or not, this lesson is valuable.

Share prices are the present value of all future cash flows that a company is expected to deliver. Thus, when markets have confidence in management, share prices rise, when they don’t, share prices slide. Thus, based on current information, we can see that investment banks have no confidence in the ability of semiconductor companies’ management to deliver future income.

The Earth is Flat

The cash flows of all semiconductor-producing companies are set to fall for one simple reason: lack of innovation (change from process to product innovation, no disruption). These companies are giving customers what they think they want, which is a problem in and of itself.

This is the disruption dilemma. Henry Ford famously said “If I asked people what they wanted, they would have said faster horses.” What Henry Ford observed is what management at Apple, Google, and Amazon (for example) know too well — if you only focus on current market need, you will be disrupted and you will lose. Research in Motion (the maker of Blackberry) learned this the hard way when the iPhone was launched.

The downside of social networks is that they only tell you what you already know, and if your social network “knows” the Earth to be flat, beware.

So why are bright, motivated, and well-compensated management teams following a path to their own destruction? You might well ask.

The semiconductor business has significant barriers to entry. The first being cash and the second knowledge. Foundries are intensely expensive to build and operate. However, modern devices are at the frontier of what silicon can deliver.

5G networks, for example, won’t perform optimally on Silicon and are better to suited to materials such as Gallium Nitride, but it’s tough to make and requires new investment in new foundries and processes.

However, modern devices often run at higher temperatures, are radio frequency and/or opto-electronic devices and thus require chip sets with new architecture.

All this means risk to the establishment, so they prefer to stick to their knitting.

Change the Establishment

I was recently at a leading semiconductor plant and was asked to go in with a customer as this site was experiencing assembly and other production problems on a new device.

What astounded me was that the engineering teams were relying on data from 20 years ago from processes that are only capable of producing tools and components to (at best) 12 microns when they are trying to assemble devices made in the nanometer scale.

Its no wonder those that know better downgrade this sector. I have worked in electronics manufacturing for 25 years and the industry is a mess. Very few want to take risks, they want to use tools that are not fit for their purposes and, as a result, these same engineers cannot reliably produce semiconductors at 10 nano-meters. How can they reach 7 nanometers? How could they move to 5 or 3 or 1 nanometer?

The world’s innovators are not going to have their wings clipped by their supply chain, they will build it the way they want it.

Apple and Amazon have already made clear statements on this. Both Apple and Amazon either have in-house capabilities or they have partnered with a leading foundry and shaped their own futures.

Samsung has gone even further. They have not only invested heavily in their own capability, but are also supplying their competitors with devices. It has long been Samsung’s strategy to learn on other peoples’ money — build for someone else at a discount, then launch your own products when your people have the skills to do it right.

These strategies of partnership and in-house production are likely to be the model going forward for those companies that wish to innovate and deliver the next generation of advanced devices. In many ways its simpler and feels less risky. It’s easier to manage as the teams have one customer (as it were) and do not get distracted and have more reliable funding. It feels less risky for management because they are part of something bigger and have a clear trajectory.

The downside for consumers is that only a few companies can do this and so there is risk of concentration and possible monopoly. We can see this starting as prices in certain areas of tech are rising and these increases are not cost-driven. We are also seeing increased consolidation in the semiconductor market and this is also a sign that profits are falling and are likely to continue to do so.

So, what?

What does this mean for the shareholders of the likes of Intel, Global Foundries, TSMC, Broadcom, Qualcom, et al? The most likely outcome is that returns will diminish in the long run and the companies disappear. Disruption will come to silicon-based devices in the same way it came to valve manufacturers and the good old transistor and, as is always the case, the incumbents wont see it coming.

They won’t even consider it a threat when it does first appear on the horizon.

The question, then, is when should I get out?

Well, if you hold semiconductor shares in your portfolio you might want to think about which companies you hold and which you sell over the next few months. There could be some more consolidation, but one thing is for sure: 5G is here and many organizations are likely to be coming up with designs for consumer devices that just wont perform the way they should if built on silicon.