In mid-October, I published a brief article and first-pass valuation model for Microchip Technology MCHP with the promise that I would take a closer look at the company.
In my original article, I thought that Microchip looked fairly valued with a midpoint right around the present price in the mid-$70 per share range. After doing a deeper dive and talking to company management to understand the company’s maintenance capital and investment needs. I believe that my originally calculated fair value range was too low. I will publish a valuation this week, but this article offers a summary of Microchip’s business.
This article will allow you to place my upcoming valuation work in better context and gain a little insight into the world of microcontrollers and analog semiconductors.
Business Overview
When I followed Microchip as a semiconductor analyst in the mid-aughts, the company was essentially a producer of microcontrollers. Microcontrollers are chips that accept information about the environment and set off an electro-mechanical process in response to that information.
For instance, a microcontroller might serve as the electronic brains behind a digital thermostat or an electronic alarm system.
The type of chip that senses environmental variables (e.g., temperature, sound) are analog chips, and Microchip’s product portfolio also contained analog chips as well as some specialized memory products, but 90% or so Microchip’s revenues were generated by microcontrollers (MCUs).
As mentioned in my previous article, Microchip has been on an acquisition spree over the last decade and has fundamentally changed its asset portfolio over this time. Now, Microchip’s portfolio is roughly half MCUs, 30% analog, and the remainder split between FPGAs (special chips conceptually similar to MCUs), Memory, an IP licensing business, and miscellaneous products.
Figure 1. Source: Company Statements, Framework Investing Analysis
According to Microchip’s FY2018 annual report (Microchip’s fiscal year ends in March), in calendar year 2017, the total size of the MCU market was on the order of $18 billion and the size of the analog market was on the order of $50 billion.
Both of these markets are hot right now. When you hear the term IoT, these are the kind of chips that make IoT happen – sensing something about the world (e.g., Is it day or night?), setting a process off under given conditions (e.g., turn on the light), and reporting the action to a remote computer (e.g., an indicator on a mobile app showing that your living room light is on).
In addition to IoT applications, even non-connected devices have much greater functionality these days. Remember when microwave ovens used twist dials to set cooking time? Dials like that have been replaced by printed circuit boards with microcontrollers at their heart.
Microcontrollers
Because of the proliferation of high-function and connected devices, industry analyst, IC Insights, forecasts an annualized revenue growth of around 7% per year through 2022 with a unit growth of around 11% (i.e., implying that unit costs will consistently drop as well).
Microchip quotes a Gartner report saying that as of calendar year 2016, Microchip was the third largest MCU manufacturer in the world, behind Japanese Renesas and Dutch NXP NXPI. NXP, a firm we valued in the context of the bid for it by Qualcomm QCOM, vaulted over Microchip in the MCU league tables when it acquired American chipmaker, Freescale, in 2014. Microchip claims that its share of the global MCU market is around 16%.
Microchip’s MCU sales are skewed toward 8-bit products, which are the most ubiquitous and will certainly be around for a long time, but which are also not the parts that would go into the smartest, most interactive devices, due to a lack of processing power.
Microchip’s CFO, Eric Bjornholdt, told me that his firm was fifth and sixth on the league tables for the more sophisticated 16- and 32-bit MCUs, respectively, but suggested that the firm was gaining market share in those segments.
It’s worth noting at this point that MCUs are not the same kind of chip, from a commercial standpoint that Intel-style CPUs or Qualcomm-style application processors are. The difference is that while you might be tempted to get a new mobile computing device that has the newest, fastest chip (even though your device sporting last year’s fastest chip is still working fine), you will be less tempted to buy a new refrigerator because the MCU embedded within it is a newer, 16- or 32-bit model. When your fridge does start leaking fluid out of the bottom, you might upgrade to a fancy, display-panel equipped device, but that swap might be years in coming.
Also, from an engineering standpoint, when a certain MCU is designed into a device, that MCU usually has a relatively long production run because even if some aspects of the device’s external design changes, usually the internal control functions do not. Engineers operate on the “If it ain’t broke, don’t fix it model”.
Also related to engineering, because MCUs are standard parts that can be used to control different types of devices, a software program must be written to tell them what they are supposed to do when some condition is met. This software is programmed in a development environment and transferred into the MCU’s memory modules (which explains why Microchip has a memory business).
While similar MCUs from different companies carry out tasks in the same way, engineers get used to working in a particular chipmaker’s development environment. As such, the quality of the development environment and the learning curve one must ascend to design efficiently creates stickiness for an MCU maker.
According to an Aspencore survey quoted by Microchip management in a shareholder presentation, 68% of a group of 719 engineers responded that “The ecosystem supporting the chip (software, tools, support, etc.)” was the most important determinant of what MCU they used in their product design. Microchip management quotes other surveys showing that engineers show a preference for using Microchip parts in their next project, presumably because of Microchip’s support ecosystem.
Interestingly, Microchip (with its acquired Atmel business) is head and shoulders ahead of the pack among respondents referring to 8-bit designs, neck-and-neck with Texas Instruments TXN for 16-bit designs, and prominent among many for 32-bit designs. This is another triangulation point pointing to the relative importance of 8-bit business to Microchip.
A last point related to the demand environment is the importance of the MCU in the design of an overall device. The MCU is supported by an analog sensor and connected to other devices through I/O (input/output) ports. A device designer must figure out all the chips to buy when working on his or her design.
According to Bjornholdt, engineers are relatively likely to first select the controller – since the controller represents the “brains” of the device – then work out what other chips and devices should work in conjunction with that MCU.
Bjornholdt contends that tendency puts Microchip – a MCU company that also sells analog chips – at a relative advantage over a company like TI – which is mainly an analog chipmaker that also sells MCUs. I haven’t spoken with engineers to check this contention, but it seems sensible.
Analog
While technically different from MCUs, analog chips share some commercial similarities. Analog chips are the first semiconductor interface with the real world, so are found in the same designs, working in conjunction with MCUs. Like MCUs, they tend to have long economic lives for many applications – the “If it ain’t broke, don’t fix it” dynamic again. Last, analog chips do not operate on Moore’s Law, so do not need the newest, bleeding edge technology and equipment to manufacture them.
Because of these characteristics, analog chipmakers’ businesses are often characterized by high profit margins, low-ish but steady growth, and very slow changes in market share between companies.
A difference between MCUs and analog chips is that sales of the former tend to be to skewed towards consumers applications; there are plenty of analog chips that are used in consumer applications too (e.g., mobile phones), but there is much greater exposure to industrial markets (e.g., telecommunication equipment, military use, and the like.
The sales of the company that Microchip acquired earlier this (calendar) year, Microsemi, is skewed toward industrial markets; here’s the relevant mention of its markets in the firm’s last published 10-K:
The principal end markets that we serve include Aerospace & Defense, Communications, Data Center, and Industrial. Today, Microsemi products are found in applications such as: communications infrastructure systems, both wireless and wired LAN systems, implantable pacemakers and defibrillators, radar systems, military and commercial satellites and aircraft, and enterprise storage and hyperscale data centers.
This industrial-focused business is good for Microchip, whose MCU business took a big hit in 2008-2009 as consumer spending and housing starts collapsed. Of course, Microsemi’s business is not immune from vagaries of the economic cycle, but pacemaker and missile system buyers alike will continue to buy during a recession.
Industrial profit margins also tend to be healthy, since the cost of an analog chip tends to represent a small proportion of the cost of the device overall. Analog chipmakers can lean forward on pricing without the buyers pushing back too hard, especially considering the “If it ain’t broke…” dynamic mentioned earlier.
Microsemi is not the only analog asset that Microchip owns, but it is important for a few reasons. Microchip borrowed a good bit of money to fund the $10 billion purchase (including the retirement of Microsemi’s debt), so the firm’s financial leverage has increased. The leverage does not look excessive to me, but it has picked up a good deal compared to recent history.
Figure 2.
Also, judging from manager comments on Microchip’s last conference call, Microsemi – which had grown by acquisition as well – had not integrated the businesses it purchased as well as Microchip does. While it may seem like a small point, logistical issues like consolidating Enterprise Resource Planning (ERP) and back office accounting systems can take an enormous amount of time and managerial attention.
Last, Microchip is working quickly to inculcate its new employees to a its own corporate culture, and this is proving to be unsettling to former Microsemi employees. Microchip, a company that is famously conservative about its revenue recognition process, has criticized Microsemi’s management’s practices. These criticisms have led to a lawsuit by Microsemi’s top managers, all of whom were fired after the acquisition closed.
In addition, Microchip said it was cancelling sporting venue box seats and other wasteful corporate perks that must have been enjoyed by Microsemi employees in the past. More jarring, though, was Microchip CEO Steve Sanghi’s comments that he was unifying compensation arrangement by cutting commission payments to Microsemi salespeople.
I’m not a salesperson, but the phrase, “cutting sales commission” seems as though it has the potential to be taken badly.
Microchip’s Bjornholdt believes that Microsemi can be integrated over the next two years or so; this may be so, but I have seen that integrations can be tricky. Also, because of the nature of analog chip design, the intellectual property embedded in an organization is very important to the value of that company. Microchip will need to keep its engineers happy. According to one industry source, Microchip’s earlier acquisition of Atmel prompted defections among key technologists at the acquired firm.
Microchip’s disciplined management style is probably robust enough to meet this challenge, but the acquisition of Microsemi may turn into more of a struggle than Microchip executives originally thought.
Business Strategy
I knew Microchip as a company that was a very intelligent semiconductor fabrication center (“fab”) investor. The equipment Microchip uses does not need to be bleeding-edge in the same way the equipment in an Intel or TSMC plant does.
When there were a greater number of smaller, bleeding-edge chipmakers out there, the semiconductor cycle would invariably turn down and one of those small chipmakers would be caught too far out over its skis. Heavily levered, it would go bust or at least have to sell a fab or two.
Microchip had a knack for buying the fabs of these bankrupt chip companies for less than they were worth. If it didn’t need the extra fab capacity, it could mothball the facility for a few years, then recommission it when management thought they needed more capacity.
The company, under the leadership of long-time CEO, Steve Sanghi, was a competently run, conservative, and intelligent allocator of shareholder capital.
Over the last decade, there has been an enormous consolidation of the semiconductor space, and Microchip has taken its skill in asset allocation to the next level to buy complementary businesses. The increased share of analog, an eclectic FPGA business, and IP licensing business all stem from these acquisition efforts.
“Microchip 2.0” as the company calls its strategy now, is to provide an entire printed circuit board’s worth of chips to a potential customer — leveraging its MCU franchise to capture greater customer dollars.
This is a sensible strategy, but the undeniable fact is that Microchip is a good bit smaller than other companies in its peer group.
Figure 3.
Microchip is a well-run firm and CFO Bjornholdt told me that it plans to be one of the firms with a chair when the music of industry consolidation stops, but its size and financial position limits its possibilities.
After the Microsemi acquisition, the company does not have the capacity — financial or managerial — to make another acquisition for a few years. Its revenues will grow — perhaps at a quicker pace than its competitors — but its competitors may have greater options.
Texas Instruments is the top-ranked analog chipmaker with roughly 18% share of that market that also has an MCU business. TI’s MCU business was ranked #5 in the 2016 MCU league tables.
If Bjornholdt’s contention that an MCU business is a good entrée to sell analog chips is true, certainly someone at TI must be thinking that acquiring Microchip might not be a bad idea. TI’s debt to equity ratio is a quarter of Microchip’s and it is hard to imagine that it could not borrow money to fund a purchase if it wanted.
I have not looked at Texas Instruments’ financials for a long time, but will start to dig in after I finish the Microchip model. There may be technological or organizational reasons that I do not yet understand why TI would not want to buy Microchip, but the idea seems naïvely attractive from my perspective.