Description

Magnachip Semiconductor (MX) is an analog and mixed signal semiconductor manufacturing business. The company’s predictable cash flow stream is misvalued due to an inventory correction and a recent accounting change that has delayed the publishing of its financial statements. Once financials are published, and the market understands the true earnings power of the business, I believe the stock will rally to $25, or 10x 2015 FCF of $2.51/share (80% upside).

What Do They Do?

Magnachip operates a hybrid model, offering foundry services (outsourced manufacturing) for fabless semiconductor companies (50% of revenue) and its in-house designed power and display products. MX operates three fabs with a combined 130k 8” equivalent WSPM capacity in Korea at the 110-800nm production node.

End Market Exposures

The best way to describe MX’s many products is as integrated circuits on analog chips that handle high voltage management within TVs, monitors, PCs, phones, game consoles, set top boxes, automobiles, energy (wind/solar) devices, and other electronics.

The first diagram, above, shows where MX’s chips help with voltage regulation between digital chips and other components, while the second diagram shows us exactly which components they manufacture and where these components lie within a smartphone. MX’s products are fairly similar in PCs, tablets, TVs, and other consumer electronics, since the underlying functionality is the same.

Segments:

Display (34% of revenue and 38% of gross profit)

Within display, MX sells large and mobile display solutions. These include voltage and frequency regulating drivers to power on and off the displays within consumer electronics.

Power (16% of revenue and 10% of gross profit)

Within power, MX manufactures (1) MOSFETs, (2) LED drivers, and (3) DC-DC converters, among other products.

(1) MOSFETs are used in applications to switch, shape, or transfer electricity based on changing power requirements (i.e. voltage regulation) in mobile phones, LCD TVs, computers, consumer electronics, and industrial equipment. Essentially, the voltage coming out of a battery is much higher than the amount that the CPU of a device can use—MOSFETs help to manage this.

(2) LED Drivers (used in LCD TVs, monitors, PCs, lighting, etc.) similarly use voltage regulation for dimming.

(3) DC-DC converters efficiently convert voltage lower (e.g. 8v to 4v) for all sorts of consumer electronics/industrial applications to dynamically manage voltage ranges.

Foundry Services (50% of revenue and 52% of gross profit)

Within foundry, MX’s process technologies include (1) Mixed-signal, (2) Power, (3) Non-volatile memory, and (4) High voltage CMOS.

(1) Mixed-signal devices have both analog and digital circuits on the same die. Examples of use might include an A/D converter that changes the analog signal coming from a sound microphone or ambient light sensor (as an electrical signal), into digital values to be used to digitally process the signal.

(2) In foundry power, MX manufactures products similar to their in-house designed own power segment, but for third party customers.

(3) Non-volatile memory integrates flash memory onto a chip so information can be stored even when the chip is off—used for long-term persistent storage.

(4) High voltage CMOS is used when power levels are extremely high and must be managed down through thick metal wires/processes—when voltages are extremely high, only large metal wires placed on very large analog chips can handle the current.

Business Strength:

Extremely Stable, Cash Flow Generative Manufacturing Business

While the market tends to discount low growth technology businesses due to their limited revenue growth and potential product obsolescence, MX is far from obsolete and the 15-20+ year runway of stable/growing cash flows offers a high and sustainable free cash flow yield at current prices.

Analog Does Not Scale (i.e. Moore’s Law Does Not Exist in Analog)

Moore’s Law is the process by which new manufacturing technologies allow transistors to be more densely placed within a semiconductor chip, allowing the size of the chip to decline, while leaving the transistor count constant. This does not exist in analog, mainly for two (related) reasons. (1) Analog chips handle high voltages. (2) Analog chips regulate down voltages. (1) As described above, many analog chips handle voltages close to the electrical outlet or the battery. As in water flow, when water or in this case, electrons flow in large quantities/high voltages, the pipe/wires linking the buckets/transistors must be extremely large and cannot shrink. As such, it is impossible to reduce the size of the chip and place the transistors closer together when dealing with high voltages. (2) Analog chips handle voltage regulation. Similar to (1), when voltages are regulated down, they must move from larger wires to smaller wires, resulting in a similar inability to place transistors closer together, which we see in digital semiconductor chips. As such, analog will always exist, as long as voltage regulation exists. And as we use more electronics and our electronics become more complicated, we will continue to see low single digit predictable revenue growth for analog semiconductor manufacturers like Magnachip.

Fully Depreciated Fab with Limited Capex

MX’s fabs are old Hynix DRAM fabs. These fabs are fully depreciated and, because the manufacturing nodes at which MX makes its products are held constant (i.e. no node shrink, AKA Moore’s Law described above), they require very little (and extremely predictable) annual capital investment. Management estimates this to be in the 7% of sales range. This dynamic, combined with our predictable revenue growth, offers predictable free cash flow generation/growth.

Favorable Supply/Demand Balance

As mentioned above, because analog fabs are old and fully depreciated, it is extremely unlikely that we see additional manufacturing supply enter the market. The most recent transaction of equipment/manufacturing space similar to MX’s was TXN’s acquisition of equipment in bankruptcy that became RFAB. The price paid was $12.2m/1k WSPM, which would value MX at $1.5bln in reproduction value or 3x the current $550m EV. A greenfield fab build would be multiples higher. Furthermore, while MX’s fabs were previously DRAM fabs that were converted, each of the other semiconductor products, DRAM, NAND, and digital logic are all at 20nm, 15nm, and 20nm, respectively, as they have each shrunk their nodes through Moore’s Law. It is now impossible to convert these nodes back to analog at MX’s 110-600nm manufacturing. As such, limited new capacity is entering the market that might disrupt the supply/demand balance.

Transition to Higher Gross Profit/Wafer Power/Display

MX is in the process of moving some of its capacity from lower priced, lower gross profit dollar wafer foundry products ($490/wafer) to higher priced, higher gross profit dollar internally designed power ($865/wafer) and display ($795/wafer) products. By transitioning capacity to specialized products, MX can grow revenue and gross profits without increasing manufacturing capacity (i.e. limited capex growth), given fungibility of manufacturing lines. This should allow MX to hit their goal of 5-15% revenue growth and 1-2% gross margin growth with limited opex and capex growth (discussed later in the valuation section). This transition should also result in revenue growth through a shift to higher growth AMOLED products in their display business and LED lighting/energy harvesting in their power business. 

Why is MX Currently Misvalued:

Earnings Restatement Ordeal

In Q3/Q4 2013, an inventory correction occurred in the semiconductor business due to a slowdown in smartphone sales growth, which led to an abundance of inventory at MX’s distributors. Since MX recognized revenue on a sell-in basis, they booked the GAAP earnings from sales to these distributors while the distributors failed to make payments, resulting in ballooning accounts receivables and limited cash flow generation at MX. The former CFO decided to restate the financials in January 2014 to recognize revenue on a sell-through basis from these distributors to better tie earnings and cash flow. This restatement process has gone on since January 2014 and the market is now concerned that the issue might be greater than a simple accounting change.  The CEO and CFO were both fired by the board (controlled by Avenue Capital) because of this concern.

The five month delay in reporting financials could indicate fraud.  However, I believe that this is a simple case of incompetentency, with a CFO who failed to fathom the length of the restatement process. Distributors in Asia keep poor records of historical transactions and product delivery. However, with the accounting change, MX must now identify exactly both when these transactions occurred at their distributors and when the product was delivered to the end customer for the last two years. In similar accounting restatements, other companies have spent 7-8 months identifying the transaction records before notifying Wall Street of the accounting change, reducing the time to restate to days or weeks, instead of months. The CFO’s incompetence in properly assessing this timeframe has resulted in the delayed financial reporting, but there is absolutely no fraud here. Moreover, the recently fired CEO is very old and his commitment to the company has waned. I believe the board was seeking an opportunity to replace him, and found it in the restatement delay. They appointed his protégé as the interim CEO, which they would not have done had there been fraud at the company.

After attending a Magnachip customer conference, where I spoke both with middle/senior company management and many customers, I have become extremely comfortable in my belief that the operating team is of extremely high quality, the manufacturing process is solid, and their pricing is the most competitive in the industry, and I believe we should see current financials filed by September/October.

Valuation/Price Target:

Based on a shift from foundry to power/display, as well as low single digit overall revenue growth in analog, I see about 3-5% revenue growth, conservatively below management guidance of 5-15%.

Based on management guidance of 1-2% gross margin expansion from the shift to specialized products, limited opex growth, somewhat offset by higher cash taxes due to NOL burns, I see FCF/share in the $2.50 range for 2015, resulting in a $25 price target at 10x FCF or 80% upside. See comps list below.

Catalyst