In my experience, an important part of successful value investing is the ability to recognize the common traits of underpriced companies. With TowerJazz (NASDAQ:TSEM) I was fortunate because, not only does it have business relationships with many of the companies I've previously written about, it also has many of their advantages. TowerJazz's new JV is reminiscent of Micron's (NASDAQ:MU) bargain purchase of Elpida which transformed the company and gave it critical mass which allowed it to compete effectively, while its impending tidal wave of EBITDA looks a lot like the situation ChipMOS (NASDAQ:IMOS) went through over the past few years (driven by rapidly filling high fixed cost capacity) which has left it flush with cash. Meanwhile, TowerJazz is organically seeing growth in sectors that I know well, image sensors and RF, from some of my top picks of the past year like RF Micro Devices (NASDAQ:RFMD), OmniVision (NASDAQ:OVTI), and Himax (NASDAQ:HIMX). This is a rare confluence of events in a company that already has a substantial market cap of more than half a billion dollars which should make it relevant to institutional investors as my thesis plays out. I believe that the convoluted nature of the opportunity at hand has left it undiscovered and underpriced, with 2015 EV / EBITDA of 3 and 100%-150% upside over the next 12 months.

TowerJazz (also known as Tower, or Tower Semiconductor) is a specialty foundry company for semiconductors. This means that they make chips for companies that don't have their own fabrication facilities, or that want multiple sources for a product in order to decrease risk. As a specialty foundry, TowerJazz specializes in processes necessary to make the increasingly growing variety of chips in the electronics we use every day. There are two basic kinds of semiconductors: digital and analog. Digital devices do processing, such as the central processor in your computer. Analog devices interact with real world signals -- the sensor in your camera, the radio frequency chips in your phone, the display driver in your screen, etc. Mixed-signal devices combine digital and analog functions. TowerJazz specializes in analog and mixed-signal technologies.

TowerJazz's position as an analog and mixed-signal foundry puts it in a prime position to grow along with the fastest growing segments of the electronics industry. Semiconductor fabrication requires increasingly high capital expenditure and an increasingly broad array of specialized technologies. Many electronics companies find themselves unable to keep up technologically, or unable to fill their own fabs to levels sufficient for profitability. TowerJazz has uniquely positioned itself to take advantage of this trend of outsourcing production with a new transformative joint venture, cost cutting measures, and organic growth from new and current customers, leaving the company extremely undervalued.

Fallow Ground In Japan

Japan has been in a state of persistent economic decline since its zenith in the 1980s, brought on by various macroeconomic issues. While this economic malaise has been hard on Japan's manufacturing industry, it's been especially painful for the electronics sector. Sony (NYSE:SNE) has been unprofitable in five of the last six years and is spinning off failing businesses, such as its Vaio laptop line. Elpida, Japan's last remaining DRAM chip maker, went bankrupt. Sharp, manufacturer of the world's most advanced LCD displays, only staved off bankruptcy with financing from competitors including Samsung (OTC:SSNLF) and Qualcomm (NASDAQ:QCOM), and with prepayments from Apple (NASDAQ:AAPL). Renesas SP Drivers, a joint venture between Renesas, Sharp, and Powerchip, was the second largest LCD driver manufacturer in the world, and will soon be a unit of the American company Synaptics (NASDAQ:SYNA). I was surprised when I learned that much of the worthless property in downtown Detroit is being converted into urban farms, but I was shocked to learn that Fujitsu, Hitachi, Panasonic, Olympus, and Sharp are all converting some of their fabs into high tech farms. These fabs each cost hundreds of millions of dollars to build, and now Fujitsu is cultivating hydroponic lettuce in a Japanese fab, while Sharp is growing strawberries in a fab in Dubai.

(click to enlarge) (Source)

The cabbage...I mean carnage in the Japanese electronics industry has left some fabulous loot up for grabs for those astute enough to take it. A prime example is Elpida, which boosted Micron's production capacity by 45% after Micron bought Elpida in bankruptcy. This proved to be a transformative move for Micron as it gained critical mass at the same time as the memory market turned around. I believe that the same situation is playing out for TowerJazz with its formation of a joint venture with Panasonic, the TowerJazz Panasonic Semiconductor Company (TPSCo).

Panasonic, like most of the rest of the Japanese electronics industry, has seen increased competition from Korea, China, and Taiwan. Consumer electronics have suffered most of all -- market share has declined and profits have been sparse. Growing vegetables in some of the most expensive buildings in the world is inventive in an odd sort of way, but it's a sign of desperation. If the Japanese electronics companies could profitably build chips in these fabs, they would. If they couldn't use the fabs, but could sell them for a reasonable price, they would. Lettuce and strawberries are a last resort maneuver.

TPSCo presents a mutually beneficial alternative to growing produce. With the formation of TPSCo, TowerJazz paid a paltry $8 million worth of TowerJazz shares for 51% of three Panasonic fabs in Japan. This was below book value, causing TowerJazz to take a gain on the purchase of $150 million. In my view these fabs have a true value of several hundred million, masked by depreciation. Also included with the transaction were $40 million and work-in-process materials, both provided by Panasonic. One of the three fabs is a state of the art 300mm fab (TowerJazz's only 300mm fab) with 45nm and 65nm analog production -- this fab alone would revolutionize TowerJazz's business. Also included in the deal is Panasonic contracts for a minimum of five years (and realistically indefinitely) with an annual run rate of $360 million-$420 million. The Panasonic contracts take up less than half of the production capacity of the JV fabs, leaving room for considerable growth, and yet are enough to leave the JV very profitable from day one. Panasonic is also licensing various of its processes to both the JV and to TowerJazz, and transferring the experienced personnel in the three fabs to the JV.

The benefits to TowerJazz are clear: increased revenue (76% growth quarter over quarter, 87% year over year), immediate profit recognition, a huge amount of new capacity, and a wider variety of equipment and processes to offer to customers. Panasonic also benefits, with the opportunity to profitably fill mostly empty fabs, which would be very difficult for them to do on their own with no experience as a foundry. The JV allows Panasonic to remove the risk of expensive capital expenditures for equipment that might remain idle.

Lean Operations

TowerJazz consists of four fabs: Fab 1 in Israel, an old 150mm facility that has seen significant improvement over the years; Fab 2, a more recent 200mm factory, also in Israel; Fab 3, a 200mm fab in California acquired when Tower took over Jazz; and Fab 4, a 200mm Nishiwaki, Japan factory acquired from Micron. TPSCo adds to that three Japanese fabs -- "Arai E", "Uozu E", and "Tonami CD". Uozu E is a 300mm fab, and the other two use 200mm wafers.

Nishiwaki (Fab 4) was acquired from Micron in 2011 along with a three-year contract extending through June of 2014. This contract was in support of a former Micron imaging business which later became the independent company Aptina (which Micron retained a 30% ownership of) which has just recently been bought by On Semiconductor (NASDAQ:ONNN). The Micron business consisted of $30 million per quarter, and this past quarter was the final quarter of the contract.

With the cessation of the contract, TowerJazz has lost around $120 million in Micron revenue. TowerJazz is adjusting for this by closing Nishiwaki which will eliminate $132 million in fixed costs. Nearly all non-Micron work being done in Nishiwaki has been transitioned to other TowerJazz fabs. The cost savings is beginning in the current quarter, and is expected to be fully realized by next quarter.

New Growth

Although the Panasonic JV is the most obvious avenue of growth, organic growth has also been strong. In the most recent quarter the top five customers by revenue grew 50% as compared to the revenue from them in the year ago quarter. The top ten customers had a growth of 40%. TowerJazz expects this trend to continue, with a 47% annual increase from the top ten customers in the third quarter.

Radio Frequency

As I mentioned in my article about RF Micro Devices, a 2G phone has around $2-$3 of RF content, while a 4G phone has up to $10-$12 of content. China, the world's largest smartphone market, is in the midst of a massive LTE build out, switching from a predominantly 2G market to 4G. TowerJazz is seeing growth in both the wireless and network infrastructure sides of the RF device market. From the last conference call:

"We have been gaining significant market share as technology has shifted from gallium arsenide, a technology which we do not serve, to RF CMOS, RF silicon-on-insulator and silicon-germanium, for which we are technology leaders. In addition, the market itself is growing faster than the overall semiconductor market"

Major RF component manufacturers such as Qualcomm, Skyworks (NASDAQ:SWKS), RFMD, and Avago (NASDAQ:AVGO) are TowerJazz customers.

Image Sensors

Image sensors have historically been a strong business for TowerJazz, and in its most recent call TowerJazz spoke extensively about its new opportunities in image sensors thanks to TPSCo. It can now supply cutting edge sensors from 300mm wafers at 65nm with 1.12 micron pixels. TowerJazz also serves various niche areas, with several new customers that will be producing x-ray sensors in TPSCo.

Other Areas

This is in fact a good time to be in the foundry business in general as capacity increases are not keeping up with demand increases. During a recent Himax call, CEO Jordan Wu said:

"On top of that we all know 8 inch capacity is very, very tight." A few minutes later he then reiterated, "So having said all of that this very moment the 8 inch capacity it's very, very tight."

Many investors fixate on the newest cutting edge production technology and are concerned with who is ahead at the latest production node on the largest wafers. However, most of the chips we use are produced with old equipment on smaller wafers in order to keep costs low. No one is building new 8 inch or 6 inch (200mm or 150mm) wafer fabs these days, but those in existence are still useful for less demanding chips because the equipment is already paid for, so the cost to produce a chip with it is minimal. There are many areas where a specialty foundry can put capacity to use. This is demonstrated by the number of new mask sets that TowerJazz receives from customers. Two to six quarters after a design win, customers send mask sets to TowerJazz. Three to six quarters after receiving those mask sets, volume production begins. In the second quarter TowerJazz received 242 mask sets, up from 130 mask sets in the second quarter of last year. Production from those customers should ramp up throughout 2015.

India

TowerJazz and IBM are being contracted by the Indian government to construct a $5.5 billion fab in India, which TowerJazz will run for five years. According to one Indian news report:

"According to Calcalist financial news website, TowerJazz, which will provide know how, consulting and support services, will have revenue of Rs.1,630.43 crore ($300 million) over six years from the project. It will not have any expenses and therefore will have a profit of 90% of the revenue."

I haven't included this in my projections at this time, but will keep it in mind as a large long-term opportunity for TowerJazz.

Modeling TowerJazz

Now that we've looked at the business, let's build a model for it. TowerJazz is a bit messy to model, and perhaps that explains why it's at a level that I believe to be extremely cheap. Let's look at the most recent statement, line by line.

(click to enlarge)

Revenue: 2Q14: $234M, up from $132.7M in 2Q13. 3Q14 is projected to be $225M +/- 5%. The second quarter was the last quarter with Micron revenue of $30 million, so this decrease of $9 million from the previous quarter means that non-Micron revenue is increasing by $21 million quarter over quarter. Remember that something around $90 million to $100 million is from Panasonic in the TPSCo JV.

Cost of revenues: 2Q14: $227.3M (GAAP), $171.7M (non-GAAP). non-GAAP excludes depreciation and amortization which seems a bit like cheating to me, but that's fine, we'll account for that. However, one positive to excluding depreciation is that it's not a realistic number. In the third quarter, TowerJazz is expecting depreciation of something like $45 million, but they are only spending $20 million per quarter on capex, so the depreciation is misleading. Depreciation is expected to trend down linearly to the $20 million-$25 million level over the next few years. 3Q14 will see some of the benefit from the closing of Nishiwaki, and 4Q will see all of the benefit. Nishiwaki's closure will result in a fixed cost savings of around $132 million per year, or $32 million-$33 million per quarter.

Interest expense: 2Q14: $8.8 million. TowerJazz received some interest on $192 million in cash, but then paid more than $10 million on high yield convertible bonds. TowerJazz hopes to refinance its debt that is at or near 8% which should decrease the interest payments. Also, the bulk of the debt comes due in 2015 and 2016, so the interest expense is relatively temporary.

Tax: TowerJazz has accumulated a massive $1.3 billion in net operating losses in Israel, which means that Fabs 1 and 2 won't be paying taxes for the foreseeable future. Fab 3 in California and the TPSCo fabs in Japan should pay something similar to normal corporate rates. The Israeli NOLs are not kept on the balance sheet, and the deferred tax liability on the balance sheet of $100 million is just an accounting artifact (explained on the second quarter call), not a real liability.

Non-controlling interest: Panasonic owns 49% of TPSCo, diluting its profits.

Shares: 59.8M (GAAP), 50M (non-GAAP). There's a mess of warrants, convertible notes, and options. The convertibles mature in December, 2015 and 2016, with a smaller number also maturing in 2018. It would seem to be impossible to buy back the 2015 convertibles due to an Israeli law that prohibits it unless there's been two consecutive profitable years. Conservatively assuming that most of the convertibles turn into shares (excluding the 2018 notes) you get a share count of around 86 million. See pages 48-51 of the 2013 20-F for details.

Bringing It Together: The second quarter had EBITDA of $33 million. The third quarter is expected to have $9 million less in revenue, but partial cost savings from the closing of Nishiwaki. Let's conservatively assume that the fourth quarter has the same revenue as the third quarter (even though management has guided for continuous organic growth), but full cost savings of around $32 million. That would bring fourth quarter EBITDA to $56 million, and on an annualized basis, $224 million on $900 million of revenue. Let's assume that TowerJazz had $135 million of revenue in this past quarter, while TPSCo had $99 million. That is about to change, as we can soon expect TowerJazz's Fabs 1, 2, and 3 to be full. From the most recent conference call:

Unidentified Participant

"Right. So when the non-Panasonic Fab's are full, they are doing $160 million of revenue per quarter and your thinking that possibility for reaching that level of revenue would be how far away?"

Russell Ellwanger - Chief Executive Officer

"Again, we don't provide long-term guidance. But as I stated, that those masks would be reaching their peak volumes in the, you know its 2015, 2016 time frame. I would assume that I will be at our target anyway. It's to be running at full utilization in the Q1, Q2 time frame of next year."

That's $25 million per quarter, or $100 million in additional annual revenue for the year at very high incremental margin, in the original three factories which have no non-controlling interest. Two of the three pay no taxes. I believe the incremental margin could be as high as 75%-80% because the incremental cost is so low -- much of the cost is fixed, meaning equipment, employees, etc. that are there with or without this new business and are already paid for. CFO Oren Shirazi gave this example when talking about the quarterly Micron business in Nishiwaki:

"So just for an example, basing on the previous numbers of $30 million revenue a quarter, so against that you have approximately $10 million of variable cost plus the $32 million fixed cost that you mentioned. So the total is $42 million actually."

Just over a quarter of the cost there is variable. Another example, here's an old Morgan Stanley report. Note how the revenue scales from one year to the next, and then note how the incremental cost scales.

If we say that $100 million in new annual revenue in the original fabs has 75% incremental margin, that's an extra $75 million in annual EBITDA starting early 2015, bringing it to $299 million in annual EBITDA on $1 billion in revenue.

Added to that, we know that in the first 3 months alone, TPSCo signed up more than $100 million in annual contracts which will start in the second half of 2015.

"Within the first three months post launching of this venture, we signed strong contracts and entered business agreements with multiple third party Image Sensor companies and top-tier integrated device manufactures, which should reach annual run rates well about $100 million of high margin revenue, on top of the previously disclosed $360 million to $420 million forecasted TPSCo annual revenues from the Panasonic business."

That puts us at $374 million in EBITDA in early 2016, though I subtract around $62 million for non-controlling interest, leaving $308 million. Interest will disappear as the convertible notes are converted, and tax is not applicable to the money made in Israel, so both the "I" and the "T" in EBITDA are not overly onerous. Depreciation is a non-cash item and is declining from its current level, as mentioned above. For a company with a fully diluted market cap under a billion dollars, that's pretty fantastic.

Risk Factors

TowerJazz's risk factors are relatively few due to the captive Panasonic contracts. Electronics are a cyclical industry so TowerJazz bears the same macroeconomic risks as its fellow chip companies. Also, should competitors decide to aggressively increase capacity, this might hurt TowerJazz's pricing power in the long run.

Conclusion

Ascendiant Capital Markets covers TowerJazz with a price target of $24. They use some different assumptions than I do regarding share count and revenue, but end up with an EPS number of $1.93 for 2015. Although I use a different valuation method, I believe their price target and EPS numbers are in the right ballpark. I believe that a price target in the mid-$20s is reasonable, if not conservative -- the projections I've used above still leave $100 million or more in spare capacity for TPSCo, which is new and is likely to expand. There's also the possibility of finding a use for Nishiwaki, or for the tide of money that I believe TowerJazz will see, and the longer-term benefit of the Indian-government sponsored fab.

I believe that the strength of TowerJazz's new joint venture, and of its business as a whole, make it severely underpriced, with a relatively low risk thanks to the new guaranteed contracts from Panasonic. The degree to which TowerJazz is underpriced is hidden by the complex transformation at hand, but I believe TowerJazz's new advantages will become increasingly obvious over the next few quarters. This is the best bargain I see on the market today, and I think it will see better than a 100% return over the next 12 months.