CEO interview: What’s next after Tower’s turn-around?

CEO interview: What’s next after Tower’s turn-around?

Interviews |
By eeNews Europe

Q1 What were the big changes at Tower over the last ten years?

“There have been multiple phases of change. The first big change was the drive to become profitable,” said Ellwanger.

“When I arrived, the company wasn’t market driven. The company had $100 million in sales and negative EBITDA of $20 million and negative $55 million cash flow from operations and was $550 million in debt [EBITDA is earnings before interest, tax, depreciation. and amortization].”

“Before I arrived it had a central R&D group and design center in Israel and a sales office in Santa Clara, California. We had to decentralize the product lines and give each line a general manager. We targeted a positive EBITDA by the fourth quarter of 2005 and positive cash flow by 4Q06. The teams worked incredibly hard and we did hit those two financial goals.”

Ellwanger explains that the second phase was dropping digital work and focusing the company on analog, RF and sensors. Over the years 2005 to 2008 things gradually improved for Tower but aside from CMOS image sensors the company was still generally a second or third source of digital circuits manufactured using 130nm or 90nm CMOS. SanDisk, for whom Tower made flash memory controller ICs, was the biggest customer. “So we told SanDisk we would have to raise prices.”

This was not an easy conversation because Eli Harari, chairman and chief executive officer of SanDisk at the time had also been a director of Tower and instrumental in bringing in Ellwanger.

Next: All that Jazz


“And in 2008 we acquired Jazz Technologies Inc., which was experienced in mixed-signal, power management and RF and ran RF-capable CMOS, BiCMOS and silicon-germanium processes in its wafer fab in Newport Beach, California.”

“We were able to take two $200 million [annual revenue] companies and take out a lot of redundancy and boost buying power with suppliers. There was almost no overlap in products and we achieved a $100 million reduction in costs. 2009 was a very bad year for the industry but for us TowerJazz was a very effective merger and in 2010 we had $500 million of revenue,” said Ellwanger.

The wafer fab at Newport Beach, California. Source: Tower Semiconductor.

The third phase was the creation of the TOPS business. TOPS stands for transfer, optimization and development of process services. This is where Tower works with IDMs that don’t want to commit to the cost of adding manufacturing capacity. “Tower agrees to take in their process and use it exclusively for them to fulfil long-term manufacturing contracts. We have worked with Vishay, Fairchild and International Rectifier. It involves IP transfer but behind a firewall.”

The long-term nature of those contracts in areas such as power MOSFETs and discrete parts provides both Tower and the TOPS customer with price stability during the chip industry’s boom-bust cycles. It also removes the year-on-year pricing pressure found elsewhere in the industry.

Next: The final phase



The final phase in the rejuvenation of Tower was the acquisition of additional manufacturing capacity in the eastern hemisphere. This has resulted in the creation of the TowerJazz Panasonic Semiconductor Co. Ltd. – which operates three former Panasonic wafer fabs and in which Tower holds 51 percent (see Tower buys three wafer fabs for $8 million).

These fabs are about 50 percent loaded fulfilling a $400 million per year, five-year contract for Panasonic, said Ellwanger. This means the remaining capacity is available to be filled in a highly profitable way. “This includes a 300mm wafer fab with a world-class 65nm CMOS image sensor flow,” Ellwanger added.

Q2 What were the most significant changes in the industry in the last 10 years?

“Mobile platforms which encompass three mega-trends: green energy consumption, wireless everything, and smart sensors,” says Ellwanger. These are all areas in which Tower operates from CMOS image sensors, through antenna switches to power management ICs.

“In addition many IDMs have had to move up to offer system-level reference designs and chipsets. This means that they need to buy in analog manufacturing and even design capability. Intel is not a fab-lite company but they need lots of chips beyond their principle digital expertise. We recently announced that we are manufacturing a near infrared light sensor used for depth sensing within Intel’s RealSense 3D camera.” (see Tower makes IR sensor for Intel RealSense)

“And of course the fab-lite trend feeds the foundries. It started in the digital domain but now it is the same with analog. In the pure-play digital foundry world no one can compete with TSMC, which has billions of dollars to invest. But in the analog space it is an even playing field and much lower cost.”

Next: But what about Israel?



Q3 Even with approaching a billion dollars in annual revenue Tower doesn’t have much capital to re-invest in mature wafer fabs. What happens to Tower’s footprint in Israel over the next ten years?

Ellwanger immediately makes the point that its two Israel fabs are doing profitable business for Tower and that, if he could, he would add to the Israel portfolio, through acquisition.

“The 6-inch factory is running with nice utilization. The capex budget is tiny, just sustaining capital. We have technologies that have no advantage or cannot be done at lower nodes or do not need extremely high volume, which suits a 6-inch fab.” Ellwanger also makes the point that serving IDM TOPS business and Tower’s full service design capability – where it designs ICs for and on behalf of IDMs on condition that it manufactures them – are well suited to the 6-inch fab.

Tower’s manufacturing complex at Migdal Haemek, Israel. Source Tower Semiconductor.

Bigger wafers do bring economies of scale, said Ellwanger, but only up to a point. “A mask set at 300mm is more expensive, 300mm production tools are more expensive and maintenance is more expensive,” he said.

Next: Whats made where?



In Israel the 6-inch fab is making discrete transistors, power semiconductors, high-speed logic and power management ICs up to 700V. At the 200mm facility in Israel Tower runs RFSOI for antenna switches, power management on BCD and CMOS image sensors, he said.

“We still have area to build out in our 6-inch and 8-inch factories. There is capacity expansion that can be done where a $3 million to $5 million spend gets you an additional 1,000 wafers per month.”

Q4: Tower is part of a consortium with IBM dealing with the Indian government to put down a fab in India What’s happening there?

Ellwanger says there is nothing to report and it is up to the India government to make announcements with regard to the progress and timing. “It is not yet part of our detailed business plans. But if and when it does happen it will be highly accretive to our bottom line.”

Q5: What will be the next significant technology direction for Tower?

“Well, we already do MEMS. We are doing a tuneable antenna for Cavendish Kinetics (see Cavendish RF MEMS released to manufacturing). We also make a MEMS resonator for SiTime as a replacement for quartz timing devices, and we do some small array microphone MEMS,” said Ellwanger.

“But the problem with MEMS is there are no PDKs [physical design kits]. It’s a process-per-product technology. To grow to be a leader in MEMS you need to take say 30 to 40 customers so that you can get some products that go to really high volume. But the trouble with that is that MEMS development is a non-standard, lengthy procedure that can really kill your fab cycle time.” Ellwinger is more bullish about markets adjacent to image sensing such as infrared for thermal imaging for security applications.

Next: More technologies?



What about packaging and multi-die components?

“We’re not really involved in packaging. We have partners and we have access to packaging capability through Panasonic. As to 3D packaging, I am not convinced about the value proposition there. IPDs – integrated passive devices – are interesting and we’re looking at the possibility of making silicon interposers.”

At work in Tower’s wafer fab. Source: Tower Semiconductor.

And in the area of power, what about gallium nitride?

“GaN: well there might be a demand one day. Right now the amount of GaN produced is very small compared with the volume of planar, trench silicon MOSFETs. In fact we do have GaN capability. We manufacture something for Panasonic but we do not offer the process as a foundry flow. But there is not yet much foundry demand.”

Next: No more Moore’s Law?



Q6: When will Tower need 32/28nm CMOS and how will it gain access to it?

“I am not sure it ever will. 0.18-micron is state-of-the-art for many analog applications. TowerJazz Panasonic Semiconductor does have a 32nm process in R&D but it is digital.” Ellwanger said he sees no need for Tower to go back into digital IC manufacturing.

Q7: How does Tower address China?

“We have very good sales and application team in China. Some Chinese customers want to work with us; they don’t want to fab inside China because of IP risk.”

At the same time there is a trend towards more IC production within China. “China can move faster than India. But you would be keen to spread out the opportunity to avoid capex. Another model would be for Tower to act as a technology provider,” said Ellwanger.

When asked if he was considering entering China Ellwanger said: “More than just considering; actively looking at the opportunity.”

Q8: At this time we are seeing lots of mergers of semiconductor companies. Is that something that affects foundries and could it be a case of acquire or be acquired?

“We have seen some; 1st Silicon and X-Fab, Grace and Hua Hong. This will continue to happen. But the real opportunity is where a system-level company wants to exit chip manufacturing,” responded Ellwanger.

He adds that at the time Tower was making its deal with Panasonic another systems company wanted to participate in a larger, three-way venture. It was decided to focus on the deal with Panasonic in order to keep the negotiations simple and tractable but the possibility is still out there, said Ellwanger.

Next: In conclusion



So will Tower be getting more Japanese legacy wafer fabs? “Maybe we will and maybe we won’t. But that’s where there is opportunity: a system company that doesn’t see semiconductors as part of their core value,” said Ellwanger.

Whether deals in China or Japan materialize or not, Ellwanger is expecting in the fourth quarter of 2015 Tower will be on an annual run rate of $1 billion in revenue, $250 million EBITDA and $220 million cash flow from operations and with about $60 million of debt. Which should set Tower up nicely for the next ten years.

Related links and articles:

News articles:



CEO interviews updated


Tower buys three wafer fabs for $8 million

Tower makes IR sensor for Intel RealSense



Cavendish RF MEMS released to manufacturing



Tower makes military MEMS for Physical Logic

Startup launches record resolution 150Mpixel sensor



India to build billion-dollar analog wafer fab



Austria’s AMS is shopping for a wafer fab


If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles