Four years ago, senior technologists Ng Meng Thai, Cheah Hun Wah and Tan Chun Chiat got together to start a company that designs integrated chips (ICs) for semiconductors — Oppstar Technologies Sdn Bhd. The trio had been colleagues at technology giant Intel and had about 75 years of working experience between them.
They felt that the market was ripe for a contract design company after more than 40 years of the electrical and electronics (E&E) industry being in Penang and a proliferation of contract manufacturing companies such as Globetronics Technology Bhd, Inari Amertron Bhd and Flextronics Technology (M) Sdn Bhd.
The three of them had been partly responsible for getting Intel to transfer some of its IC design functions to Malaysia. The tech giant had wanted its Penang office to stick to manufacturing, but the trio had been part of the pioneer team that convinced the senior management in the US that Malaysian engineers were more than capable of taking on some of the design function.
Ng, who is CEO of Oppstar, says there were already companies here doing contract design, but it was the “warm body” model, where they simply provided CVs for the larger companies to pick from. Thus, the companies acted more like employment agencies.
The weakness here is that if the multinational corporation (MNC) likes a particular engineer, it usually hires him or her, leaving the less talented ones behind. Because of this, local companies can never build their competencies.
The next type of model is task or outcome-based where the MNC defines the task and the IC design company figures out who to assign it to and how to go about doing it. “The client does not have to worry about who designs it and does not have to manage or monitor them,” says Ng.
The third model — and the most ideal — is the offshore design centre. “Instead of doing the project in our client’s office, we do it in ours,” says Ng.
He adds that it is timely to go into the market because more and more MNCs are outsourcing their design function. In the past, most companies handled everything themselves, from the definitions and design to hardware and board manufacturing to marketing. Now, most of them prefer to focus on their core competencies and outsource everything else.
“IC design costs continue to rise. This is a very expensive industry and outsourcing is going to be the trend. There are a lot of opportunities, not only in Malaysia but the rest of the world, for companies to tap into,” says Ng.
How expensive? “To design, from definitions to silicon, we need about 25 to 30 types of tools. The cheapest tool per user per year is US$25,000 to US$30,000. The most expensive is about US$1.2 million to US$1.8 million per copy,” he says.
When the three of them started out, they had the brains, capabilities and experience but not much capital. So, they opted for the “poor man’s investment model”, where they would send their guys to work in the client’s office. The client would supply the tools while they supplied the talent.
Oppstar had to start from scratch and build up a track record in the market because although the founders were experienced, nobody had heard of the company. It started with only two engineers (in addition to the founders) and took on very low-level tasks.
Ng points out that there are 7 to 10 levels of IC design and Oppstar initially took on the easiest — layout artwork. Its strategy was to assign engineers who could do so much more. Once the client realised that the ones doing its layout artwork were capable of much more, it assigned more complex tasks. That was how the company grew.
“Our first client was Altera Corp. It needed someone who knew about FPGA (field-programmable gate array) design locally. At the time, Altera’s contractors were mostly regional guys and it was worried about immigration risks, especially if the laws changed. By building up a local source, it was actually mitigating its risk,” says Ng.
From simple layout artwork, the company moved on to auto-place and route, which uses sophisticated software to place hundreds of thousands or even
millions of transistors (switches) properly on a chip and route or wire-connect all these switches according to a mega design.
The company progressed slowly and when it was given more tasks, it hired more engineers. Basically, building a chip is like building a house with 7 to 10 different disciplines, says Ng.
“When you build a house, you have someone handle the plaster ceiling, another person handle the wiring and yet another handle the plumbing. And the guy who handles the plaster ceiling knows nothing about the plumbing,” he points out.
At the beginning, it only hired very experienced engineers to carry out the tasks and this strategy bore fruit soon enough. Oppstar had planned on expanding abroad two to three years after it was set up, but it was this competence that speeded up the process.
It was doing a job for Altera, which was one of the earliest customers to use the excess capacity of Intel’s wafer fabrication foundry in Kulim. It was a fairly involved task, using a 14nm process technology (by comparison, Mimos Bhd’s process technology is 350nm and Malaysia’s wafer fabrication foundry, Silterra Sdn Bhd, only goes down to 110nm).
“We worked on that project and found a lot of methodology and process complications. We gathered the key learning points and compiled them into 50 PowerPoint pages for Altera which, in turn, provided the feedback to Intel,” says Ng.
That got Intel’s attention. Who was this upstart? “Intel decided that we were a serious company and put us in touch with one of its customers in China that was facing the same implementation issues as Altera,” says Ng.
Spectrum Technologies China, Huawei and LG were all exploring the possibility of using Intel’s foundry and Oppstar found itself in a unique position to help. “We thought we would have to go knocking on doors to get overseas customers. No, they came to us,” says Ng.
The company has grown by leaps and bounds. It has 65 engineers and plans to increase that to 300. “In this industry, size matters. If you are too small, you get bullied. Our competitors are regional design houses. They are not just competing for the projects but also the resource pool,” says Ng.
Oppstar now has experience in 7nm process technology (the most advanced in the world) and its engineers work on state-of-the-art projects in everything from the latest mobile phone technology and the Internet of Things to autonomous cars. The company is able to attract the right kind of talent because the engineers get sent out on cool projects all over the world — China, the US, South Korea and Japan.
“We support Fortune 500 customers locally and abroad. We started venturing out from Malaysia in early 2016. We actually send our engineers to work with our customers’ design team. Most of the products are so confidential that they cannot leave the facility,” says Ng.
In the early days, Oppstar provided staff and had no involvement in the design process. Once it started building its track record, it moved on to outcome-based tasks, working with its customers’ tools to deliver certain functions or projects for them. Now, it handles complete intellectual property (IP) turnkey projects.
“Ideally, we would have our own tools and do it in our own office. Of course, we would need to put up a lot of firewalls and each customer would require total isolation. Our dream is to have a large campus with a separate building for each customer, like those we saw in India, with Infosys and Wipro,” says Ng.
The company’s office is housed within Collaborative Research in Engineering, Science and Technology’s (Crest) headquarters in Bukit Jambul, Penang. In its four years of operation, it has already seen two “harvests”, or filings for IP.
On the financial side, because of the nature of its work, its cash flow cycle is very fast and billings are fairly lumpy. As Ng points out, the company is anything but cash flow tight. However, it needs sizeable cash injections to buy the tools of its trade as well as to hire top-notch engineers and beef up its team.
Today the company has 65 staff and it intends to increase this to 300.
He says the company has been approached by both regional and international players. “The way they talk it is as if they want to collaborate. But I know that once they get in, we may get swallowed up. We are not ready for that. Anyway, I prefer the initial public offering route to being bought over. Give us two years. Three years tops.”