That makes the announcement of Dynex in offering a foundry service as worrying. The company already makes a range of IGBT power devices, and now has a deal to supply a ‘leading semiconductor
That means that there is spare space in the fab, and that is not a good sign at the peak of the semiconductor cycle. Again the established wisdom is to fill up the fab with your own products. So does this mean Dynex is struggling to be successful? Although it hasn’t been hugely profitable in the last couple of years and revenues have fallen (from $46m to $40m, although the exchange rate hasn’t helped), it has grown at the site from 351 to 368 people in the first nine months of 2017.
Now the power business tends to do things its own way. The fabs are smaller than their CMOS cousins and are fully paid for and running specialised processes that are optimised to particular types of devices. You can’t run standard CMOS in these fabs, and you can only run a limited number of different styles of power device without some hefty, and expensive, process development. All this limits the options for foundry customers to differentiate their devices from those of Dynex.
Dynex is a Canadian company with a fab in Lincoln, UK, that carries 59 years of British heritage through Marconi MEDL and GEC Plessey Semiconductors. Since 2008 Dynex has been part of a Chinese conglomerate.
That conglomerate, Shanghai-based CRRC Times Electric, is the key.
Next: The Chinese connection
It buys a substantial part of the output of the Lincoln fab and so as well as the owner is the major customer. If CRRC isn’t able to take more product, which is a surprise given the boom in infrastructure modernisation across its home market, then the deal seems to be to offerup the extra capacity to foundry customers. Then the foundry parts will have a different design to the mainstream Dynex parts that CRRC uses, so they are not actually competing.
This leaves Dynex stuck in the middle. It can try to expand sales of its standard products, thereby selling to competitors of CRRC, or offer up its own process technology to its own competitors via the foundry. That’s a tough line to walk. If prospects improve and CRRC takes more product there’s less capacity for foundry customers. When the downturn comes (and Dynex management say that’s the case now, in contrast to the rest of the industry) there’s less product business and less foundry business, a double hit. This approach has been tried many times before with limited success.
So the new boss at the foundry division, Mark Kempton, Business Unit Director for Semiconductor Devices, has his work cut out walking that line, but is upbeat of course.
“The formation of the Dynex Foundry Services business is a logical expansion of the company’s commercial product and service offerings,” he said “The receipt of this first external order for high voltage IGBT die builds on decades of high-quality semiconductor fabrication experience at Dynex. The excellent chip performance, confirmed by the customer’s own testing, serves to highlight the quality of our design and manufacturing. We expect this business to grow into a substantial part of our operations over the coming years as we expand our customer base.”
There’s no doubt about the quality and design expertise, but the history of the hybrid product/foundry business model does not bode well.