Set-top box and home gateway are application areas where STMicroelectronics has already designed ICs for implementation in the FDSOI process that it has championed in the industry, albeit with limited take-up so far.
However, the news that ST is withdrawing from those markets (see ST exits STB chip business, plans lay-offs) and will make no further designs for that sector should not be seen as a blow to prospects for FDSOI, Chery told eeNews Europe. Indeed Chery says ST is preparing to make microcontrollers on 28nm FDSOI with options on an embedded non-volatile memory still open, but with phase-change memory as a leading contender.
Of the re-organization that eliminated the digital products group that Chery had led, he said: “This absolutely does not impact FDSOI. What is important is that, from the first figure of merit, FDSOI can provide processor cores, usually ARM processor cores, at very low operating voltage. The second figure of merit is in RF for communications. Together this makes FDSOI perfect for the Internet of Things.” He added that its radiation immunity also makes it well suited for automotive applications as well as for space and defence ICs. “It has already been announced that the EyeQ4 processor from Mobileye NV will be in FDSOI and NXP/Freescale is using FDSOI for ADAS (autonomous driver assistance systems),” he said.
Chery emphasized that, following the announcement of ST’s withdrawal from STB and home gateway markets and of a proposed redeployment of 600 engineers, the company is now focused on automotive and Internet of Things applications and that therefore FDSOI is a core manufacturing process. Indeed it could be argued that moving engineers familiar with FDSOI from the STB group into MCUs and automotive will help to proliferate the technology through the company.
Next: MCUs going to 28nm FDSOI
The reason for the exit from STB is excessive competition in Asia from the likes of M-Star and also in the US while the R&D costs to serve that market remain high as they demand leading-edge and highly integrated circuits, Chery said.
On the microcontroller front Chery said: “We have announced our first MCU on 40nm bulk CMOS ramping over the next few quarters. We are developing the next node on FDSOI at 28nm. For flash [non-volatile memory] we have a couple of options. Clearly the best choice is PCM [phase-change memory]. We are looking forward to very ultra low power MCUs.”
Moving MCUs over to FDSOI at the 28nm node is an approach that is also being considered at NXP Semiconductors NV.
It may seem rash to nominate PCM as an embedded memory option when stand-alone PCM has proved problematic for decades and is continuing to cause problems for Intel and Micron in its latest incarnation known as 3D-XPoint memory (see 3D XPoint steps into the light). “We have another option,” said Chery but he declined to state whether that was resistive RAM or perpendicular magnetic RAM, two options being considered by NXP (see NXP embraces 28nm FDSOI for MCUs).
But despite such plans for the medium term, the company has hit a downward spiral of declining sales, which could be hard to break.
ST announced fourth quarter 2015 sales were $1.67 billion, down 8.8 percent year on year and the outlook for 1Q16 sales is for a 3 percent sequential decline. Although ST outsources much of its high-end digital manufacturing it retains manufacturing capacity for mixed-signal, specialty processes and such things as MEMS. Low utilization of that in-house manufacturing capacity is contributing to pressure on profit margin.
Carlo Ferro, ST’s chief financial officer, said in an analysts’ call to discuss the financial results: “We need $1.8 billion per quarter for the manufacturing machine to be comfortable.”
So where will those increased sales to turn back the tide come from for ST?
Next: Better than M&A
“From all three of our [new] business groups,” said Chery. He said that transferring engineers from STB design to MCUs would allow the company to proliferate parts, create more software and development boards and accelerate design wins.
“The second growth driver is sensors although that is still in transition mode but with good confidence to get back to growth,” said Chery (see Updated: ST sees results of MEMS diversity thrust). “The third is specialized imaging sensors; the fourth is BCD manufacturing process and another is digital for automotive.”
Going back to microcontrollers where STMicroelectronics has been a force, it is noticeable that recent mergers such NXP, Freescale and Microchip, Atmel, will contrive to push ST down the ranking of suppliers. The merger and acquisition spree that the industry has been on has largely passed a pre-occupied ST by. Does it now need to engage with that to regain some critical mass in the automotive, industrial and IoT sectors it wants to pursue?
In answer Chery said that ST has looked for and continues to look for M&A opportunities but that transferring 600 engineers into the MCU and digital business unit is a similar yet superior approach. “This is much better than a merger. We are moving 600 people in R&D and product support but without the issues of managing corporate culture. Basically we are transferring resources from difficult markets to growth markets.”
Related links and articles: