MIPS-Chipidea merger's goal: 'virtual' SoCs

New York -- MIPS Technologies Inc.'s $147 million acquisition last week of Chipidea Microelectronica S.A., a privately held supplier of analog and mixed-signal intellectual property, advances both companies' ambitions for transforming system-on-chip design.

In today's typical SoC design cycle--which can take a year--the designers decide on a processor and operating system, think through the architecture and work on proprietary technology. But they often postpone decisions on analog components until the eleventh hour, creating an integration gap that leaves analog designers with a host of customization chores at the last minute. The resulting delays can stand in the way of finishing the SoC. "I've seen many customers--even large semiconductor companies-- unable to migrate to the next node, or having to skip an entire generation of technology," John Bourgoin, MIPS' president and CEO, told EE Times.

With the combined forces of MIPS and Chipidea (Lisbon, Portugal), "for the first time in the industry, customers can see a full pool of technologies--the beginning to the end--to design their SoC," said Chipidea CEO José Franca. "This could be a very powerful proposition."

MIPS' bailiwick, of course, is processor cores; Chipidea supplies analog and mixed-signal IP for the wireless, digital consumer and connectivity markets. Gartner Group analyst Christian Heidarson observed that while "MIPS has had a very successful year," it has "still looked like a bit player next to ARM." With the Chipidea acquisition, he said, MIPS strengthens its IP arsenal to mount a challenge to its dominant rival.

The deal is "huge news for the IP industry," Heidarson said. "The market for design IP is consolidating very rapidly. This market doesn't look good for startups. Unless you have strong patents and a genuinely unique and valuable technology, the design service business model is probably a better idea."

Bourgoin called the acquisition "an evolution of the IP business"; Franca termed it "a step toward a virtual SoC." Either way, their aim is clear: to enter the customer's SoC design process early in the cycle. Rather than follow the conventional method of inserting the analog IP as a gap filler, requiring analog design customization, they expect to offer customers standard blocks that combine digital and analog.

"If MIPS can shield mixed-signal integration issues from the SoC designer, this would be of tremendous value," Gartner's Heidarson said. MIPS should absolutely put a priority on developing complete MPU-plus-mixed-signal subsystem IP products."

In the short term, MIPS can integrate connectivity and power management drivers from Chipidea into its software platform, Heidarson added.

The deal
MIPS' mostly cash outlay of $147 million for the Portuguese company--whose revenue for the 12 months preceding the deal was $32.8 million--is significant. The companies acknowledged that they had been in talks for almost a year. They also confirmed that there had been a bidding war for Chipidea, although they declined to name the other bidders.

Chipidea has been a rising star in the design IP market. Despite the complexity of analog technologies and the risky nature of the IP business, Chipidea's revenue grew at a compound annual rate of 38 percent between 2000 and 2006.

Franca said he'd told MIPS his company wasn't for sale when the topic came up a year ago. Chipidea at the time had amassed sufficient venture capital funding to carry it through the next two years and "did not need to exit," he said. But the company ultimately saw more opportunity in being acquired than going public.

The cash deal could be a sign that the sellers--Chipidea's founders--are bailing out of the business. But that's not the case, Franca said. "We are committed to Chipidea even stronger than before."

Bourgoin said both sides had wanted a cash deal. MIPS shareholders did not want to dilute the stock, and Chipidea's VC investors--who held 65 percent of the company--preferred cash payment.

The acquisition will effectively make MIPS "the second-largest design IP vendor," after ARM, Bourgoin said.

According to Gartner Group's ranking of the top 10 IP vendors by market share for 2006 (under which only revenue from design IP was counted and patent royalties were excluded), ARM, with $440.5 million in design IP revenue, ranked first, with a 33 percent market share. MIPS, with $76.2 million in design IP revenue, or 6 percent of the market, was third. Chipidea, with revenue of $31.5 million, or 2 percent, ranked ninth.

The integration of microprocessor and analog IP is a visible trend in SoC development, and the addition of Chipidea's analog IP to the MIPS portfolio strengthens the latter's position and expands its customer base. MIPS previously had made "zero investment" in analog technologies, Bourgoin acknowledged. "That's why this is such a nice fit."

MIPS chose Chipidea both for its strong customer base and its financial robustness. Moreover, two-thirds of MIPS' customer base overlaps with Chipidea's. Those customers include Toshiba, Sharp, NXP, STMicroelectronics, Zoran, Qualcomm and Texas Instruments.

Currently, 88 percent of analog design is done internally at most leading semiconductor companies, according to Bourgoin. Nonetheless, Chipidea has successfully competed against in-house analog design teams, possibly because it has anticipated a shift to third-party licensed analog IP, much as the market has already shifted away from in-house processor core design.

"When it comes to analog, it either works or it doesn't," said Bourgoin. But for many chip vendors, analog is "not a differentiator." He thus questions the value of investing to design wireless or power management chips in-house.

"Analog is hard to do. It requires special skills," he said. "Over the last 10 years, Chipidea's 250 engineers have delivered a variety of analog IPs in GDS II format that actually work." GDS II is a de facto standard for IC layout data exchange.

"Seventy-five percent of Chipidea's business is [from] repeat customers," said Mark Tyndall, MIPS' vice president for business development and corporate relations.

MIPS hired Tyndall 14 months ago to scout acquisition targets. Bourgoin told Tyndall to seek companies involved in the control plane (such as processing cores), analog and other IP, embedded memory, and data plane devices (such as DSPs). "I told Mark any of those four quadrants would be fair game, including tools and interconnects," Bourgoin recalled.

Tyndall came back a couple of months later with a recommendation to buy Chipidea. "Analog represents [the] highest growth opportunity," Tyndall told EE Times. "Many captive semiconductor companies, including IBM, have been slowly releasing their analog work to the outside market over the last couple of years." Chipidea counts 13 of the top 15 semiconductor companies as customers, he added.

Bourgoin said he expects the companies' operational merger to be straightforward. Chipidea will become a business group of MIPS Technologies, and its product portfolio will continue to be sold under the Chipidea brand. Franca will become group president and general manager and has been appointed to the MIPS board. The sales, finance, legal, HR and marketing functions will be consolidated, Bourgoin said, but the MIPS and Chipidea design teams will remain intact and work in parallel.

Except for a few commonalities in drivers and tools, MIPS and Chipidea design teams have very different knowledge bases, with "not a lot of engineering overlap," said Bourgoin.

Gartner's Heidarson pointed to some potential bumps in the road to a unified company. "There are culture differences to consider between the digital and analog engineering operations," he said. "MIPS needs to be wary of attrition among analog engineers who might prefer to work for pure analog players."

Heidarson also suspects it will be difficult for the MIPS sales force to sell analog products effectively.

"MIPS should consider providing full-chip solutions to some customers. That could have a very positive impact on their financials," he said. But he added a caveat: If MIPS takes that route, "fabless customers may worry that it puts MIPS into a competitive position [with them]."

Would MIPS, with its expanded portfolio, venture a competitive play against fabless chip companies like Broadcom?

"Never," said Bourgoin. "[But] Broadcom is better off outsourcing analog IP to us--rather than developing their own analog IP for their SoCs, as they do now--because analog is not really a differentiating factor."

 See related image