High-tech

Due Diligence: SuperH Inc

By The Chilli analysts

1. Introduction
2. Hitachi's stealth architecture
3. Birth of SuperH Inc
4. Portfolio
5. Succeeding in semiconductor ip
6. Surgery required

1. Introduction

"We want 30% of the SoC cores business by 2004," an executive was quoted as saying at the launch of SuperH Inc in April 2001. If only it was as simple as copying a play from the ARM game plan of several years ago. The Chilli takes an in-depth look at SuperH, a recent entrant to the semiconductor intellectual property (IP) business, backed by Hitachi & STMicroelectronics (ST). Just where is it heading?

2. Hitachi's stealth architecture

Back in the late 1980s, Hitachi, as a leading member of Japan Inc, had a booming semiconductor business, based on memory devices and 8- and 16-bit microcontrollers. At the end of the decade, Hitachi executives saw a need for a product line to power the emerging range of digital consumer products. This led to the development of the SuperH (SH) 32-bit reduced instruction set computer (RISC) processor family. The first variant, the SH1, was launched in 1992.

The SH1 featured a 16-bit instruction set architecture, reducing the memory footprint, and cost, of an embedded system. Further variants were developed over the years adding extra features such as cache memory (SH2), digital signal processing (SH-DSP), memory management unit (SH3/SH3-DSP) and superscalar execution & a floating-point unit (SH4).

The SH1 was designed into consumer products such as digital cameras, but the key moment came when Sega selected the SH2 for its Saturn games console, using two devices per console. The volume shipments from this design propelled the SH architecture up the rankings, making it one of the leading 32-bit embedded processor architectures. The SH3 powered most of the early handheld PCs (HPC) running Microsoft Windows CE. The SH4 was selected by Sega for the Dreamcast console, and also became the de-facto standard for car multimedia and telematics.

Following the release of the SH4, a number of factors, including the industry downturn, customer demand for a credible second source and the high investment required in technology, led Hitachi to seek a partner with which to share future development costs and proliferate the architecture. ST, realising that bringing the proprietary Chameleon technology to the market to replace the ageing ST20 was a non-starter, were also looking for a partner.

The 64-bit SH5 core was jointly developed by Hitachi & ST, with both organisations contributing resource and IP. The instruction set architecture (ISA) would remain the property of Hitachi, and both companies would use the core for a range of standard products, application-specific standard products (ASSP) and custom application-specific integrated circuits (ASIC).

The SuperH range, as it stands today, can be summarised as follows:

• Spanning 20 - 714 MIPS performance range
• Instruction set enhancements for 3D graphics, floating-point arithmetic and digital signal processing (DSP)
• Wide range of standard chip products with a mix of peripherals
• Strong position in automotive, telematics/car multimedia and consumer products
• Powering over 4600 different end products
• Over 210 million SH1 - SH4 devices sold to date
• Alliances with key platform software vendors including Microsoft, Wind River Systems, etc
• Well supported by hardware and software third parties for application middleware, development tools, companion chips, etc
• Global sales and marketing channel including both direct sales and distributors through Hitachi and ST

3. Birth of SuperH Inc

During the mid to late 1990s, Hitachi had provided several vendors with access to specific SH cores; Epson, Nippon Steel and Sony. VLSI Technology (later acquired by Philips), had rights to use the SH4 core, although this was not exercised in a commercially available product.

During 1999/2000, both Hitachi and ST were considering the idea of making the SH architecture available for licensing, with an eye on the IP licensing business enjoyed by ARM and MIPS. According to Gartner Dataquest, the market for semiconductor IP is estimated to be worth $2.6bn in 2004, with processor IP making up $900m, based on license fees, royalties and professional services.

SuperH Inc was founded in 2001 with backing from Hitachi and ST, to target consumer (set-top box, personal video recorder, etc), pocket multimedia (personal digital assistants, smartphones, etc), automotive (engine control, telematics and car multimedia) and networking (voice over IP) applications, specifically those requiring a 60:40 mix of general-purpose processing and DSP. According to figures from ST, Hitachi contributed $37M of cash for a 56% stake. ST contributed $14.9M worth of cash and $14.1M of IP for a 44% stake. Both parents have committed to provide further capital, and patents relating to the SH4, SH5 and SuperHyway bus have been assigned to the company.

The original executive team consisted of Toshimasa Kihara from Hitachi as ceo, with Jean-Marie Rolland, formerly r & d director of core development at ST, as coo and Rick Chapman, formerly a director of business development at ST, as vp sales and marketing. Rolland replaced Kihara as ceo during 2002.

SuperH has offices in Bristol, San Jose and Tokyo, with approximately 100 employees, with 80% in engineering and the balance in sales, marketing, finance and administration. The headquarters are in San Jose, ostensibly because "The collaboration that produced the SH5 was centred there, so it was a logical place to headquarter the subsequent spin-out company", according to Rolland.

The board of directors has six members, and according to Chapman, are "equally divided between Hitachi and ST, with Matthew Trowbridge of Hitachi as non-executive Chairman. In the byelaws of SuperH Inc, the ceo is not a member of the board of directors." An even number of directors, divided equally between both parents, will slow down decision-making [Editor's note: see the Trade Secrets piece on "The founding team"]. The exclusion of the ceo from the board will reinforce the impression that SuperH is not an independent company, but very much under the management control of Hitachi and ST, perhaps discouraging other semiconductor vendors from licensing from SuperH.

SuperH is seeking new investment and hopes to appoint a VC to the board during 2003. The company currently has no merger or acquisition plans. The company aims to generate $150 million in revenues before IPO time, with an IPO tentatively planned for the beginning of 2005.

4. Portfolio

The product portfolio of SuperH consists of:

• SH4 and SH5 processor cores (SH6 under microarchitectural study)
• SuperHyway bus for system-on-chip applications
• Software and hardware development tools

According to Rolland, the license fee "ranges from a couple of hundred thousand to several million dollars," depending on the type of license, although he accepts that low ball license fees often attract the wrong kind of licensees, who are not serious about using the technology, leading to the IP being prematurely commoditised and the generation of a support legacy. He declined to provide figures on royalty levels, burn rate and revenue, but confirmed that both "Hitachi and ST pay a royalty to SuperH for use of the SH4 and SH5 cores. Hitachi pay a royalty for those ASSP devices that use the new SH4 core from SuperH, in preference to the older, internally-developed SH4 core." Licensees, apart from Hitachi and ST, include fabric server company Server2Net, and TVIA, a graphics chip vendor formerly known as IGS Technologies, with whom Hitachi had been negotiating an SH4 license for the last couple of years. Rolland stated that there was another, as yet unannounced licensee, "an American integrated device manufacturer (IDM) who is transitioning over to a fabless model."

The SuperHyway bus can be used in a similar fashion to ARM's modular bus architecture (AMBA), although SuperH has yet to determine the commercial model for its use and support independent of an SH core. A significant campaign would be required to promote this, given the amount of AMBA legacy IP available.

5. Succeeding in semiconductor ip

Platform solutions. Customers want solutions to business problems and will pay a premium for that which removes pain. Customers do not want to reinvent the wheel - they want to be given a virtually complete solution to which they may add their proprietary value-add as a differentiator. Most IP fails this test, and is much like plasticine - it is shapeless, and relies on the user to turn it into something useful, adding to their workload. The SH architecture is very well supported by third parties and is available from Hitachi and ST in the form of ASSP devices for use in applications including digital TV and automotive. In contrast, the offering from SuperH is limited to cores, which is not really a credible alternative.

As stated before in The Chilli, the only way forward for IP vendors is to focus on a growing niche and attack it with a complete platform with all hardware and software IP integrated, optimised and pre-verified & pre-certified for the application [Editor's note: take a look at the features on ARC and Siroyan, and the special report on IP, for more details of platform IP]. Rolland feels that "Developing complete platforms is very resource intensive" and that SuperH Inc is "Not yet ready for this." The Chilli contends that to rely on the plasticine approach to IP will result in SuperH spinning its wheels. Worst case, it will attract the wrong kind of customers in the wrong markets, asking for "specials", increasing the cost of sales with no decent return.

Platform IP requires specific skills, including legal, administrative and marketing, both to create a compelling offering targeted at a specific niche, and to support the completion of the business transaction. It cannot be operated like an ASIC business, or an electronic design automation (EDA) business. Expertise in specific market segments, alliance partnerships - domain expertise - is required.

Composition of the processor IP market. Processor IP revenue is made up of a mix of products, including DSP cores, embedded processor cores, configurable cores, etc. The majority of the volume is made up of mid-range cores, such as the ARM7TDMI, used in many mobile phone chipsets. The high-end processor IP market is small, and is limited to specific applications such as games consoles, where there is high volume and a requirement for high performance. MIPS and PowerPC hold this segment of the processor IP market. The two cores in the portfolio of SuperH, the SH4 and SH5, serve the same high-end segment of the market. It can be seen that SuperH is attacking a crowded niche. Rolland asserts that "It is an expanding niche and will be profitable provided SuperH secures two large licensees."

Decline in ASIC business. The number of ASIC design sockets is falling. Recent research from Gartner Dataquest has found that companies spend an average $5-10 million per chip design just on engineering, with mask charges around $1 million. The engineering cost is expected to rise in the future, raising the barrier to profitable ASIC development. There are some high-end ASIC sockets, but they are few in number and already served by MIPS and PowerPC. There are not enough crumbs to feed another entrant.

Design starts for ASSP designs have increased at the expense of ASIC design starts. For an IP vendor to make a serious effort in the ASSP business, they must offer a platform solution. Many segments suited to platform solutions, e.g. application processor for mobile handset supporting multimedia messaging, can be satisfied with a mid-range processor. Offering a mid-range processor in its portfolio would allow SuperH to expand its addressable market, offset its dependence on high-end applications, and earn revenue sooner.

The decline of ASIC sockets has also created a window of opportunity for field-programmable gate array (FPGA) technology, with processor IP vendors such as ARC, ARM and IBM providing cores with mixed success, as the resulting design complexity requires further support, discouraging FPGA vendors. According to Chapman, "Interestingly, the new customers of FPGA have been doing this in ASIC in older technologies and do not share the fear. We have to continually play the customers against the FPGA vendors to convince them of the marketability of the technology. Watch this space."

Independence. For an organisation to rely on an external partner to provide key technology, the minimisation of risk is key. One aspect of this is independence. Much of the value attached to the SH architecture is held by the parents, in terms of know-how for application-specific knowledge and alliances. This may deter potential licensees, who would be concerned by the control exercised by Hitachi and ST, who in effect, could potentially, unknowingly, act as gatekeepers to the value-add of the SH architecture.

6. Surgery required

It's not enough to copy a strategy that worked for ARM ten years ago. Back then, ARM was the only game in town, but in today's market with several SIP vendors, SuperH Inc must differentiate itself and offer real value. It has much strength in terms of market presence, technology pedigree, third party support, etc. However, these strengths are best realised today in the form of products from Hitachi and ST.

Companies will flock to license something that gives them an edge, and the SuperH Inc proposition does not yet provide that edge. However, they can build on what they have and consider the following recommendations from The Chilli:

• Increase the marketing resource, and make the most of the SuperH brand value
• Build one vertical market platform (see The Chilli definition), targeted at an expanding niche, as SuperH needs to provide a value-add alternative to ASSPs from Hitachi and ST. Point IP from commodity vendors (there are bargains to be had) could be integrated by SuperH into the platform
• Develop a university programme, based on vertical market platforms, requiring SuperH to work with leading universities in specific domains, e.g. Kettering Institute, which trains most of the automotive engineers for General Motors. This would significantly build on the efforts of SuperH to work with Bristol University as part of the Bristol Silicon Network
• Absorb the SH3-DSP core into SuperH as mid-range IP, thereby expanding the size of the market they can address, and providing a useful differentiator to ARM and MIPS. It will also help SuperH Inc generate revenue in the multimedia messaging service (MMS)-enabled handset market, as SH3-DSP products from Hitachi are winning design slots in that space
• Once the platform strategy is secure, the company can later exercise the option to move up the chain to become a fabless semiconductor vendor, preferably avoiding direct competition with Hitachi and ST by exploiting a niche, as explained in The Chilli value test. This should not be left too late, as the company may then be forced to compete with its licensees. The development team would also have to be re-skilled

They have already received a large amount of cash, but have yet to demonstrate they can create a viable business; this must be borne in mind by potential new investors. SuperH Inc must communicate a clear vision of what their mission is and where they are heading. However, if they are happy to be just a service centre for Hitachi and ST, they may have no incentive to propel the business beyond a couple of new licensees.

The danger with this is that if SuperH does not rise to the challenge by going all the way, its parents may dismiss it as a distraction from their current operational challenges, leading to a lack of attention, support and potential closure. With TriMedia Technologies being folded back into Philips due to a lack of traction, one can imagine several Hitachi & ST executives privately contemplating a similar strategy, although this would be complicated by the joint ownership of some of the intellectual property developed.

Comments on this story? Send an e-mail to Woz@theChilli.com.

© Chilli Publishing Ltd 2003

17FEB2003