High-tech

Due Diligence: Siroyan

Market dynamics

Siroyan has in the past stated that it plays in the market where general-purpose processing meets DSP. Neil outlined the target markets as being 3G infrastructure and terminals, wireless LAN, wireline communications and, their first target, multimedia. The Chilli believes that this is much too broad for a new architecture, with many different technologies, standards and competencies required. The highest value categories of the market, including cell phones are dominated by the combination of an ARM microprocessor core with a DSP core from the DSP Group (DSPG). 3G has been touted as offering new opportunities, but so far the newcomers have not succeeded in gaining entry past the 2G incumbents who influenced 3G development, ensuring the market segment is set for a bloodbath. Wireless LAN (subject of an upcoming feature in The Chilli) is a newer bubble segment - with too many suppliers and too little short-term revenue to sustain them.

Siroyan plans to launch a multimedia platform, which is multi-standard and programmable, supporting H.264, MPEG-4, MPEG-2, etc. There will be three variants of this platform, targeting different performance levels. The product release window will be during the latter part of 2003 - this is a very high-risk schedule for developing, compliance testing and releasing products to customers who need close support, although the company points to the fact that it has delivered two test chips in the recent past. Neil claims that a Siroyan customer will demonstrate one of the first H.264 solutions for the Consumer Electronics Show (CES) Las Vegas in January 2003.

The downside with this approach is that competing vendors may develop cheaper, fixed-function, hardwired solutions, and hence undercut Siroyan's value proposition. Such solutions will come out later, but when customers switch, Siroyan may find that the license fee and a year or so of royalties won't begin to pay back the investment they made in working with the initial customer base. As real-life examples, take the case of TriMedia or Equator - both vendors provide VLIW hybrid RISC/DSP devices/cores targeted at multi-standard, programmable end products. They have been consistently undercut in the market by cheaper hardwired solutions that have emerged as standards have solidified. If it has created difficulty for these guys in the past, it is highly likely that Siroyan may face the same challenge. It might be time to re-assess and consider the old fabless model.

Let's take a broader view of the situation. The market for DSP is approximately $8bn in revenue terms. According to our broad estimates, approximately $4bn is for general programmable DSP, dominated by Texas Instruments and Analog Devices. The remaining $4bn is accounted for by application-specific standard products (ASSPs) using both programmable and fixed-function DSP cores, which address specific markets, e.g. wireless handsets, wi-fi (802.11 WLAN), audio CODECs (MP3, AAC, etc), etc. An ASSP typically combines a mixture of microprocessor & DSP cores, with memory, peripherals, application-specific logic, and system software for that market.

Customers are often faced with the decision of whether to choose an ASSP or go the custom chip route and develop their own application-specific integrated circuit (ASIC) internally. The economics are telling. It costs approximately $10M to research, develop and produce an ASSP, including the non-recurring engineering (NRE) costs. In contrast, companies that procure or buy off-the-shelf ASSPs can spend considerably less than that. This is because the ASSP vendor can achieve a greater economy of scale by supplying several customers, and also takes on the burden of researching requirements, performing development, production, compliance and distribution. In the current economic climate, potential customers are unwilling to pay large NRE fees for custom chip projects, unless the projected volumes are in the millions of units. This is further shrinking the market for semiconductor IP vendors.

ASSPs - with optimised hardware and software - have gone on to become de-facto standards for specific market segments, including cell phones, set-top boxes, etc. The most dominant ASSP instruction sets in the embedded processor market include ARM (cell phones), MIPS (games consoles) & SuperH (home and car multimedia), from Hitachi & STMicroelectronics. By their very presence, these are "gatekeeper architectures" to market segments - so you must find a way to complement them to get in - a strategy to compete head on with them may well be a high-risk strategy.

Zooming in on the DSP IP market, DSPG, currently merging with Parthus to create ParthusCeva, is the main competitive threat to Siroyan, acknowledged as such by Neil. According to Gartner Dataquest, DSPG currently holds 70% of the DSP core licensing revenue. Their latest product, CedarDSP, has adopted some of Siroyan's OneDSP concepts such as clustering. It is not compatible with any of DSPG's previous products, and requires, like OneDSP, investment in new development tools and middleware - opening a chink in DSPG's armour. However, this is offset by DSPG's market share, customer relationships, brand name and reputation, giving Siroyan a tough competitor. New entrants to DSP IP, such as the ST100/200 from STMicroelectronics, Carmel from Infineon and Meta from Metagence (a division of Imagination Technologies) will further muddy the waters. There can only be room for two or three major players. Siroyan therefore needs to adopt a strategy that renders the plentiful resources of such vendors meaningless, and that's where focus (and the execution speed of a nimble start-up) is key.

• Introduction
• Background on Siroyan
• Technology & products
• Market dynamics
• A way forward
• What can Siroyan and fellow entrepreneurs learn from this

Comments on this story? Send an email to Woz Ahmed: woz@thechilli.com

© Chilli Publishing Ltd 2003

20SEPT2002