High-tech

Due Diligence: Tenison EDA

By The Chilli analysts

Introduction

Tenison EDA was founded in 2000 by Dr. David Greaves, who previously co-founded broadband chip set company Virata Limited (now part of GlobespanVirata Inc). In the 1990s, Greaves had carried out research on generating software models in C from a hardware description language - known as Verilog, resulting in a tool, VTOC, - at Cambridge University's computer laboratory. Tension EDA originated as a result of this research. The first incarnation of Tenison VTOC was released within the electronic design automation (eda) community for free in 1996, beta customers were signed up in 2001, and full commercial release occurred in 2002.

The company is targeting software development for system-on-chip (soc) by providing software models of the hardware representation. The theory is that early software development, well before the hardware is built and tested, should reduce time-to-market for the soc.

Vital statistics

Value proposition

The Chilli perspective

Partners

Roadmap

Competition

Way forward

Vital statistics

Aside from Greaves, the company's chief scientist, the management team includes Jeremy Bennett as ceo (formerly vp of cross-divisional solution selling at Marconi), William Stoye as vp of engineering and cto (formerly of Virata and Acorn) and Brett Clay as sales vp (ex CrystalVoice, Synopsys, Avanti, Ikos), guided by Dr. Robert Sansom (co-founder of Fore Systems), who is the non-executive chairman of Tenison EDA.

Much of the research and development was already carried out at Cambridge University by Greaves and two post-doctoral researchers. Greaves had started selling VTOC licenses at $2000 per site (unlimited seats) per year, to get early customer traction during 2000. Cambridge University was unique for a UK university in that it allowed the researchers to retain the intellectual property (ip), which allowed Greaves to retain the ip rights to his work.

Chilli R1 funding (post prototype) was provided in July 2002 by Axiomlab, Sigma, Providence Investment and Robert Sansom. The objective of the R1 funding was to fully launch the VTOC product in the market. The company recently closed a Chilli R2 round of an undisclosed amount, which will enable the company to increase its market share. The company is already generating revenue and according to Bennett expects to breakeven second half of 2004.

For a specialist software company, the mix of staff is interesting. It has thirteen staff, with six in r & d, three field engineers and the balance in sales, general and administration. Some members are part-time, helping to keep burn rate low.

Tenison has two offices - Cambridge, UK and Mountain View, California; the latter opened during Q1 2003. Independent repping (sales agents) houses are used extensively across the USA to keep costs low while still hitting potential target prospects. Current customers include two product divisions of STMicroelectronics, GlobespanVirata, Seaway Networks, Mysticom and Centillium.

Besides the US market, the Japanese market is also to be targeted soon. Bennett does not see Korea or Taiwan as an immediate priority due to the relatively small size of their markets at present, although figures for 2002 from the EDA Consortium state that China, Taiwan, Korea and Singapore showed some positive revenue increase, against a decrease in North America, Europe and Japan.

Value proposition

Tenison's VTOC product is targeted at co-design - essentially the process of developing hardware and software concurrently rather than sequentially, in order to reduce development time, risks and costs. The product is positioned squarely at hardware modelling for early software development.

VTOC generates a cycle-accurate implementation of a complete Verilog design in the form of a C++/SystemC program. This program can then be compiled using a standard compiler, running faster than a conventional behavioural simulator. The C++/SystemC code will fit into virtually any design flow: 'We are not forcing anyone to alter their design-flow,' says Bennett.

The claim is that VTOC supports the whole synthesisable language, supporting multiple clocks and gated clocks. It goes through all the steps of a logic synthesiser but outputs a C program rather than a list of gates.

According to Tenison, the models generated by VTOC can run ten times faster than compiled Verilog simulators. A number of soc development projects utilised VTOC, including GlobespanVirata's Helium500 communications processor, consisting of two cpu cores, static ram and logic for a gate count in the 3-4m range.

The list price of VTOC for a perpetual license is $50K plus maintenance per floating seat that runs the model generated by VTOC, with discounts for large design centres using lots of models. The licence allows the tool to generate as many models as wished, but it is execution of the models themselves that attracts the license fee. One model may be used by several members of the software development team, as well as by architects and verification engineers, and a license fee is paid for each simultaneous model execution. Customers are also able to secure licenses that allow the code to be re-distributed to their end customers. Tenison also provides a consultancy service in addition to evaluations and training.

Tenison has one major release per annum, with version 2.0 launched this year, and version 3.0 slated for DAC 2004. Minor releases are made throughout the year incorporating patches and bug fixes. VTOC is supported on Linux, Solaris and Windows. One patent has been granted, with several more in progress. According to Bennett, 'David Greaves is a key asset, given his understanding of the challenges and his innovative solutions.'

The Chilli perspective

In the software market, licences enable use of the software for a specific period, for a licence fee and an annual maintenance fee.

Being software, once developed, the only additional cost is packaging, distribution and maintenance. This artificially gives a very high figure for cost of goods sold (cogs). This can be in the 80 to 90 percent range, but hides the true cost of serving and maintaining a customer. A better measure for this industry is the cost of customers serviced (cocs). This parameter includes the hidden elements related to satisfying customer expectations and demand. Cocs includes the cost of:

• Sales cycle, including number of customer visits, presentations, software evaluations, mini proof-of-concept (prospect-specific), training, application support and maintenance, requiring time and combined effort from sales, marketing, application support, engineering and management.
• Customer support, which is a variable cost, as each customer has a different design flow, requiring the vendor to spend time and effort - not just a customer service applications engineer, but also some of the original r & d staff, which takes them away from true r & d work.

Cocs is an important metric for software, eda and ip companies. Keeping the cost of cocs low is a major differentiator for software startups. Many companies fall into the trap of supporting major customers who are ultimately unprofitable, as the cost of true support exceeds the profit margin from the deal.

It is often difficult to charge for support, as nobody likes unbounded costs, and the support may be required to cover an undocumented 'feature' of the product.

The sales cycle for eda products is quite long, and four to six months could be typical, which includes typically 1-2 man/months of engineering time for evaluations and proof-of-concept. Tenison is reducing this cycle time, as they get more experience from each prospect and add new features into the product. According to Bennett, 'Training and good documentation have been very important to reduce the initial learning curve. We will also be introducing a graphical user interface in addition to the existing command line interface.' Tenison has kept the cost low by keeping the documentation in electronic form and by using online distribution.

Tenison's account strategy should therefore concentrate on major strategic accounts, as their revenue model relies on fresh design starts (new hardware models) and large development teams. A large account would ideally provide multiple design centre locations and projects using similar design flows, thereby amortising support costs across several projects and centres. It is therefore a good sign that Tenison has penetrated some large integrated device manufacturers (idm) like STMicroelectronics - a large idm has the right ingredients:

• A relatively consistent design flow across design groups
• Well-staffed and multi-skilled design teams
• Higher number of design starts per annum
• Significant amount of legacy ip that may require conversion into models

The challenge with selling VTOC to other startups is that the design team may be small, and after they have made the first product, there may be nothing else for a while, as they live or die on sales of that one product.

Partners: the company is in discussion with a number of eda companies regarding potential partnerships. C-based design flows, like those from Celoxica and C-Level Design (acquired by Synopsys in 2001), could benefit from VTOC being used to convert legacy ip into C models - the legacy ip issue is a major barrier to new development methodologies, relegating them to 'niche religions'. It is important that Tenison concentrates on those partnerships that yield to real design starts. Choosing a partner who is struggling to fill their prospect pipeline would be an unacceptable risk for Tenison EDA and other similar software startups.

Roadmap: the company will shortly release a vhdl version of VTOC, (vhdl is another type of hardware description language) which will be an important requirement for getting traction in the European market, where vhdl tends to dominate, with Verilog more dominant in USA and Japan.

Competition: Tenison is in something of a niche area. Indirect competitors include Mentor (with the Seamless product) and Axys Design Automation, both of whom target the generation of models as well as profiling toolchains, at a higher per seat cost. A direct competitor recently emerging from the USA is Chilli R1 startup Carbon Design Systems, with a management team from C-Level Design and Quickturn Design Systems, with vc backing from Commonwealth Capital and Flagship Ventures totalling $5m.

When Carbon actually delivers a competitive product, this would constitute a significant threat to Tenison, as Carbon has a home advantage. Tenison needs to consolidate its position via an aggressive push in all markets, key partnerships, and leverage its several years of experience from real customer traction. To address this challenge effectively, Tenison should consider raising another round, to address the following:

• Prove the level of complexity VTOC can handle, in terms of gate count, and also execution speed
• Focus on a more defensible niche, such as with large European accounts, e.g. Philips, Nokia, Ericsson, Alcatel, Infineon, requiring vhdl support
• Aggressively penetrate the Asian market, as well expand sales and application support in US and Japan
• Continue to expand r & d to offer more complementary products
• Consider a potential acquisition, which provides complementary products and channels to market.

Way forward: the top three eda vendors have tended to grow by acquisition, pursuing the holy grail of a complete end-to-end design flow, as newcomers like Magma Design Automation deliver ground-up end-to-end design flows. The conclusion for an eda startup, as described by the ceo of another eda startup is 'That it's an all or nothing proposition - you either get acquired by one of the top three, or stay a tiny niche player.' An exit for Tenison's investors would be an acquisition. Many eda acquisitions are made on the basis of unproven technological prowess, but if Tenison can demonstrate traction with two or three tier one idms, it will be snapped up on the basis of track record, particularly if it demonstrates how seamlessly it fits into a particular eda vendor's tool flow.

The danger with a sweet spot in eda is that once identified, it becomes flooded with startups hoping to get acquired. Tenison should aim to build on its track record of delivery by exploiting a defensible niche to stay above the noise level. If it is to double revenues and sell 50-75 seats per annum whilst keeping support costs low, it cannot afford to get into price fights with local firms in the USA.

Comments on this story? Send an e-mail to editor@thechilli.com

© Chilli Publishing Ltd 2003

06SEP2003