Article Info

Wintegra – Following up on a Comment

I received a good question on the Wintegra Inc. (WNTG) post I made last week. I felt the response was worth a post on it’s own. Please remember this is not investment advice, and not a negative outlook on Wintegra as a comapny, just my opinion of something that needs to be considered when attaching a multiple and valuation to the company. References to Teknovus and Passave are in context with the previous post.

From “Watcher” –

I don’t really understand your Cisco example: according to your example Cisco doubled its purchasing from Wintegra in 3 years (2003-2005): 27% of $4.5M in 2003 to 10% of $19.5M in 2005: can you explain this again (why doubling sales to Cisco is a problem)?

You ask a good question. In the interest of brevity I oversimplified my argument. Let me explain in detail.

Wintegra NPU Revenues
Fiscal year – Cisco – Total

  • 2003 – 1.21M – 3.74M
  • 2004 – 1.58M – 7.85M
  • 2005 – 1.95M – 18.79M

Note the radical difference in growth rate.

Designing in network processor (NPU) silicon into telecom and datacom equipment requires high initial fixed costs. Companies invest the largest amount of capital to complete the first NPU design. This investment is primarily in heavy software training/learning curve climbing as well as software coding. Follow on products can build on this investment, so incremental design-ins require substantially less investment.

As an NPU vendor, securing the initial design win is a big victory, because once a company has made this large investment they will likely re-use the design for a lower cost, and the NPU vendor enjoys some degree of incumbency within the company. This gives them substantial pricing power and leverage on their customer.

During the previous bubble, the hope of NPU vendors was that a Cisco or Lucent or major tier-1 would acquire a hardware startup using their NPU, with the goal that the silicon would ‘go viral’ within the larger company. This did happen to a limited extent within Cisco.

The problem is, large companies like Cisco realize this incumbency reduces their pricing and technology negotiating power due to a reliance on a single vendors architecture. GSM, Cisco’s buying organization (very capable folks BTW), refuses to allow components of this sort in house as a result, and prefers to in-source components of this nature by doing internal silicon development. They then control their own supply chain and technology roadmap, albeit with much higher fixed costs. Microsoft (MSFT) took the same path with the Xbox 360 once it failed to get the technology roadmap and pricing reductions from Nvidia (NVDA) it felt was necessary. Microsoft developed the silicon in-house with IBM as a foundry and design partner.

Now – my point.

The revenue stream indicates Wintegra isn’t seeing horizontal ‘viral’ penetration at a large account like Cisco. A ramp from 1.21M to 1.95M is more characteristic of a single product ramping in volume after introduction. Wintegra’s overall growth has been geometric as they continue to penetrate other accounts- why isn’t the same true of the Cisco revenue stream? I suspect they are limited by the structural forces I outline above.

Passave has a device that could more easily be replaced by a chip from Teknovus – if Mitsubishi Electric decided the pricing or supply security or technology from Passave was lacking, within three months Mitsubishi could replace Passave. Not so with an NPU.

The structural problem NPU vendors face – closed architectures – prevents them from ever being completely dominant at the big Tier One vendors. Cisco knows that they could not quickly replace an NPU vendor in the same way they could replace a MAC, memory, or framer vendor. Therefore, they internalize the risk. Small companies lack the resources to make this choice.

Therefore, it isn’t reasonable to assign the same multiple to two semiconductor companies (Passave and Wintegra) that have radically different business profiles due to structural design-in barriers. The difficulty in designing out NPUs in general (not just Wintegra products) makes the largest potential consumers of these devices less likely to design them in.

Considering the attention to all of the minute risks highlighted in the S-1 (Starts on page 6), I think a mention of how large customers tend to avoid closed architectures should be in order.

More info on Wintegra can be found at Optical Keyhole.