SAN JOSE Ė Henry Samueli, chief technology officer of Broadcom, shared his thoughts on the company's $3.7 billion bid to buy NetLogic following his first meeting with NetLogic employees in Silicon Valley.
EE Times: Does this mark a change in your acquisition strategy previously focused on small companies?
Henry Samueli: This is by far our largest acquisition and our first of a public company, but it's not a change in strategy. We've looked at acquiring numerous public companies in the last 5-10 years, but never got too serious about any of them.
You can't do too many [acquisitions of this size]. This fills a huge hole in the high-end embedded processor space for us. For the foreseeable future we'll revert back to smaller, tuck-in type deals as we build up our war chest. There are so many interesting companies out there.
EET: What was the story behind this deal?
Samueli: We're always looking at companies; it's a very pro-active process. We have acquired about 50 companies in our history, that's about one or two per quarter. We have the process down pat for IT, facilities, HR--it's all in place. But historically it's been smaller 50-100 people companies.
For three to five years we have wanted to expand into the high-end embedded processor space. That was a hole in our portfolio, and they had the best road map.
EET: Why not Cavium which has more market share and a lower market cap?
Samueli: [Cavium] certainly was a possibility. We looked at the long term road maps of all the companies out there, including startups. We are not doing this for short-term gains, but for a long term investment in the people and products.
We believe NetLogic's XLP processor line will be the highest performance one out there. We are betting on the future.
The XLP's quad-issue, quad-threaded, out-of-order core is one of the most high performance cores in the industry. It's scalability to put two to 20 cores on a chip and 160 on a board was very attractive to us.
Also, their knowledge-based processors [acquired from IDT in 2009] complement our broadband switches. And their Optichron acquisition in wireless complements our own recent acquisitions in microwave backhaul and small cells. We are now developing a very significant portfolio in wireless infrastructure.
EET: Will NetLogic have to change its design flow?
Samueli: From our due diligence, it looks like there's a lot of commonality in tools and foundries. I donít expect any significant changes. They just have a lot more opportunity in IP sharing and reuse now.
EET: Do you have any plans for new integrated products based on their offerings and your IP libraries?
Samueli: Not in short term. They will pursue their existing road maps. In the medium term, there are opportunities to mix and match IP, but we donít want to be too disruptive. Once we put all the smart architects together, they will come up with interesting ideas for new products.
EET: How was your first meeting at NetLogic in Silicon Valley today?
Samueli: It was an all hands meeting and very positive with great questions. They were curious about how we do business, how we scale up our design infrastructure around a central engineering team to deal with 68 design centers around the world contributing to chip design.
EET: NetLogic's XLP II will use 28nm technology. Does Broadcom have many 28nm designs in progress today?
Samueli: We have several 28nm products in development. All central engineering is now working on 28nm libraries, including analog and RF. Our switches are 40nm today, and once they are in production the entire design team is off working on next gen which is 28nm. The same is true for basebands and applications processors.
We work with all the foundriesóTSMC, Globalfoundries and UMC. TSMC is in a strong position [at 28nm]. It's still unclear how they all will implement 14nm, and the whole issue of double patterning at 20nm is expensive--everything gets exponentially more difficult as you move to next node.
EET: Any wish list for your EDA vendors?
Samueli: Plenty. There are many new tools in each generation to solve problems that didnít even exist in prior generation. As the number of tools grows, we need help making the tool flows and interfaces seamless. We also need tools that can handle scaling as we approach billion-transistor chips.
EET: What do you think about ARM-based servers, another big comms infrastructure concept out there today?
Samueli: It's an interesting market that ARM and even MIPS has the potential to address. It's all about the software ecosystem. If the software is there, it doesnít matter so much about the hardware architecture. You'll see some alternative architectures appear over the next several, but thatís not our focus for now.