Attendees at today’s ARM Technology Conference (ARM TechCon) were treated to an unusual way to end the day: a “fireside chat” between Simon Segars, ARM’s EVP and General Manager PIPD (physical intellectual property division), and John Bruggeman, a Senior VP and Chief Marketing Officer at Cadence. As you can see in the photo below, the “fire” was provided by a plasma flat panel display and the topic was “The Evolving Electronics Industry.”
Segars opened the 45-minute session by discussing the collaboration between ARM and Cadence in developing a synthesis flow and script for the new ARM Cortex-A15 multi-processor core (discussed in today’s earlier blog “Realizing the ARM Cortex-A15: What does the road to 2.5GHz look like?”).
Bruggeman immediately picked up on the theme of collaboration. “Look across the supply chain,” he said, “there is a ton of duplication. That duplication drives costs into systems that none of the participants can afford. We need to stop duplicating effort and drive these costs out. For example, ARM, Cadence and TI worked on the first implementation of the (Cortex-)A15 design. That was a collaboration of three companies. In the future, the IP vendors, the foundry, the EDA companies, and the customer must all tighten collaboration or development will become too expensive…innovation will die. Venture capitalists will not be able to fund new companies and existing companies will not be able to develop products.”
“What will we look like in ten years if innovation is broken?” asked Bruggeman.
Segars suggested that the electronics industry itself has caused the problem. “Industry disaggregation has lowered costs, but it has raised inefficiencies.” Segars’ stated position is that these inefficiencies are no longer affordable. “We need to squeeze them out,” he said. Bruggeman then concurred, saying that the inefficiencies might have been tolerable as recently as two years ago, but the “complexity curve has exploded” and, as a result, new-product development has become unwieldy with the present disaggregated development model.
Then Bruggeman switched into a discussion of apps-driven design, one of the bedrock concepts of the EDA360 vision. “We’re missing the big part of the applications story,” he said. “The applications conundrum is not driven by the user; it’s the new business model.” Here, Bruggeman was referring to the apps-centric business model where the end customer buys a product and then purchases additional apps—which upgrade and enhance the product—on a continuous basis. The iPhone is the poster child for this new business model that delivers continuous revenue. The carrier (AT&T for the iPhone) and the maker of the mobile handset (Apple) get some revenue on the initial acquisition of the phone but then they get a continuous revenue stream through the purchase of services and apps. It’s this new business model that permits the development of expensive platforms like the Apple iPhone.
This new business model also extends far beyond phones, said Bruggeman. It’s now used in markets as diverse as automobiles and network switches. For example, Bruggeman pointed out that Huawei has adopted an apps-driven model for its switches, offering for-pay, downloadable upgrades such as deep-packet inspection to enhance revenue after an initial purchase. “These new business models will (in turn) drive new device types,” said Bruggeman, “and we, as part of the supply chain for these new devices, need to keep up with that.”
Switching to an example of where some of the inefficiencies in the supply chain exist, Bruggeman recalled his previous employer Wind River Systems, where they “spent millions of dollars” to develop various flavors of Linux for customers using different hardware platforms. “Android needs to be predictable, scalable, and all those other ‘able’s—so it’s not able to take advantage of hardware innovation unless the drivers are brutally hard coded.”
“How can semiconductor and system vendors differentiate? How do we realize the promise of Android while allowing hardware innovation?” asked Bruggeman. He then answered his own question: “It can’t be done without collaboration” and he pointed out ARM’s recent announcement of the multi-company Linaro project (www.linaro.org) to develop a standard version of embedded Linux.
Segars then added his own take on the need for collaboration, saying that the average price for a chip was currently $1.60, which provides very little money for process technology, EDA, and IP innovation. “We need to get the inefficiencies out or there will be no money for semiconductor innovation,” he said.
Then at the end of the chat, Seegars pulled a fast one on Bruggeman. “You’ve been in the industry a short time,” he said, noting that Bruggeman came from the software industry and had just passed his first anniversary with Cadence. “What has surprised or concerned you?” he asked innocently. From the smile that flashed across Bruggeman’s face, I knew there was a noteworthy answer or two about to surface and. I held my pen at the ready.
“This is going to get me in trouble,” said Bruggeman, confirming my suspicion. “First, we move slow. Cycle times are long…too long.”
“The second thing is really going to get me in trouble,” he then said. “We (EDA companies) spend too much time hating, not enough time co-opetating (competing and collaborating at the same time). We need to solve some basic customer problems. They’re holding us back. Things like power format (referring to the difficulty in creating one standard format for files that express a designer’s power intent). This doesn’t make sense to me. The software world is more careful about this. Software vendors spend more time creating standards and competing where they can add value.”
“It takes an industry to solve these problems. What if we could cut the cost of a startup from $125 million to $30 million? What if we could cut three months from a chip-development schedule? If Cadence could do this, it would be great. But it can’t do it alone. There must be collaboration.”
Perhaps next time there’s a fireside chat like this, there will be more representatives on the stage besides one from ARM and one from Cadence.