Every product and service is also a marketplace for further products and services that enrich the opportunities customers can pursue. We provide value by selecting and recruiting these resources, categorizing, testing and certifying them. We are in a prime position to collect and publish customer feedback, continuously improving the ecosystem. When we increase the access of members to each other by making markets we increase productive exchanges. Increasing exchanges creates more value faster, and advances the whole.
NVIDIA, Dell, AMCC, Linaro, Xilinx
The leadership lesson of this chapter is simple: Go for it. We live in a time when the conditions to develop business ecosystems have never been better. At its heart, a business ecosystem is an initiative based on shared purpose, with open communication and peer-oriented relationships, where the participants share and grow together. And as a general rule, most initiatives can be made stronger with an ecosystem perspective. Partly this is a function of the global diffusion of knowledge, and the ease of connecting to others. In many cases the best resources are outside of one’s own company, and/or can be better accessed in a network.
The connected community has a qualitatively new culture of collaboration — very broadly shared, cross-industry, ultra competent and with a large global population of members. This worldwide growth medium, this “yogurt culture” for business, is more or less invisible to most observers, but it is the secret sauce in the stew, the Miracle Grow in the soil.
In the following pages we will examine four distinct ecosystems that are planted in the soil of the connected community. These demonstrated four different ways to succeed. Taken together these stories also illuminate the overall benefits available in the connected community.
But the major lesson here is “go for it.” This is your time. There will be successors to the connected community. But your reason to act now is that you can see this one. There are, as I’m sure you have noticed, many in tech who can’t see or can’t figure out how to understand the current landscape. Here is a warning: that may be us in the next era. The important features of the next era may literally be invisible to us. So make your move now, while you can see this one.
Consider a recent collaboration between NVIDIA and ARM. Phil Carmack is the head of NVIDIA’s mobile business. About a year-and-a-half ago he developed an idea for improvements to the ARM Cortex-A9 processor, which NVIDIA wanted to integrate alongside an LTE cellular modem. He went to see ARM CEO Warren East in his office in Cambridge UK travelling from California where he works and lives.
“I drove four hours to get to Warren’s office, sat in his lobby so that he would talk with me …
“‘There’s a theme here Warren. I think it’s a great opportunity [to improve the Cortex-A9 processor design] if we could work together on this.’ And he said, ‘OK, Phil, we will do it.’”
The collaboration began by examining the real feasibility of Phil’s suggestion. As Phil says,
“If you’re a participant in a market as efficient as this, you’re running flat out. You don’t have much time to adapt and change course. So it should be an important decision. [We needed a solid process, we just needed it fast.] Sure, I would have liked just for them to believe that NVIDIA has all the answers and whatever NVIDIA says, that’s what we should do. That’s not very realistic.
“On the other hand, we recognized that this new design wasn’t [going to make it into the next manufacturing cycle] unless we got it done really quickly, three or four months. So we managed it to be as simple as humanly possible. We both put effort in to make it work.
“The result was a dramatic improvement in terms of performance per watt, while still keeping a really small [ARM processor] core area.
“There was no consulting arrangement. We both did it for the good of the ecosystem. These ideas we shared freely with each other just raised the value of the ecosystem and we recognized it. In the end [NVIDIA] might be helping our competitors to a certain extent. But we are also making it possible to create a product that is kind of amazing. So we are happy that it turned out that way.”[i]
A few things stand out in this story. First, Phil made the meeting happen quickly. Warren was open to it. The substance of Phil’s visit was obviously compelling — a credit to Phil‘s vision. Yet what then happened would be remarkable in other organizational communities: both were willing and able to commit themselves and their organizations. Literally a few months later the chip was being baked in silicon.
An earlier chapter highlighted a similar story in which Robert Hormuth of Dell flew to Los Angeles to meet with Ian Drew. Out of that, a multi-year joint study of servers, and ultimately a project to build servers — in turn in an ecological relationship with several ARM ecosystem fabless chip companies, and their choices of fabs. Super credit to Robert for the ideas and initiative, super credit to both for the relationship-making process.[ii]
During this time the ARM high performance chipmakers were emerging in a similar co-evolutionary fashion. We met Paramesh Gopi briefly in the first chapter. His company AMCC is one of four key companies working with Frank Frankovsky at Facebook on transforming the server architecture of the largest scale data centers.
“I left Marvell in 2008. In the summer of 2008 for about three months I spent [my time] hanging around with all the big cloud data center guys, and it was apparent that there needed to be a change relative to server architecture.”
At the time, the Intel architecture market was saturated, so it would be hard to gain a foothold.
“IBM PowerPC was the obvious choice, but it was a closed ecosystem.
“I’d worked with Apple on the iPhone when I was at Marvell, with Jobs and the whole initial iPhone team. They control a closed ecosystem for a large, very large market.”
Paramesh had experienced the downside of a closed ecosystem, even within the success story of the iPhone.
“After talking to key customers in China, and the web services guys, we all said, ‘You know, the PowerPC is a closed ecosystem … IBM controls rate of resolution. IBM controls peripherals. IBM controls architecture to the point where they dictate and preside over any type of architectural input, so as to not fundamentally change their inherent closely controlled server and services.’
“So, it became obvious that we needed to go off and think about ours, especially when [a very large Chinese customer] said, ‘You know what? If you guys are going to do anything, why don’t you build an ARM?’”
Paramesh wondered, could they “understand that the performance aspect of this is going to be so critical? It’s not your conventional ARM mobile core.”
So Paramesh contacted Tom Cronk, head of ARM’s processor division, to explore a joint commitment to ultra-high-performance. Each side had questions: AMCC had questions about performance, and ARM had questions about whether a company the size of AMCC could pull off co-designing a new architecture.
“We spent multiple meetings at [a special location near] Hyde Park because it was a secret project.”
Both sides were convinced.
“ARM wanted to make sure we wouldn’t screw things up and fragment the architecture. One of the things we [jointly] decided was they would adopt our strategy of not just building a better CPU but building a better server.”
In his view the result is a significant win.
Paramesh credits this as an ecosystem success.
“We’ve broken multiple records from a business ecosystem perspective. It’s taken Intel roughly a half a billion to three-quarter billion dollars [to do a similar program]. We’ve done it in less than, you know, less than one-fourth that budget.”[iii]
This open, collaborative community manifests open source community processes in its transparency, information sharing, joint experimentation and wisdom of crowds. Sometimes the crowd is small — the joint feasibility study by ARM and AMCC comprised less than ten people from each company, as did the joint project between NVDIA and ARM — but sometimes the crowd can be quite large.
However, in contrast to the typical open source project, in these cases the open process can be wrapped in secrecy when necessary — in the AMCC story about three years — and maintain its openness at the team and inter-organizational levels.
Even more interesting is that this community is making and managing investments in which can reach into the billions. And this in turn boosts proliferation and multiplies the effect of its ideas.
Third generation ecosystems can start small. They sound puny at first — powerful later. They are subject to an excellent set of Darwinian fitness requirements: Third-generation ecosystems get off the ground in their early stages only if representatives of individual companies see enough self-interested value in the project that they are willing to participate. And the self-interested value must be calculated knowing that their competitors — participating or not — will benefit as well.
By contrast in a vertically integrated company you must convince someone up the hierarchy. In a first-generation business ecosystem you must convince an executive of a keystone company or its corporate venture arm. Neither of these fitness landscapes is even close to as rigorous as the enlightened-self-interest of your peers in third-generation partner companies.
So far we’ve reflected on three stories of newness at NVDIA, Dell, and AMCC. Let’s now look at a different aspect of the ecosystem. We will consider another arena where one can “go for it” using a purpose-built ecosystem: Oft-times in technology there are existing problems to clean up that, when fixed, will open the way forward. This is the dreaded legacy systems problem.
When the legacy issue stems from the community spending money on duplicative and not-particularly-differentiating technology, all of the affected partners need to be involved. The point of the exercise is to get everyone using the new, common bits of technology. In general companies will not adopt something of this nature unless they helped create it.
One such initiative is Linaro, which is developing shared Linux middleware that can be used with the Android operating system on a range of devices.
George Grey is the CEO of Linaro. I met with George one morning this winter at the Main Street Café near my home in Concord, Massachusetts. My three-year-old refers to the café affectionately as “the muffin store” for the raisin bran muffins. The cafe was George’s choice and is my favorite place in town. George was relaxed, cheerful in business casual. We sat atop barstools at a pine table in the back of the café.
George described the Android and Linux situation that he faced when he began working on Linaro,
“In Linux you had Intel that invests from top to bottom. You had all these ARM companies doing their own thing.
“Now together it is actually a lot bigger ecosystem than Intel’s. But individually none of them had clout.
“And so we went to all these companies and said, ‘Look, we can’t go on like this. It’s costing you a lot of money. You’re maintaining a lot of code. Let’s all agree what the functionality is, and then we will just do it once.’
“But it is a lot harder to do this when it is after the event and everybody has their own proprietary piece of software. But so, anyway, we are getting there.
“And instead of ARM having to fund all this, which it can’t afford to do, it’s shared amongst all of the ARM SoC vendors.
“Now, the interesting thing here is that when you think about the value of ARM, to me, ARM’s value is not in low power. I mean yes, their DNA is low power. They’ve come from the mobile space and therefore low power is a technology benefit for ARM. But ARM’s value is much more in the business model, and the fact that it enables huge innovation and differentiation very quickly.”[iv]
Yes, it is a powerful model, and “ARM” has become shorthand for the system of hundreds of interpenetrating third-generation business ecosystems, thriving together in the open culture of the connected community.
Another type of ecosystem development is for a firm to create forward-facing application software for the end-customer. The conceptual model for this in the smartphone world is the Apple or Google app store. The customer is able to buy a program and download to his or her phone, or download a front-end to a cloud service. Either way the customer is — ideally — able to depend upon the quality of the app and have it “just work.” Each of these applications may be said to have its own ecosystem — thus Google maps, or Evernote or Angry Birds create communities, and have others developing extensions to their software and using it as a platform
One of the challenges for a Google or an Apple is managing the vast number of applications and ecosystems that can develop on the platform. This class of problem, we will see, is not limited to the consumer app stores — as a megatrend it is coming in some form to every part of the ecosystem that touches a large number of end customers. For example, Xilinx:
Xilinx makes systems-on-a-chip that serve some very complex applications. One of their major initiatives is in smart vision. That is, taking video data and analyzing it. This means looking for patterns in ambiguous data — making sense of it, and doing something. A lot of what they do is in automotive applications, where obviously the software must make an interpretation in moments — in order to prevent a crash with another car, avoid hitting a child or keep the car on track while sliding on ice.
Xilinx has three other major thrusts in variations of smarter networks — smarter wired, wireless, and data center. In each case “smart” involves understanding moment-by-moment network conditions, including equipment performance, outages, traffic patterns, loads, applications and adjusting overall for many different dimensions of higher performance.
Xilinx has, like most companies in the connected community, amazing underlying technology. Their chips are fully programmable, meaning that the same off-the-shelf chip can power a diversity of the very complex applications mentioned above.
Their chips are built principally of FPGAs — field programmable gate arrays. These are vastly capable arrays of transistors that can be reprogrammed. The range of “identities” they can take goes from mainstream calculations and logic, to fast algorithms for vision or communications analysis, to analog to digital signal converters that are part of software defined radios. They can be programmed to connect to input and output devices from medical equipment to auto steering, and on the other side to GPS, accelerometer, chemical sensors — and on and on.
In addition to the programmable areas on a chip, other dedicated resources can be made available. Typically this would include a general-purpose processor as well as specialized accelerators tuned for particular jobs. The routing on the chip among the areas is as programmable as the logic that does the work in each area. Xilinx SoCs are little digital ecosystems. Seen from a systems perspective the microelectronics field is ecosystem after ecosystem after ecosystem, all the way down.
I spoke with Steve Glaser and Dave Tokic from Xilinx. Steve is Senior Vice President, Corporate Strategy & Marketing Group; Dave is Senior Director, Partner Ecosystems and Alliances.
Steve and Dave are on their fourth company leading ecosystem development. They have become experts at it, systematic and comprehensive. Impressive.
Steve and Dave make a pair — energetic, quick thinking and able to finish each other’s sentences. Xilinx headquarters is in San Jose, on “Logic Drive” — you have to love it — on the southwest side of the city, near the foothills. Despite the building being a typical Silicon Valley 2-story “earthquake shelter” office building, Xilinx is only a couple of miles from open space. It’s a pretty sweet location.
Steve started, expressing the multi-ecosystem view typical of leaders in the connected community,
“We’re unique [the company Xilinx]. We’re actually part of the ARM ecosystem in a way. We are a customer of ARM’s. ARM is a customer of ours. They use our products in various ways. Then we have our own ecosystem around our platform, aspects of which involve ARM, some of which don’t. Some of which complement ARM and ARM is part of the solution. Sometimes they’re not. So it’s actually a web of things.
“We have an area we call our Smarter Vision Initiative; around image and video processing systems that are adding all this intelligence to video. Interpreting what’s in video say, helping in driver assistance. Processing of video and deciding if there’s a kid going across the street so that your car can automatically save the child.”
Steve became audibly energized as he talked about a hybrid ecosystem that mixes open source with commercial contributions,
“Obviously we have a huge vested interest to tap into these hybrid models … So we tap into open source, for example around ARM’s embedded Linaro.
“There’s something called Open CD, for example, which is an open source library of functions that people use to build these Smarter Vision systems.
“With OpenCD we even modified some things so it would automatically go into a high-speed hardware acceleration function. The customer can have it either way — use the accelerator, or simply run OpenCD on ARM cores. There can be an enhanced version of OpenCD, or a plain one.
“So if someone has a special purpose that is very relevant to meeting their system requirements, they can incorporate a variety of specialized dedicated hardware.”
Steve continued, highlighting how their Smarter Vision ecosystem is composed, looking forward.
“We have major IP partners that invest in intellectual property — large [general purpose] subsystems, real-time video engines for example, or [vertical-market applications] that do something in, say, communications for security. We can’t do all the IP ourselves.
“[So then partners are integrating] hardware, software, intellectual property, tools, design services ‘cause we can’t do all the design services. If someone wants help, you name it, we do it [within the ecosystem model]. In the case of IP, whether it’s for profit, whether it’s open source, whether it’s a hybrid of all of that, which it often is.”
I ask what keeps him up at night?
“Well, I’ll let Dave jump in a minute, but I’ll tell you the thing that keeps me up at night …
“Our goal is to provide an ecosystem that complements what we do and aggregates into solutions that our customers ultimately are demanding. Now that is a challenge. Let’s say that we end up with hundreds of millions of dollars of business associated with one ecosystem partner’s capability. What happens if that ecosystem partner gets acquired –– or goes away?
“So there’s this tension [about dependence]. Our biggest customers sometimes demand that we take more direct ownership of some of these key elements. Or they ask us to back it in some way even if it’s from a third party. They sometimes want us to step up and take a bigger role in terms of liability and quality assurance and integration.
“Ideally it all sounds wonderful to spur on all these complementary elements of an ecosystem that add up to all these different solutions that no one company could ever create.
“But then practically it creates at some point almost too much success. In certain areas it creates extremely difficult dilemmas in terms of business continuity and assurance — for some of our big customers and certainly for us when we start having a dependency that’s too high.
“Obviously you could solve this if you have several sources and they are all equal. But often in these cases they are not all equal. There are only so many people that are really good at what they do.”[v]
Dave Tokic came in.
“Ultimately Xilinx can’t do everything. We’re not going to vertically integrate. That’s not our business, it’s not our strategy
“So a key piece is having a healthy ecosystem that can supply these customer needs, and the challenge is that these tend to be pretty complex systems. And the intelligence behind them requires specialized intellectual property products, whether it’s hardware or software or expert services to help augment the capabilities of the end customer that’s trying to implement that system.
“It’s not like Apple’s app store. These guys are typically PhDs in image processing or in communication theory and there are just not 10,000 of those guys or gals out there.
“So we’ve evolved as a company to have a more strategic relationship with our customers and they’re looking to us to provide guidance. That, in turn, requires us to have a much deeper relationship with the providers. We need to proactively get them on board our latest technologies, align roadmaps, help with quality assurance.
“So I’d say it’s, it’s interesting. It’s not a ‘spread seeds all over the field and let’s just reap what is sown.’ You have to be very, very picky about planting a particular seed and watering it carefully in order to get the best blooms out of that plant.”[vi]
Xilinx has a variety of mechanisms it is using with its third parties. They range from training, to helping fund specific technology development, to taking equity investments. In rare cases they have acquired a firm.
Steve and Dave take a population approach to the partners that make up their ecosystems. They have about 600 parties that they deal with, with about fifteen in the top tier, providing mission-critical services on programs like Smarter Vision. These are companies Dave works with very closely. Then there is a second, much larger tier, that also gets active assistance, but through a less-directly-engaged consultative and training approach.
An ecosystem, like any organization or team, requires leadership. It may have a different form from a company, but it needs shared vision, management systems from aligned roadmaps to certification and testing. Partners have staffing, training and coaching challenges. Sometimes they have business model or funding problems. Steve and Dave are continuing to expand the range of ways they can help their partners be expert and successful.
In addition to providing leadership to the ecosystem members, a second role Steve and Dave have carved out is looking for gaps in their technology, in their services, and in their partner network. The goal is to identify what they call “boulders” — substantial problems holding back customer and partners, of a magnitude that is difficult to address except with guidance and perhaps funding from the Xilinx level. In this way Xilinx is operating similarly to what we have seen in earlier chapters in those working in new markets —systematically identifying customer pain and problems, and systematically putting together initiatives to solve them.
The leaders of these companies see themselves involved in many ecosystems, some broader, some deeper, all overlapping and reciprocal. “I’ll be in your ecosystem if you’ll be in mine.” Each company has its own unique and evolving configuration, shaped in constant interaction with other companies.
It should come as no surprise that the technology world continues to come up with advances in large-scale, fast-moving cooperation. The industry is deregulated, its hiring practices are very open and diversity-friendly when compared to other fields, it is international, it has a culture that celebrates science-and-innovation, it’s well funded and it creates a lot of labor-saving value and thus is highly paid by the society as a whole.
Twenty years ago observers “discovered” the growth medium that was Silicon Valley. The large, first-generation business ecosystems were established in its soil. Intel, Microsoft (a Redmond version), Oracle, Cisco, Google. Books were written about the Valley’s open, cooperative business style, its speed of thought and action. Other towns, London and Singapore, Bangalore, Austin, Pittsburgh and New York, as well as San Francisco just up the bay — worked to replicate its key features, and in many ways they were successful.
Today a new growth medium, even more collaborative and open has emerged. The stunning thing about the new connected community is that it is not centered in the United States at all. It is truly global. To the extent it partakes of a regional culture of origin that would be the reserved “just enough” British scientific and academic culture. Debate is considered excellent and vigorous when it is quiet, reflective, and settled by fact and clarity of thought.
This book is an attempt to describe some of its features. The leadership lesson in this chapter, it seems to me, is that this new environment is so rich, so supportive, that it broadly makes sense to go for it in establishing ecosystems. By this I certainly don’t mean to be imprudent, but I think the general advice of the leaders in this chapter is that almost any initiative that is of interest to you might be enriched by an ecosystem approach.
[i] Phil Carmack, NVDIA, personal communication, April 2013
[ii] Robert Hormuth, Dell, personal communication, April 2013
[iii] Paramesh Gopi, AMCC personal communication, April 2013
[iv] George Grey, Linaro, personal communication, March 2013
[v] Steve Glaser, Xilinx, personal communication, April 2013
[vi] Dave Tokic, Xilinx, personal communication, April 2013