Intel has a frustrating pattern of choosing the right companies to acquire, and then failing to capitalize on their advantages. Now, Wind River, in our view one of its most strategically important purchases of recent years, is joining a line that also includes DEC StrongARM/XScale, TI’s cable modems, Level One, Dialogic and Trillium. Intel is selling off Wind River to investment firm TPG (the same firm which acquired half of Intel’s McAfee unit last year). The company – headed by the unit’s current president Jim Douglas – is already pledging to pursue an aggressive strategy in Internet of Things (IoT), edge compute and intelligent devices.
But those are all areas which Intel has identified as highly strategic, as it looks for growth opportunities to offset the decline in PCs and rising pressures on servers. When it acquired the venerable embedded systems vendor in 2009, for $884m, it quickly became apparent that the firm could be a valuable weapon for Intel to expand into these new markets.
So why will Wind River succeed better on its own, than with all Intel’s resources, when its objectives do not appear to have changed greatly?
Some will cite the curse of Intel M&A, which has seen more of its acquisitions fail than succeed – though the jury is still out on recent strategic purchases like Altera and Mobileye, and at least Infineon Wireless is coming good at last, with a rising share of the modem space.
In Wind River’s case, the strategic fit was clear, but cultural and structural factors seemed to prevent Intel realizing the full potential of this jewel in its crown, or of achieving its objectives for synergies and additional business.
Now the company will have greater independence and agility to push forward its various offerings – increasingly intelligent embedded systems and real time operating systems for industrial and IoT systems in markets like transport and manufacturing; and Network Functions Virtualization (NFV) technology for operators.
Douglas said in a statement: “TPG will provide Wind River with the flexibility and financial resources to fuel our many growth opportunities as a standalone software company”.
Intel says Wind River will remain “an important ecosystem partner” for its “things group” and help Intel “collaborate on critical software-defined infrastructure opportunities to advance an autonomous future”.
But the fact remains that it has been steadily offloading activities which were once supposed to support its aim of establishing a default platform for the edge-based IoT, just as it once did for the PC. Gateway chips, connectivity, virtualization and embedded systems were key ingredients of this, but last year it ditched a series of elements in this.
In 2017, it axed its Basis wearables division; its embedded computers such as Galileo, Edison and Joule; and the Arduino maker board (a challenger to Raspberry Pi) plus its associated Curie module.
The maker market and the wearables space might reasonably be seen as non-core, or peripheral opportunities. But these divestments were followed by the end of the Quark microcontroller core, which was launched with much fanfare when CEO Brian Krzanich took the helm in 2013, and now by Wind River.
Quark and Wind River were examples of the new post-PC Intel, concerned with critical enablers of high growth markets associated with the edge and IoT. Now it appears to be retreating to its core approach – supporting new platforms indirectly in the hope they will spur new chips sales, rather than controlling the whole platform itself.
For instance, not long before the Wind River sale was announced, Intel had already decided to contribute the unit’s Titanium Cloud portfolio of OpenStack-based NFV infrastructure to open source – to the Linux Foundation’s Akraino project, heavily supported by operators like AT&T, which is concerned with “edge cloud infrastructure for carrier, provider, and IoT networks (just as Intel had professed itself to be).
Only last month, at the Open Networking Summit, Melissa Evers-Hood, senior director of cloud and edge software for Intel’s Open Source Technology Center, said: “We are doing this with the explicit intent of accelerating edge technology and innovation. Wind River Titanium Cloud is in production today with deployments of IoT and industrial edge deployments, all the way up to carrier-grade implementations of edge technology. This is intended to accelerate the ecosystem with regard to providing assets that are low latency, provide high availability, the ability to update in the field, etc.”
But there is a big gap between accelerating progress by donating some technology to open source, and relinquishing control of that technology altogether. Indeed, as examples from AT&T’s ECOMP (now ONAP) to Google’s Android show, open source platforms almost always succeed commercially when they have a guiding hand, with deep pockets, driving their development and adoption.
It is questionable whether a TPG-backed independent Wind River will have that clout in Akraino, to achieve what Intel had hoped for when it donated its code – to establish a single de facto standard for the edge cloud stack. Evers-Hood said: “We feel there needs to be one edge stack. There are lots of projects dabbling with trying to create an edge stack that is hardened and reliable and ready for production. We are announcing with this host of amazing partners that we are standing up and joining Akraino to make this project THE project for the edge.”
Weakened ability to drive Wind River’s technology to be that unified edge stack could be just one lost opportunity for Intel. When it bought the company in 2009, many were baffled, but the rise of Android, IoT and edge compute successively seemed to justify the purchase as something very strategic.
Wind River brought skills and alliances in three areas which, a few years after the 2009 deal, had become vital to Intel’s strategy for telecoms markets – virtualization, Android, and embedded systems for the IoT. Wind River brought the ability to add software intelligence, and a broad ecosystem, to emerging chip platforms for data centers, carriers and the IoT.
For instance, in 2014 Wind River launched an ‘app store’ offering a host of virtual network functions (VNFs), including virtual firewalls and load balancers, and a scheme to certify products from partners to ensure compatibility with its flagship platform in network virtualization, the Carrier Grade Communications Server.
This positioned Intel to attempt to establish a de facto standard, with the Wind River technology at its heart. As for any company of this size, Intel needs not just to sell products but to define and lead key new technologies, and in this respect Wind River has been a crown jewel.
Intel knows its best shot at a strong power base in the wireless carrier sector lies in harnessing the multilayered move towards virtualization in infrastructure and devices, moves which, for once, put Intel technologies in the cat seat in a mobile world. Having developed these kind of platforms in the enterprise, its x86 architecture has a significant headstart over ARM platforms. SDN also gives Intel the hope of carriers moving to an architecture where it is a leader and pacemaker, not on the back foot as it is in carrier networks.
But to succeed, and to gain the confidence of carriers (traditional and disruptive) Intel needs to look well beyond silicon and create top-to-bottom platforms – which is where activities like Wind River became strategic, with its long experience in hypervisors and virtualization, not just its original business of real time OS and, later, Android tools.
With some justification, Wind River characterized its offering as the “first commercial, carrier-grade platform for NFV”. Its credentials for making such a claim rested largely on its role in UK telco BT’s groundbreaking Network Evolution R&D project, some of which was fed into the ETSI work on NFV specifications. That enabled Intel to claim a stack which was specifically devised for telcos, not adapted from an enterprise offering.
Wind River’s stack included its own Linux OS, open virtualization, the KVM virtual machine, and OpenStack software with carrier-grade enhancements, all optimized for Intel Xeon processors and Data Plane Development Kit (DPDK). The idea was to offer an out-of-the-box NFV host which will then serve multiple virtual machines supporting different use case such as evolved packet core (the first application for many operators, and already demonstrated by NEC using the Intel platform).
But Intel failed to capitalize very much on these promising foundations, nor on the IoT partnerships Wind River brought in, such as the work with Axeda on a Machine Cloud platform enabled by Wind River’s venerable VxWorks real time OS and its Linux operating systems, as well as its Intelligent Device Platform (IDP).
The price paid by TPG for Wind River was not revealed, unlike the sum it paid for 51% of McAfee in 2016. That was another example of an opportunity squandered by Intel. This looked like throwing the baby out with the bathwater, since security is so essential to everything Intel is trying to do.
When it acquired McAfee, it talked convincingly about the need to embed security capabilities right down at chip level, something rival ARM has pursued with its broad TrustZone ecosystem. The strategic importance of McAfee was seen in 2014 when Intel launched a reference platform for an end-to-end IoT solution with security at every layer and link.
Ensuring robust security in networks of huge numbers of endpoints is probably the most critical challenge facing the IoT, especially when the devices are transmitting sensitive data back to the cloud. The Wind River and McAfee technology provided a middleware layer to handle device management and security, and software reference designs for a broad range of products from network appliances to CCTV cameras.
Of course, Intel will still use McAfee and Wind River technology, but it is questionable whether its future agenda will fit well with that of an investment company – short term profits are not always achievable when devising big picture architectures.
Instead, the $1.1bn McAee deal looked like a short term one, part of Intel’s radical cost-cutting exercise and its bid to double down on a few core activities. Intel, whose endpoint security is already well regarded, lost the chance to dominate one of the most important areas of technology, ensuring security at the network edge once that is supporting huge numbers of connected devices. Now it may have forfeited its chance to rule other important aspects of the virtualized, edge-focused IoT too.