Intel never causes shockwaves with its CEO choices. It has never selected an external candidate, and the newest leader, Bob Swan, has been doing the job on an interim basis for seven months already, since the abrupt departure of predecessor Brian Krzanich. Swan is only the seventh CEO of Intel in its 50-year history and while that has given the company the benefits of continuity and consistency of vision, it also limits the scope for innovation and a radical approach. And a hefty dose of radical thought is in order, for a firm which lost its ‘world’s biggest chipmaker’ status to Samsung (off and on) in 2017, and which is seeing intense pressure in its core markets, coupled with only modest growth in new ones.
“It’s a daunting task to turn around this aircraft carrier,” Jim McGregor, principal analyst at Tirias Research, told EETimes.
In recent years, there have been many reasons to doubt Intel, but its slow and steady approach has proved the dire warnings wrong. Its data center processor business is robust and attacks from the ARM community have largely bounced off like arrows off a rhino’s armor. Even the PC microprocessor market has not shrunk at the speed once predicted.
But the firm remains very over-reliant on these core sectors. Its 15 years of trying to break into the cellular devices big-time resulted in serial failures, and finally it bowed out (though it remains a top three provider of smartphone modems thanks to its acquisition of Infineon Wireless). That failure was disastrous because it has kept Intel out of nearly all the device categories which have started to replace the PC.
Success in a new wave of IoT gadgets and gateways is not assured either, and while ARM consolidates its position in the low power use cases, there is the looming threat of the open Risc-V processor architecture too.
Krzanich’s Intel made a shaky start in IoT. It focused too heavily on consumer devices, particularly smart glasses, watches and other aspects of ‘smart clothing’. These have grabbed a lot of headlines but have been disappointing in terms of growth – and more importantly, they tend to be made by the same companies which rejected Intel in the smartphone. Samsung, Huawei and so on have their chip suppliers in place (sometimes their own chip subsidiaries).
Apple’s adoption of Intel modems for the iPhone is important for the former Infineon business but does not progress the Intel processors at all, since Apple has its own (and may even move away from x86 in the Mac over time).
Being a top five player is not enough for a company of Intel’s scale – if it is to seize its ‘global number one’ title back from Samsung it must grab the lead in some categories with real growth potential, and high value. That means making serious efforts to chase new and premium types of IoT processors, where there are no incumbents to topple, and the market is wide open. Innovative modem/processor/memory combinations for connected vehicles, trains, robots, manufacturing equipment and so on are where Intel has its real opportunity.
Under Krzanich, it has made some of the right initial moves, investing in start-ups and engaging in trials of systems for drones, driverless cars and others. But Swan needs to be far more aggressive in trying to drive an overall platform for the massively fragmented IoT device market.
There is no one chip design that will suit all these connected end points, from trains to lightbulbs, but there can certainly be a common architecture and stack. The x86 design spanned low power notebooks (and less successfully, handsets) right up to supercomputers, and the new generation of chip architectures (Intel or not) must have an even bigger spectrum, but also be the physical foundation of a full developer stack with security at every layer, rich developer tools and open APIs (application programming interfaces).
ARM is building this stack, with help from the finances of its parent Softbank’s Vision Fund, and there will be open source contenders too. Back in the 1990s, when Sun decided to provide its Sparc processor architecture as OpenSparc, the idea of an open processor platform was ahead of its time, but in the era of the IoT, there is a far stronger case for a processor that any foundry can build without a huge IP licensing cost, in order to drive costs down and scale up for all the massively high volume use cases we envisage – billions of low power devices right up to hyperscale cloud data centers. Initiatives like RISC-V, with its fully open source design, are gaining acceptance, and there has been speculation that IBM might donate the Power architecture to open source.
Away from devices, the more promising growth area for Intel is to extend its activities in data centers, FPGAs and memory further into the cloud. Krzanich made some important strategic decisions which could lead to breakthroughs in some areas. One was to buy FPGA maker Altera, and a fully unified Xeon/FPGA platform would strengthen Intel’s lead in Cloud-RAN and many hyperscale segments (see separate item on Xilinx). Another was to establish a foundry business (though Altera was the biggest customer). Both of these could drive expansion, but it is early days to assess their impact.
And other strategic opportunities were missed – the sale of Wind River and of McAfee for instance, two jewels which Intel did not leverage to their full potential. It remains to be seen whether Altera is another in that long line of Intel acquisitions which make perfect strategic sense but are then under-exploited.
The escalating political and trade wars between the USA and China only add to the risk, since Intel has been cultivating many investments and partnerships in China, including its stake in modem maker Spreadtrum, to try to ensure it is part of that country’s bid for semiconductor self-sufficiency, rather than merely the loser from it. The harder it becomes for China to buy US chips, or sell to US partners, the more difficult that critical aim will be for Intel to achieve.
Despite the sense that Intel finds it dangerously difficult to try anything really new – and its choice of new CEO will only confirm this – it is currently doing fine with its tried and tested formula. It recently recorded its third consecutive year of revenue growth and is forecasting another record year in fiscal 2019.
To keep the momentum going, Swan needs to convince investors and customers that Intel does have a long term strategy that will still make sense even if the PC dies (including the ‘post-PC’ market that has helped slow the decline), and if some of the data center revenues are lost to alternative architectures or the internal chip developments of the webscalers.
One of the biggest near term decision will be about next steps for Intel’s manufacturing processes, which remain its crown jewels and make it very hard for others to catch up – especially the fabless firms which cannot define the agenda in the way Intel can. However, Intel has been challenged in achieving good yields from its 10-nanometer process, and the massive cost of making nearly all its chips itself is leading some investors to call for it to outsource manufacturing.
Swan’s background in finance, rather than engineering – he was previously Intel’s CFO, and before that CFO of eBay and others – may make him more likely to take such a step. And, by contrast with other Intel chiefs, he has only been at the company since 2016. Krzanich came from the manufacturing side himself, though his predecessor, Paul Otellini, was the first non-engineer to hold the top job.
Supporters of the choice of Swan hope that his short period at the company so far means he will still have an outsider’s view, but combined a deep knowledge of Intel’s processes and challenges, which an external hire would not have. However, if he was the first choice for the post, it begs the question why it took seven months to confirm him, and inevitably there have been reports that other candidates turned down the opportunity.
Andy Bryant, Intel’s chairman, and former CFO, said the company’s search committee chose Swan after conducting “a comprehensive evaluation of a wide range of internal and external candidates to identify the right leader at this critical juncture in Intel’s evolution.” He hinted that the long wait for the decision was because Intel needed to assess Swan’s performance as interim CEO (and, reports say, persuade him he wanted the role on a permanent basis, since he had previously denied that).
“Bob’s performance, his knowledge of the business, his command of our growth strategy, and the respect he has earned from our customers, our owners, and his colleagues confirmed he is the right executive to lead Intel,” Bryant added.
However, the new CEO will face new competitive challenges, and the way to deal with this will be to harness five decades of experience and many billions of dollars of investment in silicon and process R&D, and use that weight to drive a full stack which will be hard to fight against. That will mean making some elements far more open, though Intel will stop short of open source. It will surely, though have to consider licensing some of its elements to broaden its reach into a far more diverse and fragmented industry.
Under Krzanich, the company took some steps forwards towards this broader platform and ecosystem, but also some back. Intel acknowledged, finally, that x86 CPUs could not do everything. For many specialized workloads, at server or device side, other chips and accelerators will be needed, and those may be powered by FPGAs and GPUs, or by new types of chip. So the acquisition of Altera was important, as were the smaller purchases of the ARM-based platforms of LSI Logic and Picochip. These started to diversify Intel’s architecture and make it better positioned for demanding cloud applications like virtualized RAN.
Another step forward was to open x86 up to partners. While various companies licensed x86 in the early days of the PC era, in recent times AMD has been the only significant licensee. That did not change on Krzanich’s watch, but there were some interesting deals in China, which showed he was open to greater flexibility in relationships with partners. For instance, in 2014 Intel agreed to provide basic mobile SoC designs to Chinese chip vendor Rockchip, which then created its own variants optimized for the needs of the local market.
Such moves raised hopes that Intel was on the way to building an ARM-like family of partners to take its platform into all kinds of new markets which it finds it hard to penetrate directly, for all kinds of reasons of cost and history. The IoT will be full of such segments, and when Intel unveiled its Quark IoT-focused architecture, it said this was “open”. The chip giant would offer “fully synthesizable” chips with hooks in the silicon to add others’ IP blocks, to customize designs for specific tasks.
But Intel did not follow through with bolder moves into licensing, in a world where IP and common platforms are the route to power and profit, not churning out every low margin chip and sensor. And it killed off most of the projects based on Quark, finding that the IoT maker market it had targeted, as a bridge into a full offering in the microcontroller space, was outside its comfort zone.
This is another example of how Intel often has the right vision, but is over-cautious about how it follows through. For all his competence, there have been few signs during his period as interim CFO that Swan is the leader to change that. Of the five EVPs (including Swan), Murthy Renduchintala – the previous Qualcomm co-president who joined Intel in 2016 and is now group president of technology, systems architecture & client group – would have seemed more likely to innovate. It is said that the Intel board was split between him and Swan.
Swan may prove to be a more radical thinker than his interim CEOship and his CFO background suggest. But this will not be the time to tinker. After all, Intel’s glory days were founded on the truly radical decision in the 1980s, by then-CEO Andy Grove, to transition Intel from the mainframe memory business to microprocessors. He remembered in his memoir: “The fact is that we had become a non-factor in DRAMs, with 2-3% market share. The DRAM business just passed us by! Yet, many people were still holding to the ‘self-evident truth’ that Intel was a memory company. One of the toughest challenges is to make people see that these self-evident truths are no longer true.” The same thinking, and the willingness to challenge sacred cows, will be critical to Swan’s tenure as CEO.