Your browser is not supported. Please update it.

21 March 2013

ARM and Intel to gain new CEOs at the same critical moment

Significance: ARM’s CEO Warren East is to step down in June, just after Paul Otellini leaves the helm at arch-rival Intel. The two firms gain new leadership at a critical juncture, when both are battling in almost every processor market from mobile and embedded devices to servers and base stations. In contrast to Intel, which is said to be struggling to appoint a new chief, ARM will have a smooth handover to the current president, Simon Segars. However, while ARM may be steering a stable ship, its OEMs are navigating tumultuous waters. Some have fallen by the wayside, most recently STE-Ericsson, while others are chasing new markets, notably Nvidia, whose latest roadmap pushes further into the enterprise and high end tablets. Intel needs to look across the pond at its potential nemesis and realize the benefits of a smooth transition, as it faces key decisions about its future ‘ even whether to license the ARM architecture again, in order to broaden its foundry business.

There is a certain neatness to the fact that Intel and ARM, battling more fiercely than ever to dominate the processor landscape, should be gaining new CEOs at the same time. Intel chief Paul Otellini is to step down in May, though no successor has yet been found ‘ and the process is taking longer than expected, according to sources, highlighting the critical nature of the decision. Meanwhile, ARM CEO Warren East will stand down on June 30, handing the reins to his right-hand man, Simon Segars, currently company president. But if the ARM handover will be a smooth one, and none of the major changes expected at Intel will ensue, the market its processor platform supports is in turmoil. ST-Ericsson is the latest to give up the battle to rival Qualcomm in the mobile device space, but Nvidia has stepped up its bid to fill that slot with an updated roadmap for its Tegra family.

For firms like Nvidia and Broadcom, Qualcomm, not Intel, is the company to beat, but the strategists at ARM, keen to push the architecture well beyond devices, see Intel everywhere they try to play ‘ blade servers, wireless infrastructure, embedded chips for the internet of things. The plans and abilities of the new CEO, whoever he or she may be, will be watched by nobody as closely as Segars. Of course, there would be an added neatness if East were leaving to cross the pond to Intel, as some have speculated in the past, but he insists he is retiring.

He would have faced all kinds of challenges in adapting from an IP-based company to a massive manufacturing juggernaut, but he would have brought Intel perhaps the thing it needs most ‘ a change of focus away from obsessing with its own Moore’s Law, and towards laws centered on low power consumption. ARM has thrived wherever low power processors were required, and that landscape has broadened dramatically from relatively niche mobile gadgets in its early days, to just about everything electronic, even servers and base stations. Meanwhile, though Intel’s capabilities in performance growth are still essential, Moore’s Law needs to be superseded by, or at least given equal weight with, others like Stanford University’s Koomey’s Law (just as Moore predicted that processing power, measured by the number of transistors on a chip, would double every 18 months, so Koomey rules that energy efficiency also doubles in the same time period).

The ability to drive down power consumption has been the single biggest factor placing ARM at the center of a growing and powerful web of chip providers, collectively targeting Intel on all fronts. This will be East’s legacy, and a far more unambiguous one than Otellini’s, tainted by the failure to break into the mobile world more effectively over the past decade. By contrast, East will leave a record of remarkable growth during his 12-year tenure. During that period, ARM has received royalties for over 40bn chips based on its cores, while the number of ARM-based chips has risen from 420m in 2001 to 8.7bn last year. That has driven a fourfold rise in earnings per share during the period and a current market valuation of almost £13bn.

John Buchanan, ARM chairman, said in a statement: ‘Warren has transformed ARM during his time as CEO. In 2001 ARM had one processor product line found mainly in mobile phones. ARM now provides the broadest portfolio of technologies in the industry, used by more than 300 semiconductor customers in nearly 9bn chips last year. As CEO he has created a strong platform for growth and consistently created value for shareholders, even in a challenging external environment … On behalf of the Board, and the wider ARM team, deep thanks are due to Warren for his passion, service and leadership.’

East, who is 51, joined ARM in 1994 and held various roles, including head of consultancy, head of business operations and COO, before taking over the helm from Robin Saxby. Segars joined ARM in 1991 as an engineer and has since taken a major role in supporting East in growing and expanding the firm’s business. It is now pushing its core IP into post-PC devices, low power servers, wireless infrastructure and embedded products for the internet of things, and also has the Mali range of graphics processor cores.

Indeed, the need to combine processing power with low energy, and the push into one anothers’ markets, have brought many similarities of outlook to ARM and Intel. In 2008, when Intel celebrated its fortieth birthday, its former CTO Pat Gelsinger set out the top priorities for the next decades. These were threefold – ‘many-core’ microprocessors for massive processing and graphics power; the push into very low power applications like cellphones; and the combination of microprocessors, sensors and embedded technology to make computing ubiquitous.

The trio of priorities could have come from the mouth of East, who has driven ARM’s expansion into graphics, multicore, embedded chips and 64-bit to diversify the revenue opportunities. Few of those concepts were in ARM’s mind in 1990 when it emerged from the ashes of the UK’s attempt to build a microcomputer industry of its own. Since 1985, it had been a division of Acorn Computer, the leader (with Sinclair) of that ill-starred attempt to take on IBM and Wintel. But Acorn had valuable processor IPR, which led it to collaborate with Apple ‘ a more successful challenger to IBM PC – on the Newton chip. Acorn and Apple spun off the chip activities on November 27 1990 to form ARM, whose goal was to exploit the IPR behind the Acorn Risc Machine, later renamed Advanced Risc Machine, which from an early stage was seen as a potential global standard by Saxby, the original CEO.

This was sweet revenge indeed for Acorn, which had sown the seeds of ARM’s success by choosing a RISC (reduced instruction set) design for its original platform because it would require fewer transistors and therefore be cheaper (the opposite of Intel thinking, and momentous in the small device world, allowing for simpler, lower cost designs ‘ and, critically, with lower power consumption.

Little fundamental change can be expected under Segars, given his close alliance with East and their common view of the company’s strategy. ‘What we’ve been doing is a very appropriate strategy for responding to the dynamics of the market in which we operate,’ Segars said. ‘There will be no fundamental shift in strategy. It’s about execution and making sure we take advantage of the markets and continue to grow market share.’

East, who is retiring at the young age of 51, also stressed the need for continuity in times of change. ‘We’re doing things and making plans now for activities that will fuel the growth of ARM in 2020 and beyond,’ he said. ‘It’s important for the leader overseeing those plans and implementing those plans to be in place all the way through. If I see that through, I would be CEO for 20 years, which is too long for the business.’

This approach will certainly provide stability and predictability to ARM licensees at a time when they are engaged in very complex battles of their own. Some, notably Qualcomm, are riding high on the low power trend and others are extending their use of ARM architectures to address this need (Freescale adopting ARM rather than PowerPC for its small cell base station chip, for instance). But others have given up the fight ‘ Freescale and Texas Instruments both quit the mobile processor sector to pursue less over-competitive opportunities; and now ST-Ericsson has finally thrown in the towel, leaving the future of its NovaThor mobile platform (ARM-based multicore processor plus modem) in doubt.

The grand hopes which attended the creation of the joint venture were entirely extinguished as STMicro and its Swedish partner announced they would close down the struggling firm. The partners finalized their JV four years ago and promised a European powerhouse to balance the rising dominance of Qualcomm. Instead, they got saddled with rising debts and losses and in recent months, have failed to find a buyer for STE. Now they will wind up the JV altogether, with the loss of about 1,600 jobs.

The two firms will divide the assets and the remaining employees. Ericsson said it would take on about 1,800 of the unit’s 4,350 employees and contractors, in countries including Sweden and Germany. Perhaps surprisingly, it will also continue to develop STE’s flagship 3G/4G modem technology, despite a recent push to divest non-core businesses, especially those related to devices (like Sony Ericsson).

Its Swiss-based partner will take over other existing products. Some may be integrated back into the core range, where STMicro offers many components for mobile devices. In particular it has focused recently on chips for the internet of things, and has shown off breakthroughs like ‘non-touch touchscreens’. It will also assume 950 employees, mainly in France and in Italy, and incur costs of as much as $450m. However, neither parent mentioned the ARM processor platform.

The equally owned venture has stacked up net losses of $2.7bn since its formation in February 2009. Despite some ambitious roadmaps, it has struggled with the transition from legacy products to its new NovaThor platform, and despite some strengths such as TD-SCDMA in China, it has failed to win the major handset deals from Qualcomm and others.

There were still some hopes that the parents would find a buyer for all or some of STE. In the past two years there has been a rush for modem technology among mobile processor makers, though many have already made their choices ‘ Intel relieved another European dinosaur of its mobile unit when it took over Infineon Wireless; Nvidia bought UK start-up Icera. So the timing for STE was poor, since most app processor vendors are now on the way to an integrated processor/baseband platform, considered essential to compete in mainstream smartphones and with Qualcomm.

Sources say Samsung was one potential buyer approached, but declined to make an offer. Though the Korean firm is building up its mobile silicon armory, it already has LTE modem technology and, if it wants further assets, is likely to look to a local player, or at least to one with less baggage, in terms of older products and European employment contracts. Reintegrating employees into the parent firms should minimize job losses and placate some European governments such as that of France, which still owns a stake in STMicro.

Ericsson, which put its former Ericsson Mobile Products unit into the venture, will now run the modem business as a standalone division, which is likely to report an operating loss of SKr500m ($77m) in the fourth quarter, the Swedish firm said. It has already set aside SKr3.3bn in 2012 to cover costs related to strategic changes at STE.

STE’s CEO Didier Lamouche resigned last week and the JV’s COO Carlo Ferro was named temporary CEO to oversee the transition. The formal transfer of assets to the two parent firms is likely to be completed in the third quarter.

As one star falls, another rises, and Nvidia is intensifying its attack on Qualcomm. Unlike MediaTek, coming from low end modems into the integrated smartphone sector, Nvidia is moving from the opposite direction. It has harnessed some of its PC graphics technology for its mobile platform, Tegra, and combined it with ARM cores, initially going after tablets, where space and power consumption limits are somewhat more generous than in handsets. This is what most observers expected Intel to do, taking the natural stepping stone from the PC world via larger mobile devices, but instead ‘ as though to establish its new low power credentials ‘ it has kicked off its belated Atom push into mobility with distinctly midrange smartphones.

Nvidia eyes that space too, hence the importance of getting its Icera technology fully integrated with Tegra. However, while it needs the midmarket for volume growth and to sign broader contracts with large OEMs, its differentiation will always be at the high end, and particularly in graphics-intensive platforms geared to gaming ‘ where the firm has even announced its own games engine.

This week, at its GTC (GPU Technology Conference), Nvidia unveiled its latest roadmap, which includes two generations of Tegra, called Logan and Parker. CEO Jen-Hsun Huang boasted that the platform would deliver 100fold performance improvement from the first generation to the fifth, over the space of less than five years (Parker is due in 2015). The current Tegra4 went commercial earlier this year and Logan will emerge late this year.

Drawing on its graphics heritage even more than before, Nvidia will incorporate its Kepler GPU in Logan. Kepler has a strong position in the discrete graphics processor sector, and the US firm will also include another technology from its high end, CUDA, in the mobile platform. CUDA is a programming model for parallel processing on GPUs, which has been demonstrated right up to supercomputers.

Parker, the fifth generation Tegra, will be the first processor from Nvidia’s Project Denver R&D project, which introduces full 64-bit capabilities and was first described at 2011’s Consumer Electronics Show. There have been few details since, but it aims to take Tegra into high end products like servers, allowing Nvidia to supply both CPU and GPU in most of the products where Intel plays ‘ and where it currently merely provides graphics alongside an x86 processor. Parker will also use a new generation GPU codenamed Maxwell, the successor to Kepler.

All this graphics firepower will be important for high end devices, and to fend off rising levels of competition in tablets. Nvidia powers both the Google Nexus 7 and the Microsoft Surface, but admitted that sales of the latter device, and Windows RT in general, had been disappointing. At the other end of the product spectrum, it will be vital to deliver low power, integrated systems-on-chip for handsets, a market where Nvidia has scored few wins so far. The key step is the recently announced Tegra 4i, which integrates an LTE modem on the same silicon as the processor.

Importantly for ARM, Nvidia is also helping to drive the platform into the enterprise space, applying its graphics expertise to business as well as more traditional technical applications. It rolled out its Grid Enterprise initiative, a new layer of systems and software to provide business users with access to apps running on Nvidia GPUs. Dell, Hewlett-Packard and IBM have all signed up to make Grid Enterprise servers and appliances for large and small businesses, and Citrix, Microsoft and VMWare are among those contributing the software.

Nvidia also showed off the future, post-Maxwell roadmap for its GPUs; a new ‘tiny little computer’ called Kayla; and an end-to-end server system designed to replace dedicated workstations. Kayla is a computer on a board, with a Tegra3 processor and Kepler GPU, dedicated to intensive tasks such as simulation. The important aspect of Maxwell is that it shares virtual memory with the CPU within the CUDA environment ‘ a challenge to AMD’s similar approach with HSA, although the latter is moving beyond proprietary designs and has enlisted support from Microsoft, ARM itself and others.

These, then, were the clearest holes in Nvidia’s strategy ‘ an architecture plan which, though true to its heritage of high performance excellence, remains very proprietary; and a lack of significant progress in mobile devices. Attendees also noticed the shortage of discussion about the Shield Android gaming device, launched at CES but seen by some as a risky diversion.

Nvidia will be pleased not to have to cope with sudden changes of direction from ARM as well as from its markets and customers. While ARM is going through a smooth CEO transition more typical of Intel ‘ which habitually grooms its new chiefs well in advance and has never appointed an outsider ‘ the larger firm is still in limbo. It is under pressure to look outside to get a chief with new perspective, who might accelerate change. However, the job is a daunting one and according to a research note from Piper Jaffray, the decision has been delayed.

‘We believe Intel has narrowed down CEO candidates to two internal and one external, but the final decision appears to be slipping,’ wrote analyst Gus Richard. That is a problem not just for confidence in Intel ‘ among customers and shareholders ‘ but because some major decisions need to be made, including how far Intel goes down the contract manufacturing route (amid talk it may even make chips for Apple in future), and even whether it might license the ARM architecture for that purpose. Intel used to have an ARM licence, which it acquired with Digital Equipment’s StrongARM platform and which underpinned one of its previous failed attempts to make a mobile impact, XScale, later sold to Marvell.

Intel has just two months to appoint a CEO before Otellini leaves in mid-May and the firm insists it is still aimng to hit that target. But Richard points out: ‘Without a CEO Intel is, in our view, a rudderless ship driven more by external forces than internal decisions. The most important of which is: will the company license ARM and improve its ability to be a foundry? We think any major decision like this will have to wait for a new CEO. The fact that the decision looks to be delayed is not a positive sign in our view.’

Sources told CNet that the shortlist of three, alluded to by Piper Jaffray, consists of Intel’s current COO, Brian Krzanich; the general manager of the Intel Architecture Group, Dadi Perlmutter; and Sanjay Jha, former CEO of Motorola Mobility and before that, COO of Qualcomm. In the wake of announcing Otellini’s plan to retire, Intel promoted three executives to EVP level, taken as a signal that each of them would be candidates for the top job, but only one ‘ Krzanich ‘ is still said to be in the running (the others are CFO Stacy Smith and head of software Renee James). In the past, Intel has tended to promote its COO to the top job, as it did with Otellini, so that will put Krzanich in the spotlight. He was promoted to COO only in January but has been at Intel for 30 years, most recently as head of manufacturing operations.

Perlmutter, already an EVP, often stands in for Otellini at major presentations and has become something of a ‘face of Intel’. He also runs the same unit Otellini did before being promoted.

But while he is well known to the investment community, shareholders are likely to favour new blood and new thinking. Among the many names suggested from other firms, Jha has always been a frontrunner and a very obvious candidate. He knows the enemy, Qualcomm, inside-out, and has close ties, from both his recent major roles, with Microsoft, Google and other key players. He oversaw the separation of Motorola Mobility and its sale to Google ‘ widely regarded as a masterstroke for the otherwise struggling firm. And unlike some of the wilder cards, Jha comes with the engineering background that remains dear to Intel, even bringing expertise in an acknowledged Intel weak spot, DSPs.