Intel CEO Brian Krzanich Is Abruptly Out. What Did His Five Years Amount To?

He solved some of the problems Intel ran into while looking for its future, and created some new ones.

Brian Krzanich, the sixth CEO in Intel's history, has ended his five years in the job and, with it, a 30-year career at the company. Intel's announcement named the reason plainly: an affair with an employee.

"Given the expectation that all employees will respect Intel's values and adhere to the company's code of conduct, the board has accepted Mr. Krzanich's resignation," Intel said. He had violated Intel's "non-fraternization policy," which bars managers from dating subordinates who report to them.

CNBC, citing an insider, reported that the relationship had in fact ended some time ago and that Intel only discovered it recently — but Krzanich still had to go.

According to Intel's announcement, the relationship was "consensual," which means no harassment, no sex-for-power trade, no coercion — he can hardly even be called a chauvinist. Few CEOs at the world's giant companies leave their posts over a romance with an employee the way Krzanich did, and Intel's statement gave him no graceful way out: "for family reasons" or "for personal reasons" are the usual formulas when an executive departs.

Nor was he pushed out for doing the job badly. Over his five years as CEO, Intel's stock rose 120% in total, 54% in the past year, and another 16% in the first six months of this year.

For a company that has spent years feeling its way through a transition, Krzanich's record is a mix of credit and blame — but on the whole Intel has held steady and edged up.

No hope in smartphone processors, but Intel got into mobile through the equally important communications chip

By the time Krzanich took over in 2013, Intel had already missed its best chance to make CPUs for smartphones.

His predecessor, Paul Otellini, turned down the business for the first-generation iPhone in 2007, on the grounds that phone chips might not make money. Otellini also kept putting Intel's newest technology into desktop processors rather than mobile ones.

That judgment proved extremely expensive. Intel shook off AMD and took the server market, but missed smartphones, which grew much faster. It has been trying ever since and has never made up the ground.

Because Intel had spent years tilting most of its resources toward PCs and servers, it was not until the end of 2013 that it shipped a system-on-chip, codenamed SoFIA, to serve as a phone processor. Besides the Atom X3 processor, the board carried a 3G baseband — communications-chip technology that came from Infineon, which Intel bought in 2010 for $1.4 billion.

Phone brands including Lenovo, Motorola, ZTE and Asus later released products using Intel phone chips, but the big shippers — Samsung, Xiaomi, Huawei, OPPO, Vivo — never adopted them.

In recent years, though, Apple and Qualcomm have fallen out repeatedly over patent fees, taking the fight as far as the courtroom. Intel benefited from the quarrel and got a new opening.

In 2015, Krzanich said Intel had a team of more than a thousand people designing chips for Apple and might start supplying it with modems in 2016. Finally, beginning with the iPhone 7, Intel got its chance to supply basebands to Apple alongside Qualcomm.

An estimated 30% of iPhone 7 units used Intel's LTE baseband chip, the Intel XMM 7360, with a theoretical top download speed of 450 Mbps — a quarter slower than Qualcomm's part. To keep the two versions feeling the same in use, Apple deliberately capped the data throughput of the Qualcomm iPhone 7. The move caused an uproar.

But the people doing the shouting were mostly trade press and a handful of hardware enthusiasts. In ordinary use, you would not feel the difference on your phone.

On last year's iPhone 8 and 8 Plus, Apple stuck with its mixed-supplier strategy. Intel and Qualcomm supplied the XMM 7480 and the MDM 9655 for the iPhone 8 Plus. The XMM 7480's theoretical downlink speed was 600 Mbps; Qualcomm's X16 baseband could reach 1 Gbps.

No matter how wide the gap, though, Qualcomm was destined to lose Apple's baseband orders. In February this year, Ming-Chi Kuo, the Apple analyst at KGI Securities, said every iPhone would use Intel basebands from this year on and Qualcomm would be struck off the supplier list entirely.

Much of Intel's edge over Qualcomm came from its PC business, which throws off enough profit that Krzanich could let Intel take less. At Qualcomm's priciest, the hardware fees and patent royalties Apple paid it came to nearly a quarter of the cost of each iPhone; Intel's hardware cost $20 less per unit, though it has never said on what basis it charges Apple patent fees.

Phone CPUs have little to do with Intel now, and even the high end has little to do with Qualcomm: Samsung, Huawei and Apple all have their own processors.

The baseband processor is a piece of hardware a phone cannot do without and that costs a lot. Winning Apple meant Intel had won the richest customer in the industry.

Qualcomm is responding. Last month it lowered its 5G patent licensing rates, hoping to settle with Apple. Whether Apple's position shifts remains to be seen.

Reports that Apple is developing its own baseband chip have also surfaced from time to time in recent years. For Intel, winning Apple's orders for now is not victory — it is the start of the war.

Data-center server CPUs became Intel's pillar, but during the chip-flaw crisis Krzanich made decisions investors would question

In 2016, Intel under Krzanich carried out its biggest layoff in nearly two decades, hitting 11% of its global workforce, 12,000 people in all.

"Data center and Internet of Things businesses can deliver 40% of our revenue. Building on that, we must move from being a PC company to a company that powers the cloud and billions of smart, connected computing devices," Krzanich wrote, bluntly, in a memo to employees.

That meant Intel's once-central PC microprocessor business would be pared back too. Krzanich and Otellini before him had both noticed the global PC market sliding: from 2006 PCs began a 12-year decline, and Intel began its own step-by-step layoffs and transformation.

Starting in Otellini's time, more than $3 billion saved through several rounds of layoffs went into building what is now Intel's most important business, data centers — which has also proved to be its most profitable. In the first quarter this year, Intel's net income reached $4.5 billion and earnings per share 93 cents. Intel credited the growth to its data-center business, which accounted for nearly half of first-quarter revenue growth. Data centers brought in $5.2 billion that quarter, up 24% year on year.

Because Intel started early here, it has long held a monopoly position in the market. In 2016, by the count of analysts at Mercury Research, Intel had 99.7% of the commercial server market, and its rival AMD did not put out a competing product until 2017. Last November, its old opponent Qualcomm also moved into cloud servers, only to announce six months later that it would "stop developing chips for data center servers" and shut down its Centriq server processor division.

But the competition has not ended. Last June AMD released its EPYC 7000-series high-performance data-center processors. Nvidia — a graphics company that grew out of gaming — laid out a data-center chip strategy built around artificial intelligence, releasing GPUs aimed at high-performance computing and AI work. Some cloud providers are also designing chips for their own AI and machine-learning tasks. Those chips are unlikely to replace general-purpose server processors, but they are proof enough that these companies are keeping a wary eye on Intel.

On this month's earnings call, chief financial officer Robert Swan said Intel's data-center business should keep growing quickly in the second quarter of 2018, but that the pace would slow.

Swan said Intel would face "more intense competition" in the second half of 2018 than in the first, as several data-center processor suppliers step up shipments of new products. Analysts at Nomura Securities think Intel could lose 15-20% of its data-center processor market share in the fight with AMD.

AMD's choice is either to release better products and eat into Intel's share, or to give ground on price and build an advantage there. If Intel matches AMD on price to hold its share, revenue growth could slow and gross and operating margins could fall.

And for Intel and AMD alike, most data-center processor customers are large companies such as Google, Amazon and Facebook. Selling to companies that size tends to leave a supplier little room to bargain and little say in the deal.

After the Nomura report came out, Intel replied: "We do need to prepare for a 'more challenging' environment in 2018, but we are confident, because Intel is still in a leading position."

How far the "chip-flaw crisis" that has run for most of this year will dent that "leading position" remains to be seen. Intel is not the only company whose devices were affected, but because its processors hold the largest share in both PCs and data centers, it may take the biggest hit.

Intel's own tests found that computing performance suffered once the patches were applied. Most ordinary users would not notice so small a change on their PCs, but for the data centers and cloud providers that install Intel processors at scale and sell computing performance as a product — Amazon AWS, Microsoft Azure, Alibaba Cloud and the like — the effect was impossible to ignore.

More striking still, Krzanich sold $39 million of Intel stock a month before the crisis broke, which earned him a reputation for being "irresponsible." An Intel spokesperson said the sale followed a pre-arranged stock-sale schedule, had nothing to do with the "chip-flaw crisis," and left him holding the number of shares the company requires.

That account is not hard to pick apart. Bloomberg reported that Krzanich drew up the sale plan in 2015 and that this round of sales was scheduled for February 2018 — yet he actually sold in November 2017. He amended the plan in October, when the flaw had not yet been made public. But according to the records of Jan Horn, the Google Project Aero member who found the flaw, Horn had notified Intel of the vulnerability in June 2017, so Krzanich was unlikely to have been in the dark when he changed the plan.

More awkward still, he sold 880,000 shares at $44 each. After the flaw was disclosed, Intel's stock did drop to $42.50 that week, but it soon climbed back and closed the first quarter at $52. Had he not gone to the trouble of changing the schedule, he would have made more.

By the time he left the CEO job, Krzanich held only the 250,000 Intel shares the company requires of him.

Looking for "the next smartphone-sized market" in VR, drones and other new hardware did not work

Krzanich's five years as CEO were five years of Intel trying things and getting them wrong across many fields. It kept looking for the next mobile-processor market after smartphones, and the healthy profits of the data-center business gave those attempts, sound and unsound alike, at least a base of capital.

In 2013, aware that the PC market's day had passed, Intel pinned its hopes on nurturing the tablet market. To put more Atom chips inside cheap Android tablets, it subsidized Chinese tablet makers heavily for a time. Bernstein Research analyst Stacy Rasgon said that in 2013 Intel paid Shenzhen manufacturers $51 for every Intel-powered tablet, close to a quarter of a tablet's price. In 2014 it also sold them chips cheaply, barely above cost, at almost no profit.

But tablets did not become the next mobile-processor market. In 2014, subsidies got Intel to 40 million units shipped. That same year it lost $4 billion in mobile. By the middle of 2015, shipments of Android tablets running Atom processors had fallen to 10.8 million.

Jan Dawson, an analyst at Jackdaw Research, reckoned that Intel wasted at least $10 billion in the mobile market.

Then came AR/VR and wearables. In 2013 Intel set up a New Devices Group, or NDG. Across 2014 and 2015 the unit bought the health-watch company Basis and the cycling head-up display company Recon, and worked with fashion watchmakers such as Fossil and TAG Heuer to supply chips for their smartwatches. Intel also made a run of smart products you have probably never heard of: smart earbuds, health bracelets and the like.

None of these businesses ever contributed anything to Intel's revenue, and they were plainly not the next mobile-processor market it was hoping for. More than one giant and more than one startup have since shown, at their own expense, that the entire wearables market was a bubble. In 2016 it was reported that Intel had cut 80% of Basis's staff. In July 2017 the business was killed off for good.

NDG then turned to VR hardware. Interviewed by TechCrunch at the end of 2016, Krzanich still had grand hopes for the VR market, saying Intel's VR project was in its early days but would eventually become one of Intel's key businesses. In fact it was abandoned five months later. Intel then tried RealSense depth sensing and a WiGig-based wireless headset system, and failed at those too.

In April this year, Intel announced it was dropping NDG's last big project, the Vaunt smart glasses. That same month, it said it would soon close the unit, which had already cost it hundreds of millions of dollars; the remaining staff would move to other Intel departments or be let go.

Finally, one attempt that barely made a ripple: drones. In September 2014 Intel began working with the American drone startup 3D Robotics to develop the Edison chip. That November it put $10 million into the drone startup PrecisionHawk. The following May it invested an undisclosed sum in the American drone company Airware, and in August spent $60 million on the Chinese drone company Yuneec. Early the next year it bought the German drone company Ascending Technologies, again for an undisclosed sum.

There is no way to know exactly what all this shopping cost Intel, but it is clear that drones were, for a while, one of its bets in the mobile-processor market. In 2016 Intel finally produced its first Intel-branded commercial drone, the surveying Falcon 8+, and its first drone built purely for light shows, the Shooting Star — and then almost nothing was heard of either.

The Shooting Star last caught the public's attention in February this year, in a drone light show at the opening ceremony of the Pyeongchang Winter Olympics. "But we're not going to sell it in stores," Krzanich said.

Krzanich bought Intel a place in the driverless market with huge sums, but the price was steep

In July 2016, at Fortune's Brainstorm Tech conference, Krzanich talked about the company's future once again.

"Unfortunately we've already missed the smartphone business. I hope we don't miss smart cars too," he said, adding that smart cars would be the next generation of the mobile-device business.

Nvidia got there far earlier. As far back as 2011, Tesla used Nvidia graphics chips in its first all-wheel-drive sport sedan, the Model S. By 2014, 14 carmakers including Audi and Lamborghini were customers. In March 2015 Nvidia released Drive PX, an in-car computing system for self-driving cars built on two Nvidia Tegra X1 chips, which it claimed was 3,000 times faster than DAVE, the autonomous-driving technology developed by the U.S. Defense Advanced Research Projects Agency. That May, Nvidia also began selling software and hardware toolkits to carmakers, supplying the technology for self-driving cars to recognize and avoid obstacles.

It was not until this same point in 2016 that Intel put together a smart-car R&D team, announcing it would provide data processing and analysis technology for the driverless cars BMW planned to mass-produce in 2021. The work would be done by Intel together with Mobileye.

For a latecomer like Intel, the simplest and most effective way to keep the gap from widening was to buy its way in. After working with Mobileye for a while, Intel spent $15.3 billion last March to take it over outright. Before that it had already bought a pile of companies from various countries with some connection to driverless cars — Yogitech, Arynga, Itseez, Nervana Systems, Movidius.

Mobileye is a formidable Israeli company in the driverless field, and its main products are advanced driver-assistance chips and systems. Its autonomous-driving package was used in Tesla's first-generation Autopilot, and carmakers such as BMW, GM, Volkswagen and Tesla have all been customers. Mobileye says it has taken more than 70% of the market. What Intel bought was not only market share, R&D experience and data, but also, crucially, those customer relationships.

It was the second-largest acquisition in Intel's history. Intel's earnings report two months before the deal showed total cash flow of $21.8 billion, so the purchase consumed at least 70% of its cash reserves. For Intel and for Krzanich, how much rode on it is easy to imagine.

A few months before buying Mobileye, Intel set up a new autonomous-driving division to handle everything car-related. A month after it was formed, the division launched Intel Go, a development platform for self-driving cars. Manufacturers that choose Intel Go can reportedly pick any of Intel's existing processors as the core of the vehicle's computing unit, from the weakest Atom to the strongest Xeon, and get support from Intel's cloud data centers as well. For Intel, that was one more good chance to push its own chips.

In May this year, Mobileye disclosed that it had signed a large contract to supply autonomous-driving technology for 8 million vehicles from a European carmaker, without saying which one. If that holds, Intel, having spent so much, should finally start seeing money come back. By Intel's own earlier estimate, the driverless market could be worth $70 billion by 2030.

All told, in Krzanich's five years Intel reorganized its main business and made a series of big-money forays into new fields. Once the number-one player in PC chips and the most valuable chipmaker in the world, Intel has followed the currents of the technology industry, working hard and struggling through one change after another — and perhaps changing CEOs is part of that too.

Cover image from: Intel