Intel: Hard Reboot

英特尔：强重启，难

 

Can Pat Gelsinger turn the chipmaking giant around?

帕特·基尔辛格能否带领芯片制造巨头英特尔扭转乾坤？

 

“Success breeds complacency. Complacency breeds failure. Only the paranoid survive.” So said Andy Grove, the Hungarian emigré who helped turn Intel from a scrappy startup in the 1960s into the firm that did more than any other to put the “silicon” in Silicon Valley. They will be ringing in the ears of Pat Gelsinger, Intel’s new boss, who took over on February 15th. He takes the helm of a company that looks, from some angles, to be in rude health. With $78bn in revenue in 2020, it is the world’s biggest chipmaker by sales. It has a 93% share of the market for powerful – and lucrative – chips that go into data-centre computers, an 81% share in desktop pcs, and operating margins of around 30%.

 

“成则满，满则败。唯狂者胜。”被迫从匈牙利移民到美国的安迪·格鲁夫（Andy Grove）如是说。20世纪60年代，在他的领导下，英特尔这家朝气蓬勃的的创业公司日益崛起，几乎凭一己之力，让硅谷真正实至名“硅”。他的这句话也将在2月15日新任英特尔CEO的帕特·基尔辛格（Pat Gelsinger）的耳边响起。从某些角度看，帕特·基尔辛格掌舵的英特尔表现可圈可点。其2020年年收入为780亿美元，是全球销售额最高的芯片制造商。英特尔在功能强大且利润丰厚的芯片市场中独占鳌头，占数据中心计算机市场份额93%、台式电脑市场份额81％，营运利润率约为30％。

 

Yet Intel’s share price has underperformed those of rivals. Nvidia, a firm with one-seventh of Intel’s revenues, has a market capitalisation, at $370bn, that is half as high again. The manufacturing technology on which much of Intel’s success was built has fallen behind. It has missed the smartphone revolution. Some of its big customers, such as Apple and Amazon, are turning into competitors. Mr Gelsinger inherits quite the in-tray, then.

 

然而，英特尔的股价却落后于竞争对手。英伟达公司（Nvidia）收入只有英特尔的七分之一，但市值高达3700亿美元，再次超出英特尔一半之多。英特尔赖以发家的芯片制造技术也远落于他人之后。它错过了智能手机革命，原先的大客户（如苹果和亚马逊）也逐渐变成竞争对手。现在，基尔辛格可以说是接过了这一烫手的山芋。

 

Start with production. Chipmaking is propelled by the quest for smallness. Shrinking the components in integrated circuits, these days to tens of nanometres (billionths of a metre), improves the performance of both the components and the microchip as a whole. For decades Intel led the way, its “tick-tock” strategy promising a manufacturing revolution every other year. Now “it has lost its mojo,” says Alan Priestley of Gartner, a research firm, who worked at Intel for many years. Its “ten nanometre” chips were originally pencilled in for 2015 or 2016 but did not start trickling out until 2019 – an unprecedented delay. The technology is still not mature. In July Intel said the next generation of “seven nanometre” chips would not arrive until 2022, a delay of at least six months.

 

先说生产。芯片制造追求产品尺寸越小越好。集成电路中，组件尺寸越小，组件和微芯片整体性能就越高，如今的组件尺寸已被压缩到几十纳米（十亿分之一米）级别。英特尔几十年来一直稳坐行业龙头，其Tick-tock战略确保每两年都会进行一次制造业革命。然而，高德纳（Gartner）咨询公司的艾伦·普里斯特利（Alan Priestley）曾在英特尔工作多年，如今也坦言英特尔“已失去了魔力”。其原定于2015或16年推出的“10纳米”芯片直到2019年才缓缓问世，如此推迟在英特尔史上从未有过。该项技术也尚不成熟。英特尔公司在7月表示，下一代“7纳米”芯片要到2022年才能上市，预计至少推迟六个月。

 

Manufacturing stumbles have cost it business. AMD, its most direct rival, outsources production to Taiwan Semiconductor Manufacturing Company (TSMC), whose technology is now ahead of Intel’s. That means AMD’s chips are generally faster, and consume less power; its market share has more than doubled since 2019.

 

英特尔已经为芯片制造领域挫败付出代价。其直接竞争对手AMD公司将生产外包给台积电（TSMC），而后者的技术目前已反超英特尔。即，超微半导体（AMD）生产的芯片运行速度更快，功耗更低；该公司自2019年以来的市场份额翻了不止一番。

 

A second challenge is the industry’s growing specialisation – a problem for Intel’s traditional forte of general-purpose chips, especially if desktop pcs continue to stagnate. Technology giants, flush with cash and keen to extract every drop of performance for their specific purposes, increasingly design their own semiconductors. In 2020 Apple said it would drop Intel from its laptops and desktops in favour of custom-designed chips. Amazon is rolling out its “Graviton” cloud-computing processors, also designed in-house and made by TSMC. Microsoft, whose cloud business is second only to Amazon’s, is rumoured to be working on something similar.

 

第二大挑战是芯片制造产业不断向定制化发展，对英特尔一向引以为傲的通用芯片制造是个不小难题，台式机市场需求若持续停滞更将令其雪上加霜。各大科技巨头资金充裕，迫切希望榨取每一滴性能来满足自己的特定需求，不断更新半导体设计。苹果公司于2020年表示其笔记本和桌面电脑不再使用英特尔芯片，转而采用定制设计的芯片。亚马逊推出了Gravition云计算处理器，芯片同样由自主设计，台积电制造。云计算业务仅次于亚马逊的微软据传也有同样的动作。

 

Intel has also failed to make any headway in smartphones, the most popular computers ever made. An effort in the late 1990s to build graphics chips, which have also proved handy for artificial intelligence (AI), and to which Nvidia owes its enviable valuation, petered out. Attempts to diversify into clever new sorts of programmable or memory chips – in 2015 it paid $16.7bn for Altera, which makes them – have so far not paid off in a big way.

 

除此之外，英特尔在智能手机（史上最普及的计算机）领域的发展也停滞不前。20世纪90年代末，英特尔开始尝试制造图形芯片，事实证明也可提高人工智能便捷性，英伟达由此获得了人人称羡的估值，但英特尔对此的激情开始慢慢退却。公司尔尝试丰富产品阵容，生产各类新型的可编程或可储存的芯片，还在2015年以167亿美元收购了生产上述芯片的阿尔特拉公司（Altera），但到目前为止仍未获得丰厚回报。

 

Mr Gelsinger has yet to say how he plans to deal with the challenges. He does not look like a revolutionary. He began working at Intel aged 18, before leaving in 2009 to preside over EMC, a data-storage firm, and for the past nine years heading VMware, a software firm. In an email to Intel’s staff after his appointment was announced he invoked its glory days, recalling being “mentored at the feet of Grove, (Robert) Noyce and (Gordon) Moore”, the last two being the firm’s founders. Like them but unlike his predecessor, Bob Swan, Mr Gelsinger is an engineer, who in 1989 led the design of a flagship chip.

 

基尔辛格尚未透露他将如何应对这些挑战。他不像一位革命者。从18岁起他就在英特尔工作，2009年离职后开始掌管信息存储公司易安信（EMC），过去九年来一直担任软件公司威睿（VMware）总裁。在宣布就任后，他在给全体英特尔员工的电子邮件中回忆了往昔自己“在格罗夫（Grove）、罗伯特·诺伊斯(Robert Noyce)和戈登·摩尔(Gordon Moore)指导下”的光辉岁月(诺伊斯和摩尔是英特尔创始人)。基尔辛格先生和他们一样都是工程师，于1989年领导旗舰芯片的设计，这点他与上任总裁鲍勃·斯旺（Bob Swan）不同。

 

His first job will be to try to turn the firm’s ailing manufacturing division around. Intel already outsources the manufacturing of some lower-end chips to TSMC. Its production woes will force it, at least temporarily, to send more business to Taiwan, perhaps including some of its pricier desktop and graphics chips. Daniel Loeb, an activist investor with a sizeable stake in Intel, sent a letter to the firm’s management in December urging it to abandon factories entirely and restrict itself to designing chips that other firms, such as TSMC, would make. On paper, that looks attractive: Intel capital expenditure in 2020 amounted to $14.2bn, almost all of it on its chip factories. AMD, meanwhile, spun out its manufacturing business in 2009, and is thriving today. Nvidia has been “fabless” since its founding in 1993.

 

他将首先着手重振困境中的制造部。英特尔已将部分低端芯片的生产外包给了台积电。生产困境将迫使英特尔（至少暂时）将更多的业务转移到台湾，可能包括一些高档台式机和显卡芯片。拥有英特尔大量股权的激进投资者丹尼尔·勒布（Daniel Loeb）在12月致信该公司管理层，敦促其彻底放弃制造部门，把全部精力投入到芯片设计，芯片制造可让台积电等公司代工。理论上来看，这项提议很有吸引力：英特尔在2020年的资本支出为142亿美元，几乎全部用在了芯片制造上。同时，超微半导体在2009年剥离制造业务后公司大有起色。英伟达自1993年成立以来从未涉足过芯片制造。

 

Finding a buyer could be tricky, says Linley Gwennap, a veteran chip-industry watcher, precisely because Intel’s factories are now behind the cutting edge. Most of the world’s chipmakers, which might be tempted by the fabs, are in Asia. Since chips are a front in America’s tech war with China, politicians may veto a sale to a non-American bidder.

 

芯片行业资深观察员林利·葛文纳普（Linley Gwennap）表示，英特尔正因为工厂已落后于行业前沿水平，很难找到买家。对微型芯片生产厂感兴趣的芯片制造商大多在亚洲。由于芯片是中美科技战的前沿竞争领域，出售给非美国竞标者的交易很有可能遭到政客否决。

 

In any case, Mr Gelsinger has said he will ignore Mr Loeb’s suggestion. In January the new boss said that, although the firm may use more outsourcing for some products, he intends to pursue the hard, costly task of restoring Intel to its customary position at chipmaking’s leading edge. He also seems minded to pursue his predecessor’s strategy of diversifying into new products, including graphics-to-AI chips. “Our opportunity as a world-leading semiconductor manufacturer is greater than it’s ever been,” he wrote. The direction of travel, then, is not about to change. Intel’s shareholders will have to hope that Mr Gelsinger can at least get it back on the pace.

 

不管怎样，基辛格先生已经表态不打算接受勒布的建议。这位新任老板1月份表示，尽管公司会加大某些产品的外包力度，他仍然有意去啃这根又贵又硬的骨头——恢复英特尔在芯片制造行业的领先地位。他似乎打算遵循其前任的战略，开发更多新产品，包括显卡和人工智能芯片。他写道：“作为一家世界领先的半导体制造商，我们今天的机会更胜以往。”因而，英特尔的前进方向是不会改变的。股东们只能希望至少基辛格先生能让公司恢复以前的步调。

英文、中文版本下载：http://www.yingyushijie.com/shop/source/detail/id/2617.html