Gordon Moore
戈登·摩尔
Chips in everything
万物皆有芯
Medieval philosophers once wondered how many angels could dance on the head of a pin. Gordon Moore, who with his liking for pad, pencil and a quiet, solitary desk often looked philosophical, wondered how many transistors could be etched on one silicon wafer to make an integrated circuit. The first transistors he had ever encountered, in 1954 at a lecture by the man who had co-invented them, William Shockley, were as big as peanuts. But they were shrinking fast. The more you could pack together, the more these tiny marvels could do, until they could probably change the world.
中世纪的哲学家们想知道一个大头针帽上究竟能够容下多少天使一齐舞蹈，而戈登·摩尔却想知道一块硅片上究竟可以集成多少根晶体管来组成集成电路。他喜欢独自坐在办公桌前，像个哲学家似的手持板子和笔、沉思问题。1954年，摩尔在晶体管发明者之一威廉·肖克利（William Shockley）的讲座上第一次见到晶体管。那时的晶体管还大如花生，但随后它们快速升级，越变越小。在一块硅片上集成的晶体管越多，这些神奇的小玩意能带来的改变就越大，直至有望改变世界。
In 1965, in Electronics magazine, he laid out his thinking plainly. In the past decade, the number of components in each integrated circuit had roughly doubled every year. In the next decade, therefore, they would probably do the same. Devices containing such circuits would become more efficient, more popular and less expensive, all at an exponential rate. They would power “such wonders” as home computers, automatic controls for cars, and “personal portable communications equipment”. Even wristwatches could have them. And though he went no further into his crystal ball than 1975, he saw no reason why this extraordinary growth should not go on for a long, long time.
1965年，摩尔在《电子学》杂志上发表了一篇文章，清晰地阐述了自己的想法。过去十年间，一个集成电路能够包含的元器件数量每年都几乎翻倍，因此这一趋势很可能在未来十年持续下去。含有集成电路的设备运行更高效、价格更实惠、大众也更青睐，而这一切均将以指数级的速度推进。这些集成电路还会助力“伟大奇迹”的诞生，例如家用计算机、汽车自动控制以及“个人便携式通信设备”，就连手表都能采用集成电路。摩尔的展望虽然只到1975年，但他坚信每个集成电路所能包含的元器件数量将继续在相当长的一段时间内以惊人的速度不断增加。
The article caused a sensation. What seemed to him just blind extrapolation was seized on as a Great Truth, “Moore's law”. From that time on, semiconductor companies took it as a given that the processing power of their product ought to double every year. If it didn't, they would lose their edge. In 1975 Dr Moore reviewed it, and thought the rate of doubling should be every two years—even though, in the preceding decade, it had in fact doubled in nine of the ten years. This refinement hardly made a difference. The points the industry took away from him were, first, to make semiconductors as tiny as possible, and second, to do it fast. Or fall behind, probably never to recover.
这篇文章轰动一时。在摩尔看来，他的看法不过是显而易见的推断，而人们却对此大为关注，称之为一条“伟大的真理”——即“摩尔定律”。自此，半导体公司开始理所当然地认为自家产品的处理能力必须每年翻一番，要不然自己就将失去竞争优势。1975年，摩尔重新回顾了这一定律，并认为翻倍的速度应改为每两年翻一番（尽管过去十年当中有九年都实现了翻番）。不过，摩尔对定律的完善并没有对实际情况产生什么影响。半导体行业从摩尔的结论当中学到的要领是：第一，要制造出尽可能小的半导体；第二，完成第一步的速度要快。要是做不到这两点，企业就会落后，很可能再也无法赶超。
He was surprised to have given his name to any law, and more so to be called a revolutionary. The silicon revolution was a fact, but he himself was a rooted, reserved sort of fellow who had hardly ever lived outside the foggy country south of San Francisco, and whose greatest love, beyond his work and his wife, was fishing. The closest he had come to revolution was when, as a boy, he made explosives with a home chemistry set, and somewhere in his well-stuffed bookshelves he still had his acid-scorched primer on the nature of nitroglycerine. But when it came to forging the digital age he was simply, he thought, in the right place at the right time. He got into semiconductors at the very start, and that was a great piece of luck.
摩尔从未想过他的名字会和任何定律联系在一起，更没有想过这一定律会被称为一场革命。硅谷革命是一件事实，但摩尔自己却是一个扎根乡土、行为内敛的人，基本上从未离开过旧金山南部多雾的农村，除了工作和妻子之外最大的爱好便是钓鱼。他做过的最接近“革命”的事是小时候用家用化学工具箱自制*药炸**，在他满满当当的书架上还塞着一本已经被酸腐蚀的硝酸甘油初级读本。就打造数字时代这件事，他认为自己只是在对的时间做了对的事。他一开始就研究半导体，这基本上就是运气。
He also drove the transformation, though, and not just with his law. At Shockley Laboratories, where he went in 1956, his job was fundamental: to create a cheap transistor based on silicon, of which the world had plenty. At Fairchild Semiconductor, founded by “the traitorous eight” who broke from Shockley in 1957, he developed a smooth silicon-dioxide surface on which to print the electric circuits and pioneered the use of aluminium wires to connect the transistors. When he founded Intel with Robert Noyce in 1968 he worked on self-isolating transistors, which could be packed more closely, as well as devising a better dip to clean the silicon surface before the aluminium went on. This single change increased chip production more than tenfold.
不过，他推动数字革命的方式不仅仅有“摩尔定律”。1956年，他加入肖克利实验室从事基础工作：使用硅这种全世界储量庞大的材料来制作价格低廉的晶体管。在飞兆半导体（由1957年离开肖克利实验室的“八叛逆”创建的公司），他开发了一种可用于打印电路的光滑二氧化硅表面，并率先使用铝线来连接晶体管。1968年，和罗伯特·诺伊斯共同创办英特尔之后，他开始研究排布可以更紧密的自我绝缘晶体管，同时还发明了一种在加入铝之前清洁硅表面的液体。仅这一项发明便使芯片产量提高了十倍以上。
He called himself an accidental entrepreneur, but there was not much accident about it. He left Shockley mostly because the man had no interest in getting the product to market. At Fairchild, in which each founder had invested $500, they were making chips commercially in less than a year, and had overtaken both Texas Instruments and Bell Labs in military applications. But Fairchild became too slow for him. His business plan for Intel, vague as it looked—to make interesting things with silicon—was spurred by his burning belief in chips in everything. By the 1990s, Intel microprocessors were in 80% of all the computers made in the world.
他称自己是阴错阳差才成了企业家，但是（他的成功）其实并非偶然。他之所以离开肖克利实验室，主要原因是因为肖克利对于产品上市兴趣索然。然而在创办人人均投资500美元的飞兆，这些人不到一年就开始生产商用芯片，并且在军事应用上已经超越德州仪器和贝尔实验室。但飞兆对于他来说还是太慢了。他给英特尔制定的商业计划——用硅制造有趣的东西——尽管在当时看来并不清晰，但其实是基于他对芯片能应用于万物的炙热信心说出来的。截至上世纪90年代，全世界生产的电脑当中有80%都在使用英特尔的微处理器。
Moore's law stated that as microchips invaded appliances, sales would soar and prices would fall; but setting up to make them was costly. He therefore strived to save money. He fitted out Fairchild's first premises with cheap kitchen units and blew his own glass tubes to manipulate gases, as he had done for Shockley. At Intel he watched every cent, down to the rubber gloves which he found for $1 a pair, rather than $2.50. This was habit. Until 1961 he had kept a ledger of all household incomings and outgoings, including a nickel found in his wife Betty's pocket and a dime spent on a pencil “(red)“. After 1961, he no longer needed to; his monthly salary, as charted in the ledger, was rising as steeply as the number of transistors on each chip. By 2014 his net worth was $7bn.
摩尔定律指出，随着微芯片在家用电器当中的广泛应用，微芯片的销量会飙升，而单价会随之下降；但是，制造微芯片的成本却很高。因此，他要尽量节省成本。飞兆最初的办公场所就安装了一套廉价的厨房设备，而导气用的玻璃管也都是他像当年为肖克利吹玻璃那样自己吹出来的。在英特尔，他紧盯每一分钱的使用状况，就连橡胶手套也要用1美元一双的，而不用2.5美元一双的。这是一种习惯。直到1961年，他一直保留着记录家庭所有收支的习惯，包括在妻子贝蒂口袋里发现的五美分硬币和花在一支“红色”铅笔上面的十美分。1961年以后，他不再需要这样做了；如财报上显示，他的月薪如同每个芯片上的晶体管数量一样飙升。到2014年，他的净资产达到了70亿美元。
Typically, though, his wealth did not change him. He liked to wear well-worn khakis and go out in his rickety old fishing boat, until it became unsafe. He liked things as they were. When he became rich he and Betty gave more than half of it away. They were motivated by seeing their fishing haunts in Baja California smothered by development, and their precious wild places disappearing. He did not think small. His foundation's grants went not just to CalTech and the Bay Area but were also earmarked to save two-thirds of the Amazon basin and the whole arc of salmon rivers that curved from northern California, through Alaska, to Russia's east.
不过不出所料，他的财富并没有改变他。他依然喜欢穿那条磨破的卡其裤，开着那艘快散架的旧渔船出去钓鱼，一直开到那艘船不安全到没法开为止。他喜欢事物维持它们本来的样子。当他变得富有以后，他和妻子贝蒂捐出了一半以上的财产。之所以这么做，是因为他们看到自己经常到下加利福尼亚钓鱼的那片地方因经济发展而面目全非，而他们珍视的荒野也在一点点消失。他的格局很大。他的基金会不仅拨款给加州理工学院和旧金山湾区，还定向拨付用于拯救三分之二的亚马逊盆地和整个起于北加州、横穿阿拉斯加、最后延伸至俄罗斯东部的弧形鲑鱼洄游带。
He hoped that might be his permanent legacy. Instead his legacy, inevitably, was Moore's law, which could not last for ever. He had never said it would; exponential growth always burned itself out. By 2020 its end was predicted within a decade in America. The Taiwanese were doing well but, even so, transistors could not endlessly be made smaller. In 2021 one was achieved that measured no more than a nanometre, a billionth of a metre, almost the size of an atom. Already billions of transistors could be squeezed on one silicon chip. Naturally, he accepted the laws of physics. As a chemist, though, he could not help hoping that some new material, some new process, might yet make room for more.
他希望这些公益事业可以成为自己留给世间的永恒遗产。但摩尔定律终归是他的遗产，即便这条“定律”不会永远地持续下去。他也从来没有说过会；指数增长亦会山穷水尽。2020年，有人预计10年内这条定律将在美国失效。中国台湾在芯片事业上成绩斐然，但即便如此，晶体管也不可能无限变小。根据2021年的测量结果，晶体管的大小已经可以做到不超过一纳米（也就是十亿分之一米），几乎和一个原子差不多大了；而且目前已经可以做到在一个微芯片上装下数十亿个晶体管。理所当然地，他能接受物理有它的客观极限。然而，作为一名化学家，他不禁希望出现某种新材料、或者某项新工艺，也许还能突破极限。