數字生產流水線

項利

Technology firms vie for1 billions in corporate data-analytics contracts.技術公司爭奪價值數十億美元的企業數據分析合同。

Somebody less driven than Tom Siebel would have long since thrown in the towel. In 2006 the entrepreneur， then 53 years old， sold his ?rst ?rm， Siebel Systems， which made computer programs to track customer relations， to Oracle， a giant of business software. That left him a billionaire—but a restless one. In 2009， a few months after Mr. Siebel had launched a new startup， he was trampled2 by an elephant while on safari in Tanzania. When， a dozen surgeries later， he could work again， the enterprise almost went bankrupt. Undeterred3， he rebooted it.

Mr. Siebels fortitude has paid off4. The ?rm， now called C3.ai， raised $100m in venture capital last year， valuing it at $2.1bn. It was an early bet on data analytics， which converts raw data （from a machines sensors or a warehouse） into useful predictions （when equipment will fail or what the optimal5 stocking levels are） with the help of clever algorithms. Many investors see fortunes to be made from this new breed of enterprise software， which is spreading from Big Techs computer labs to corporations everywhere.

Worldwide， 35 companies that dabble6 in data analytics feature on a list of startups valued at $1bn or more， maintained by CB Insights， a research ?rm. Collectively， these unicorns7—some of which brand themselves as purveyors8 of arti?cial intelligence （AI）—enjoy a heady valuation of $73bn. According to PitchBook， another research company， the six biggest alone are worth $45bn . Many venture capitalists who back them are hoping to emulate9 the successful initial public offerings of less exalted10 business-services startups like CrowdStrike， which provides cybersecurity or Zoom， a video-conferencing company. And then some.

As is often the case in Silicon Valley， hype11 springs eternal， fuelled by big numbers from consultancies. IDC12 reckons that spending on big-data and business-analytics software will reach $67bn this year. But it will， boosters say， at last allow businesses to see the computer age in their productivity statistics， freeing them from the shadow of Robert Solow， a Nobel Prize-winning economist， who in 1987 observed that investment in information technology appeared to do little to make companies more efficient. Just as electricity enabled the assembly line in the 19th century， since machines no longer had to be grouped around a central steam engine， data-analytics companies promise to usher in13 the assembly lines of the digital economy， distributing data-crunching capacity where it is needed. They may also， as George Gilbert， a veteran business-IT analyst， observes， help all kinds of ?rms create the same network effects behind the rise of the tech giants： the better they serve their customers， the more data they collect， which in turn improves their services， and so on.

Consultants at Gartner calculated that in 2021 “AI augmentation” will create $2.9trn of “business value” and save 6.2bn man-hours globally. A survey by McKinsey last year estimated that AI analytics could add around $13trn， or 16%， to annual global GDP by 2030. Retail and logistics stand to gain most.

Data analytics have a long way to go before they live up to these expectations. Extracting and analysing data from countless sources and connected devices—the “Internet of Things”—is difficult and costly. Although most ?rms boast of having conjured up14 AI “platforms”， few of these meet the usual de?nition of that term， typically reserved for things like Apples and Googles smartphone operating systems， which allow developers to build compatible apps easily.

An AI platform would automatically translate raw data into an algorithm-friendly format and offer a set of software-design tools that even people with limited coding skills could use. Many companies， including Palantir， the biggest unicorn in the data-analytics herd， sell high-end customised services—equivalent to building an operating system from scratch for every client. Cloud-computing giants such as Amazon Web Services， Microsoft Azure and Google Cloud offer standardised products for their corporate customers but， as Jim Hare of Gartner explains， these are considerably less sophisticated and lock users into their networks.

The enterprising Mr. Siebel

Enter C3.ai， founded to help utilities manage electric grids， a complex problem that involves collecting and processing data from many sources. After its near-bankruptcy， advances in machine learning， sensors and data connectivity gave it a new lease of life—and allowed it to repackage its products for a range of industries. Crucially for corporate clients， C3s approach grew out of Mr. Siebels experience with enterprise software. He wanted to make data analytics hassle-free15 for corporate clients， without sacri?cing sophistication.

3M， an American conglomerate16， employs C3 software to pick out potentially contentious17 invoices to pre-empt complaints. The United States Air Force uses it to work out which parts of an aircraft are likely to fail soon. C3 is helping Baker Hughes to develop analytics tools for the oil-and-gas industry （General Electric， the oil-services ?rms parent company， has struggled to perfect an analytics platform of its own， called Predix）.

C3s chief rival in building a bona ?de18 AI platform is not Big Tech or the very biggest data-analytics unicorns. It is a company called Databricks. It was founded in 2013 by computer wizards19 who developed Apache Spark， an open-source program which can handle reams of data from sensors and other connected devices in real time. Databricks expanded Spark to handle more data types. It sells its services chie?y to startups （such as Hotels.com， a travel site） and media companies （Viacom）. It says it will generate $200m in revenue this year and was valued at $2.8bn when it last raised capital in February.

Though C3s and Databricks niches do not overlap much at the moment， they may do in the future. Their approaches differ， too， re?ecting their roots. Databricks， born of abstruse20 computer science， helps clients deploy open-source tools effectively. Like most enterprise-software ?rms， C3 sells proprietary21 applications.

It is unclear which one will prevail; at the moment the two ?rms are neck-and-neck22. In the near term， the market is big enough for both—and more. In the longer run， someone will come up with AI-assisted data analytics that are no more taxing23 than using a spreadsheet. It could be C3 or Databricks， or smaller rivals like Dataiku from New York or Domino Data Lab in San Francisco， which are also busily erecting AI platforms. The ?elds other unicorns are unlikely to give up trying. And incumbent tech titans24 like Amazon， Google and Microsoft want to dominate all sorts of software， including advanced data analytics.

Mr. Siebel would be the ?rst to admit that this scramble25 is likely to claim victims. But it certainly bodes well for buyers of data-analytics software， which is likely to become as familiar to corporate IT departments in the 2020s as customer-relations programs are today.

如果沒有十足的干勁，湯姆·西貝爾可能早就在失敗中放棄了。2006年，這位時年53歲的企業家將他的第一家公司西貝爾系統公司（制作跟蹤客戶關系的電腦程序）出售給了商業軟件巨頭甲骨文公司。這讓他成為億萬富翁，但他并沒有安于現狀。2009年，創辦新公司幾個月后，西貝爾在坦桑尼亞游獵時被一頭大象踩傷。經歷了十幾次手術后，他終能重返工作，當時公司已瀕臨破產。他沒有氣餒，而是重新啟動了公司。

西貝爾先生的堅韌得到了回報。這家現名C3.ai的公司2018年籌集了1億美元的風險投資，估值21億美元。這是提早押寶數據分析，它借助巧妙的算法，將來自機器傳感器或倉庫的原始數據進行轉化，來有效預測設備故障發生的時間或最佳庫存水平等。許多投資者都看到了這種新型企業軟件帶來的機遇，這種軟件正從大型科技公司的計算機實驗室推廣到世界各地的公司。

市場數據研究公司CB Insights提供的資料顯示，全球有35家涉足數據分析的公司被列入估值至少10億美元的初創企業名單。總體而言，這些獨角獸企業（部分自稱為人工智能供應商）總估值高達730億美元。另一家研究公司PitchBook的數據顯示，僅6家最大的初創企業總估值就高達450億美元。許多支持他們的風險投資家都希望能夠效仿那些不知名但首次公開募股就取得成功的商業服務初創企業，比如提供網絡安全的CrowdStrike公司、提供視頻會議服務的Zoom公司，等等。

正如硅谷那樣，在咨詢公司的大數據推動下，炒作層出不窮。互聯網數據中心評估顯示，2019年用于大數據和商業分析軟件的支出將達到670億美元。但支持者表示，這最終會讓企業在生產率統計數據中看到計算機時代的契機，讓它們擺脫諾貝爾獎得主、經濟學家羅伯特·索洛帶來的陰影——1987年，索洛曾提出，信息技術投資似乎對提高企業效率沒有什么幫助。正如19世紀電力的出現讓生產流水線得以實現，機器的運作不再需要依賴中央蒸汽機，今天數據分析公司有望引入數字經濟的生產流水線，按需分配數據處理能力。據資深商業信息技術分析師喬治·吉爾伯特分析，它們還可以幫助各類企業在技術巨頭崛起的背景下創造出同樣的網絡效應：為客戶提供的服務越好，收集的數據就越多，這轉而又可以提高服務質量，以此類推。

高德納咨詢公司的咨詢顧問估測，2021年，“人工智能增強”將在全球范圍內創造2.9萬億美元的“商業價值”，并節省62億工時。麥肯錫咨詢公司2018年的一項調查顯示，到2030年，人工智能分析可能使全球年度國內生產總值增加約13萬億美元，即提高約16%。零售業和物流業很可能獲益最大。

在達到這些期望前，數據分析行業還有相當長的路要走。從包含了數量龐大的信息源和聯網設備的“物聯網”中提取和分析數據，難度大且成本高。盡管大多數公司都宣稱自己構想搭建人工智能“平臺”，但沒有幾個是真正意義上的人工智能平臺，像蘋果和谷歌的智能手機操作系統那樣，允許開發者輕松構建兼容的應用程序。

人工智能平臺會自動將原始數據轉換成一種算法友好的格式，并提供一套軟件設計工具，就算是編程能力有限的人也會使用。許多公司，包括數據分析領域最大的獨角獸帕蘭提爾公司，都在銷售高端定制服務——為每個客戶從零開始構建操作系統。亞馬遜網絡服務、微軟云和谷歌云等云計算服務巨頭都為企業客戶提供標準化產品，但正如高德納全球研究副總裁吉姆·黑爾所說，這些產品操作起來非常簡單，將客戶鎖定在自己的網絡中。

勇于進取的西貝爾先生

成立C3.ai公司是為了幫助公共事業公司管理電網，這是一個涉及從多個來源收集和處理數據的復雜問題。公司近乎破產后，機器學習、傳感器和數據連接方面的進步令公司的生命得以延續——使其能夠為許多行業提供重新包裝的產品。對企業客戶來說，至關重要的是，C3的發展途徑源于西貝爾先生在企業軟件方面積累的經驗。他希望在保留復雜性的同時，為企業客戶提供省心省力的數據分析。

美國3M企業集團運用C3軟件挑出可能存在爭議的發票，從而預先制止投訴。美國空軍用C3軟件來測算可能很快會出現故障的飛機部件。C3正幫助石油服務公司貝克休斯開發石油和天然氣行業的分析工具（貝克休斯的母公司通用電氣公司一直費盡心思要完善自己的分析平臺Predix）。

在打造真正的人工智能平臺方面，C3公司的主要競爭對手不是科技巨頭，也不是那些超大的數據分析獨角獸公司，而是一家名為Databricks的公司。它是由開發Apache Spark的電腦奇才們于2013年創建的。Apache Spark是一個開源程序，可以實時處理來自傳感器和其他聯網設備的海量數據。Databricks擴展了Spark性能以處理更多數據類型。它主要服務于初創公司（如旅游網站Hotels.com好訂網）和媒體公司（如Viacom維亞康姆）。該公司表示，2019年公司收入將達2億美元，上一次融資就在2019年2月，當時估值高達28億美元。

盡管目前C3公司和Databricks公司的市場定位不太相同，但將來可能會有重疊。兩家公司發展途徑也不同，反映了各自不同的發展根基。Databricks公司源自深奧的計算機科學，致力于幫助客戶有效利用開源工具;C3公司則像大部分企業軟件公司一樣，售賣專利應用程序。

兩家公司哪家會占上風尚不明朗，目前它們勢均力敵。短期內，市場夠大，足以容納這兩家甚至更多公司。長遠看，會有公司提出人工智能輔助的數據分析方法，那不會比使用電子表格更難。有可能是C3公司或Databricks公司，也可能是規模較小的競爭對手，如紐約的Dataiku公司或舊金山的Domino數據實驗室，它們也都在忙著構建人工智能平臺。該領域的其他獨角獸公司不太可能放棄嘗試。亞馬遜、谷歌和微軟等老牌科技巨頭都希望主導包括高級數據分析在內的各類軟件。

西貝爾先生將會第一個承認，這場競爭可能會造成傷害。但對于數據分析軟件的買家來說，這無疑是個好兆頭——在2020年代，公司的信息技術部門很可能會像今天熟悉客戶關系程序一樣熟悉數據分析軟件。

（譯者單位：廣東第二師范學院）