根據(jù)報導(dǎo)，中國的無晶圓廠(Fabless)公司數(shù)量至少有450多家，而現(xiàn)在的問題是，在幾年之后，有多少家能真正存活下來；更重要的是，他們將如何成長茁壯，在全球市場發(fā)揮影響力？ 在本刊最近進(jìn)行的一系列本土IC業(yè)者報導(dǎo)中，本土一些業(yè)界高階主管們指出，在假設(shè)這些無晶圓廠都繼續(xù)他們當(dāng)前業(yè)務(wù)模式的情況下，至少還要2~10年的時間，才能看出這些驟然崛起的企業(yè)們會消失或是走上正軌。 從根本上來看，大家對該產(chǎn)業(yè)未來前景的看法與猜測都有很大不同，但基本上大家似乎都同意，以目前中國的無晶圓廠模式來看，要長期維持下去似乎不大可能。中國企業(yè)不能僅仰賴今天的成本優(yōu)勢或勞動力優(yōu)勢來維持成長。然而，殘酷的事實是，在中國，成本優(yōu)勢已經(jīng)讓很多公司雇用了二位或是三倍的工程師。而這股趨勢也讓美國的工程師更加感到恐慌。 北京WestSummit資本管理公司的創(chuàng)始合伙人，董事總經(jīng)理陳大同指出，與美國公司相比，若一家中國公司有100個工程師，他們只需要美國工程師設(shè)計所需的三分之一時間，就能快速設(shè)計出產(chǎn)品。 芯原科技(VeriSilicon)總裁戴偉民同意這個觀點。 當(dāng)被問到“為什么選擇中國，”時，戴偉民的回答簡短有力：“美國太貴了。臺灣工程師品質(zhì)很好，但數(shù)量不足以滿足我們的需求；而印度則是還沒有準(zhǔn)備好?！倍跔I運成本部份，若美國是一的話，臺灣大約是三分之一，而中國大約是五分之一，戴偉民表示。至于工程品質(zhì)方面，戴偉民指出，年輕的中國工程師可能還很嫩，但我們可以給他們最好的設(shè)計流程和工具，以彌補(bǔ)他們?nèi)狈?jīng)驗這一點?！霸谛驹?，我們每周都tape-out一款芯片，”他說。盡管芯原在全球各地?fù)碛卸鄠€辦事處，但芯原在中國的員工人數(shù)共有380人，其中250人位在上海的總部。 然而，芯原的戴偉民也對中國無晶圓廠提出警告，若未能發(fā)展出獨特的商業(yè)模式，“未來兩年內(nèi)許多業(yè)務(wù)都將無以為繼”。芯原和eSilicon與OpenSilicon類似，都是快速成長的“IC設(shè)計代工” 供貨商，主要是為客戶提供硅解決方案和SoC的一站式服務(wù)。本質(zhì)上，芯原并不提供芯片業(yè)務(wù)。相反，它是服務(wù)型企業(yè)──利用自己的IP，或是來自其它第三方伙伴的IP為客戶設(shè)計芯片撇開商業(yè)模式問題不談，目前，現(xiàn)實問題在于，中國的工程師成本也日漸升高。 一些報導(dǎo)指出，中國的工程管理人員薪資并不低，以公司設(shè)在北京的工程部副總裁這類高階主管來看，他們的薪資都已經(jīng)在2007~2008年時達(dá)到高峰，而設(shè)計工程師的薪資則仍在上升。因此，為了維持利潤率，許多公司開始向中國西部出發(fā)，尋找更便宜的設(shè)計工程師。 ARM中國總裁吳雄昂認(rèn)為，以今天中國的無晶圓廠模式來看，最大的陷阱就在于有太多同構(gòu)型太高的本地公司，他們的產(chǎn)品競爭非常激烈。這種模式必然不會長久。而這也是新思科技(Synopsys)全球副總裁、亞太區(qū)總裁潘建岳所說的，中國無晶圓廠必須掌握創(chuàng)新的力量才能生存下去。 潘建岳指出，“如果你想在全球市場中插一腳，你就會需要全球性的人才和全球供應(yīng)鏈?！? 他們學(xué)得多快？ 但是，潘建岳并不認(rèn)為中國無晶圓公司必須從頭開始。他解釋道，中國企業(yè)應(yīng)該積極招聘歐洲和日本的工程技術(shù)人才。“看看三星如何從日本人那里學(xué)習(xí)半導(dǎo)體技術(shù)?！迸私ㄔ懒信e了一些韓國企業(yè)的做法──也許在二十年前，當(dāng)時韓國企業(yè)每個周末都邀請日本工程師到韓國去；韓國人會問一些問題，日本工程師回答后，他們便一起打高爾夫球，接著送日本工程師回家──他們的商業(yè)機(jī)秘就在“第19洞”。 當(dāng)然，中國無晶圓公司現(xiàn)正面臨著一個更大挑戰(zhàn)：缺乏經(jīng)驗豐富的管理技能。WestSummit的陳大同指出，“確實，很多中國無晶圓公司今天都是由“海歸派”在經(jīng)營，但這些公司中，有許多都缺乏 專業(yè)行銷知識，而且?guī)缀鯖]有任何管理經(jīng)驗。”他表示，問題在于“他們學(xué)習(xí)的速度能有多快？”陳大同本身是在美國拿到學(xué)位，并在硅工作一段時間后才回國； 他曾在1995年與伙伴共同創(chuàng)立Omnivision，同時也出任展訊(Spreadtrum)的首席技術(shù)官。 Synopsys 的潘建岳也有類似的看法。他指出，缺乏管理能力是一大問題。他指出，要經(jīng)營一家規(guī)模達(dá)十億美元的公司，你所需要的技巧會與經(jīng)營一家規(guī)模一億美元的公司截然 不同。他以一家由三人共同成立的家庭式無晶圓廠公司為例指出，你可能只需要埋頭工作，直到讓公司成長為20人甚至100人的中小型企業(yè)。然而，這種模式卻不能應(yīng)用在2,000人的公司中。潘建岳表示，中國的高階主管通常都是好學(xué)生，但問題是，他們學(xué)得有多快？這個學(xué)習(xí)過程可能會長達(dá)數(shù)年之久。 ARM的吳雄昂認(rèn)為，缺乏執(zhí)行經(jīng)驗的公司，有時也會展現(xiàn)在無法良好定義市場方面。一旦無法明確地定義市場，就容易導(dǎo)致更高的產(chǎn)品開發(fā)成本，或是讓他們錯失商機(jī)。 本文授權(quán)編譯自EE Times，版權(quán)所有，謝絕轉(zhuǎn)載 本文下一頁：中國靈敏的反應(yīng)才是重點！

相關(guān)閱讀：
• 高通選擇代工伙伴的糾結(jié)之心
• 無晶圓廠經(jīng)營模式在14nm的雞血下仍將堅挺
• 李力游是如何讓展訊起死回生的nCQesmc

{pagination} 并購潮 然而，當(dāng)前中國成功的無晶圓公司們，幾乎都有著共同的故事：藉由并購及收購來強(qiáng)化自身能力。 2007 年底，展訊收購Quorum Systems就是很好的例子。Quorum是一家位在圣地亞哥的無晶圓廠半導(dǎo)體公司，主業(yè)是設(shè)計高整合CMOS射頻收發(fā)器。去年，展訊還收購了 MobilePeak Holdings Ltd. 48.4%的股權(quán)，MobilePeak也是一家無晶圓廠芯片公司，在圣地亞哥和上海設(shè)有公司，主要產(chǎn)品是UMTS/HSPA+調(diào)制解調(diào)器芯片組。 在最近一次接受本刊的采訪過程中，展訊CEO李力游表示，他希望展訊能被認(rèn)為是一家“能提供美國技術(shù)和中國服務(wù)態(tài)度的公司”?！澳愠Ｈ徊粫Ｍ藗冋J(rèn)為你是一家提供中國技術(shù)和美國態(tài)度的公司，”李力游笑道。 在中國，購并風(fēng)潮正迅速蔓延。今年稍早，上海的銳迪科微電子(RDA Microelectronics)和Trident Microsystems簽署了一項IP授權(quán)和發(fā)展協(xié)議。 RDA將支付1,600萬美元現(xiàn)金，以獲得為期10年的非排他性、全球性、不可轉(zhuǎn)讓的授權(quán)，用于開發(fā)、生產(chǎn)和銷售 Trident SX5 數(shù)字電視SoC平臺的衍生版本。 RDA CEO戴保家在最近與本刊的訪談中表示，該公司之前并沒有進(jìn)軍DTV SoC市場的計畫。他解釋道，收購IP更像是一種防守動作。但他仍然表示，“即使我們在IPTV市場是后進(jìn)者，但我們?nèi)該碛幸粋€巨大優(yōu)勢。因為我們熟知本地市場?！? 在中國，大家都承認(rèn)，不僅工程成本上揚(yáng)，現(xiàn)在各種基礎(chǔ)建設(shè)的成本也不斷攀升。而當(dāng)被問到中國的無晶圓廠業(yè)務(wù)還能持續(xù)多久時，Synopsys的潘建岳看來有些猶豫，接著他表示，“五年。”然而，他立即改口，“或許還有10年?！? 中國靈敏的反應(yīng)才是重點！ 他的樂觀并非來自中國的成本競爭力，而是速度。他對中國能迅速對市場做出反應(yīng)的能力深具信心。隨著所有關(guān)鍵的組件供貨商和制造商紛紛在中國設(shè)立公司(更不用提在中國擁有OEM客戶的公司了)，中國當(dāng)?shù)卣谛纬梢粋€巨大的生態(tài)系統(tǒng)，這將讓中國的無晶圓廠業(yè)務(wù)模式更能發(fā)揮效率。在這里，沒有延遲、不用開冗長的電 話會議，可以立即做出決定。這個生態(tài)系統(tǒng)也能讓當(dāng)?shù)貜S商們從僅生產(chǎn)簡單的“me-too”產(chǎn)品，進(jìn)階成可提供加值型系統(tǒng)的廠商，甚至開始開發(fā)軟件。 另一個讓中國無晶圓廠樂觀的原因，是晨星(MStar)和聯(lián)發(fā)科(MediaTek)這兩家臺灣IC設(shè)計大廠在手機(jī)、數(shù)字電視領(lǐng)域的成功經(jīng)驗。這兩家大家所稱的“M brothers”是許多中國無晶圓廠半導(dǎo)體公司正在試圖模仿一個成功藍(lán)圖。晨星和聯(lián)發(fā)科已經(jīng)體認(rèn)到，為客戶提供完整軟硬件解決方案具有重大意義?！斑@兩 家公司一直維持著較低的成本結(jié)構(gòu)，而且已透過全球供應(yīng)鏈實現(xiàn)全球化經(jīng)營，”潘建岳表示。 編譯： Joy Teng 本文授權(quán)編譯自EE Times，版權(quán)所有，謝絕轉(zhuǎn)載 本文下一頁：參考英文原文： How long for China to sort out fabless strategy?，by Junko Yoshida

相關(guān)閱讀：
• 高通選擇代工伙伴的糾結(jié)之心
• 無晶圓廠經(jīng)營模式在14nm的雞血下仍將堅挺
• 李力游是如何讓展訊起死回生的nCQesmc

{pagination} How long for China to sort out fabless strategy? Junko Yoshida BEIJING – With reportedly 450-plus fabless chip companies in China, it’s fair to ask how many more years can they survive in this business and, more important, what they must do now to grow into a significant force in the global market. (Editor's note: How Chinese fabless companies have gotten to where they are today is examined in the previous article entitled "Is China's fabless model sustainable?") In a series of interviews recently conducted in China by EE Times, Chinese executives estimated another two to 10 years before these fabless upstarts either become irrelevant or go out of business—assuming that they keep doing what they’re doing right now. In essence, while this window of opportunity differs from guess to guess, everyone seems to agree that China’s current fabless model isn’t sustainable for the long haul. Chinese companies can’t keep growing simply by relying on today’s cost advantage, and working even harder without sleep. However, the hard truth is that the cost advantage in China has been driving many companies to double or the number of triple engineers they hire. In turn, this trend instills fear – warranted or unwarranted – among engineers whose jobs are still in the United States. As Datong Chen, managing director of Beijing’s WestSummit Capital Management, pointed out, if a company has 100 engineers in China, it can design a product at a sixth or an eighth the cost of the United States, and in one-third the time. Wayne Dai, company president and CEO at VeriSilicon, echoes this point. When asked “why China,” Dai simply said: “The United State is too expensive. Taiwan is good but doesn’t have enough of the engineers [we want], and India isn’t ready.” As far as the cost of doing business is concerned, if the U.S. is one, Taiwan is about one-third, and China comes out to be one-fifth on average, Dai said. Asked about the quality of engineers, Dai said that young Chinese engineers may be green, but given the best design flows and tools, their energy compensates for their inexperience. “We tape out one chip per week at VeriSilicon,” he added. While the company has multiple offices throughout the world, VeriSilicon has a total of 380 people working in China – 250 based in Shanghai. But it was VeriSilicon’s Dai who sounded the alarm that Chinese fabless companies “will go out of business within the next two years” without a unique business model. VeriSilicon, similar to eSilicon and OpenSilicon, is a fast-growing “IC design foundry” which offers its customers silicon solutions and SoC turnkey services. VeriSilicon, in essence, is not in the business of making their own chips. Rather, it’s in the service business – designing chips for clients, leveraging VeriSilicon’s own IP’s and those from other sources. How long for China to sort out fabless strategy? Setting aside the business-model issue, the reality is that the cost of engineers has also been rising in China. Some report that high-end salaries for Chinese engineering managers – such as a VP of engineering based in Beijing – might have already peaked in 2007 – 2008. But the price of design engineers is still going up. Hence, companies looking to keep their gross margin low are going westward in China in search of cheaper design engineers. The biggest trap Allen Wu, president of ARM China sees in China’s fabless model today is that there are just too many local companies who compete purely on cost with similar products. That isn’t sustainable. And it’s why Jian-Yue Pan, corporate vice president, Asia Pacific region at Synopysis, is convinced that the survival of fabless China hinges on companies’ power to innovate. “You need to move up the innovation ladder.” Pan said, “If you want to play in the global market, you need global talents and global supply chain.” How fast can they learn? Pan, however, isn’t suggesting that Chinese fabless companies need to invent everything from scratch. Chinese management shouldn’t shy away from actively hiring engineering talent from Japan and Europe, he explained. “Look at how Samsung learned semiconductor expertise from Japanese engineers.” Pan is referring to the somewhat infamous practice – maybe a couple of decades ago – when Korean companies invited Japanese engineers to Korea over weekends; Koreans ask a few questions, Japanese engineers answer, they play golf together, and the Japanese engineers go home — leaving their trade secrets at the “19th hole.” Of course, this poses an even bigger challenge Chinese fabless companies face: a lack of experienced management skills. WestSummit’s Chen said, “Yes, a lot of Chinese fabless companies today are run by so-called ‘returnees.’ But many of them have little marketing expertise and practically no management experience.” Chen said that the question is: “How fast can they learn?” Chen himself returned to China after getting his higher education in the United States and worked in Silicon Valley; he co-founded Omnivision in 1995, and he was also Spreadtrum’s CTO. Synopsys’ Pan has similar views. Lack of management capabilities is a big issue, he noted. “To run a $1 billion company, you need a different skill set from the one you used in running a $100 million company.” Take the example of a fabless company, say, founded by three members of the same family. “You may be able to grow that business with no process and no system in place – simply by working very hard, to a 20-people company or even a 100-people company,” he said. But that won’t work with 2,000 people. Pan said, “Chinese executives are good learners. But the question is how fast can they learn? It could take years.” ARM’s Wu sees that this lack of executive experience can sometimes manifest itself in misdefinition of a market. The inability to clearly define a particular segment of the market to go after with new products could cost the business dearly, or result in a blown opportunity. M&A waves However, some of the success stories emerging from the current generation of China fabless companies have a common thread: an ability to pull off mergers and acquisitions. One good example is Spreadtrum’s late 2007 acquisition of Quorum Systems, a San Diego-based fabless semiconductor company that specializes in the design of highly integrated CMOS radio frequency transceivers. Last year, Spreadtrum also acquired 48.4% of MobilePeak Holdings Ltd., a fabless chip company that designs UMTS/HSPA+ modem chipsets from bases in San Diego, California and Shanghai, China. As Spreadtrum’s CEO Leo Li proudly put it in a recent interview with EE Times, Spreadtrum wants to be known as a company that offers “U.S. technology with a China attitude.” Li chuckled, “You certainly don’t want to be known as someone to offer ‘Chinese technology with a U.S. attitude.” Kidding aside, the M&A waves are spreading fast. Earlier this year, RDA Microelectronics in Shanghai (a fabless semiconductor company that designs, develops and markets Radio Frequency and mixed-signal semiconductors for cellular, connectivity and broadcast applications) signed an IP license and development agreement with Trident Microsystems. RDA will pay US$16 million in cash to secure a non-exclusive, worldwide, non-transferrable license to develop, manufacture and sell derivative versions of the Trident SX5 Digital TV SoC platform for a period of 10 years. Vincent Tai, RDA’s CEO, in a recent interview with EE Times, said the company’s entry in to the DTV SoC market isn’t planned until later. Its IP acquisition was more a defensive move, he explained. But he remained confident: “Even if we are a latecomer to the IPTV market, I know I have a huge advantage. Being a local, we know the local market.” There’s no denying the rise of not only engineering costs but also the cost of infrastructure in China. Asked how long China’s fabless companies can sustain their business, Synopsys’ Pan, with a little hesitation, first said, “Five years.” But then he immediately corrected himself, “Perhaps another 10 years.” It's China's speed, stupid! His optimism comes not from China’s cost-competitiveness, but from speed. He has faith in China’s ability to respond to the market quickly. With all the key component suppliers and the manufacturing base on hand in China (not to mention OEM customers in China), a vast, locally available eco-system lets Chinese fabless companies do business much more effectively. No time lag, no lengthy conference calls and no delayed decision making. It also allows them to expand from simply rolling out me-too products to offering value-added systems or even developing software. Another source of optimism for fabless China comes from the recent success of MStar and MediaTek – two Taiwanese giants who took the global semiconductor market by storm for mobiles and digital TVs. The so-called “M brothers” drew a blueprint for success that many Chinese fabless companies are trying to emulate. MStar and MediaTek have figured out the significant importance in providing their customers with turnkey software and hardware solutions. “The two companies kept the cost structure relatively low and yet totally globalized theiroperations, by using global supply chains,” observed Pan.