China is entering memory chips, the bloodiest segment in the technology industry. This could get interesting.


The 'memory wars' of the 1980s were viewed at the time as a strategic victory by Japan over America. Tokyo’s MITI-backed electronics giants muscled Intel out of the market for superfast DRAM. Intel re-dedicated itself to microprocessors, while the Japanese and a dozen others fought the most ruinous competition in the history of technology. Today, Intel reigns in processors and, by the end of a 20-year shakeout in DRAM, only three of the seventeen starters are still standing (Korea's Samsung and Hynix and America's Micron). Taiwanese companies alone invested US$50 billion in DRAM over the 1990s and 2000s and ended up nowhere. The other major type of memory NAND (or 'flash', which is slower but stores data in power-off mode) underwent a similarly bruising rationalisation. Today only Japan's Toshiba, its US partner SanDisk, Samsung, and Intel-Micron's venture are competitive.

Memory chips are a winners-take-all business; the losers quickly burn through shareholder funds and then (usually) government subsidies. A single 'fab' factory costs billions of dollars to build, billions to run and a years-long learning curve. The tyranny of Moore’s Law dooms laggards and keeps even market leaders gasping for oxygen. It was thought inconceivable that an entirely new entrant would hope to challenge. But now China, that had largely avoided the memory industry, has announced a daring series of forays into the market.

Beijing’s planners have big ambitions in the semiconductor market, which they see as a pathway to an advanced information economy. They lament that China hasn’t one company in the global Top 20 and evidently feel shame that 'China imports more chips than oil'. Now they have a one trillion yuan (US$160 billion) fund to spend on semiconductors over a decade. Their budget will be allocated between domestic and foreign spending because they realize China needs outside technologies.

Foreign companies are intrigued, concerned and excited by China’s grand memory ambitions. Beijing is splashing money overseas acquiring targets such as ISSI and Powertech. It has encouraged the multinationals that lead the market to build memory fabs in China, hoping to transfer their expertise to the local workforce. Samsung and Hynix have already obliged.

Outside observers are puzzled by a number of things.

China has made solid progress in designing and producing logic chips, so why prioritise the notoriously difficult memory industry? Possibly there are national security reasons for its obsession with homemade memory chips. Why has the mysterious newcomer Unigroup — led by a flamboyant billionaire chairman famous for elaborate financial engineering — been tasked to invest $50 billion in the next five years? How will it spend the $13 billion now being raised? Can it integrate its disparate assets? Unigroup has successfully rolled up a number of designers, notably Spreadtrum and RDA, and staked the hard-drive maker Western Digital to buy SanDisk, but it’s unclear whether these American companies can or will transfer critical memory technology to China. Some of Unigroup’s 'shopping', like a rumored bid for a slice of Micron, looks hopeless. An executive there reckons China would have to spend $70 billion to build an independent memory capability.

But Unigroup keeps knocking. And foreign technology companies are desperate to access the Chinese market and eager to satisfy Beijing’s demands to localise. An affiliate of Unigroup has enticed its new Western Digital/SanDisk partner to make and sell domestic memory solutions on the mainland. Intel bought a 20% interest in Unigroup and surprisingly has announced a new memory fab in China. Notably Micron, Intel’s partner, hasn’t joined.

The dilemma for Taiwan and Korea is painfully apparent even as they too rebuff China’s approaches. They have seen the technology-for-access movie before in the LCD sector: first Chinese import tariffs are raised, then a 'price-fixing' probe, then one foreign company buckles and agrees to share its crown jewels with Beijing, and the rest quickly fold and follow. As in LED, solar and flat TVs, once Chinese firms are established, they proliferate and grow aggressively. Margins and returns implode. Chinese semiconductor executives, supposedly motivated by Mao 'to serve the people', beam insouciance about the financial rationality of their adventure.

Underlying this complicated situation are the three biggest and baddest trends in technology. First, cheap money has fuelled a 'frenzy' of tech industry M&A, especially in semiconductors. Second, China wants influence, its pockets are deep and it’s very edgy about data security. Third, in a tremendous architectural shift, big data is changing the way the world is storing information. Just as DRAM enabled PCs to replace tape-playing mainframes in the 1980s and today NAND silicon is killing off hard-drives, all our information is moving 'into the cloud'. Memory is becoming more strategic than ever.

Predictably, governments are paying attention. Washington has started to thwart China's semiconductor ambition. American analysts are urging it to foil Unigroup’s Micron bid. The grim language Chinese leaders themselves use — 'to break free from external dominance' — reflects a determination to control this business. They are risking tens and probably hundreds of billions of dollars, to prevail where Japan failed.

Photo courtesy of Flickr user Kevin Stanchfield