Sales of dynamic random access memory — the chips that store information in personal computers — have plunged 20 percent in the last year, prompting calls for government-led bailouts in Taiwan, consolidation efforts, and doubts about the ability of Micron Technology, the last DRAM maker in the U.S., to function without raising more cash. But bailouts and throwing more cash at DRAM players isn’t going to work. The only viable solution is to reduce the number of industry players.
DRAM prices have sunk to a historic low; manufacturers currently sell their chips for a third of their productions costs — essentially reaping 33 cents on every dollar spent. Prices are expected to continue to fall, to 74 cents per unit in 2010 and 28 cents in 2012 from $2.95 in 2007, according to estimates from iSuppli.
So far, Qimonda, the sole European DRAM player, has raised 325 million euros ($450 million) from its parent company and governments, and has sold off part of its operations to Micron in order to stay alive. Taiwan, where DRAM companies employ thousands, is another high-cost market, and may spend up to $6.5 billion on bailouts for PowerMOS and Powerchip, while also encouraging consolidation. Japan’s Elpida is rumored to be buying both firms, although several other suitors are said to be interested solely in PowerMOS.
Taiwan’s second-largest DRAM supplier, Nanya, which has a joint venture with Micron, is hoping for Taiwanese governmental assistance as well. Korea’s Hynix is also suffering, and last week scored a 800 billion Korean won ($600 million) loan from its creditors.
Instead of handouts, suppliers need to figure out how to make money when their wares sell for so little. Adding to the problem is that one DRAM chip is pretty much like another — meaning there’s no reason to pay more for DRAM from PowerMOS vs. Quimonda. The answer involves stabilizing prices by matching supply and demand, and producing chips as cheaply as possible. Consolidation and voluntary reductions in production should ease overcapacity, and reduce some pricing pressure, but demand needs to be goosed as well.
Demand for DRAM is expected to stay flat, in part because PC sales continue to decline (Gartner expects them to drop by 5 percent in 2009), and because PC manufacturers can’t add much more memory to the machines they already sell. Until PC makers install more 64-bit software that requires more memory, they don’t need more DRAM. But such widespread software upgrades could be years away. In the meantime, there are eight major DRAM vendors with only one — Samsung — making money.
The flip side to this coin is making more chips for less. Continuing down the process node so chipmakers can place more transistors on a chip is important in enabling economies of scale — and making DRAM for less. However, research into cramming more stuff on a chip is expensive, so those already bleeding red ink will be hard pressed to continue R&D into such efforts.
So if governments want to save DRAM, they should step out of the way of consolidation, invest in 64-bit software and perhaps in R&D that keeps DRAM chipmaking moving beyond 50 nanometers for manufacturing — not handing out cash to a business that’s currently recouping only a third of its costs.