Samsung's $26 billion bet
At more than 50 years of age, the semiconductor market is still sprightly, with growth this year in the 20 percent range. And with high growth comes short supply, which is why DRAM and flash prices have risen this year.
Samsung, which already produces around half the world's supply of both commodities, plans a dramatic increase - almost 2.5x - in next year's capital expense (capex) budget in production facilities, to $26B. To put this in perspective, Intel's 2017 capital budget - after a 25 percent increase from 2016 - was only $12B. In fact, Samsung's bet is almost as large as the total 2017 capex budgets of the next three largest - Intel, TSMC, and SK Hynix - combined.
Given that a greenfield plant now goes for $7B, this ups the ante in the high stakes poker of semiconductor manufacturing. The risk: industry-wide oversupply, slashed prices, and major balance sheet damage. But it's not all bad: we get cheaper memory and storage! For a while.
Endgame
Each of Samsung's top competitors now faces a difficult choice. Either they up their capital budgets to provide enough supply to maintain their market share, or give up trying to compete, as Samsung's higher volumes translate into economies of scale that no one else can match.
special feature
Anyone who attempts to match Samsung faces the daunting prospect of industry-wide oversupply. falling prices, and massive losses. Semiconductor plants have to be run at full capacity to obtain lowest costs, so as long as prices cover all variable expenses, and make a contribution to fixed costs, such as debt service, it's worth it to keep them running, even at a loss. It isn't easy to decrease production.
But if they don't ante up, competitors face the prospect of shrinking market share, higher costs, and eventually getting out of the business altogether. And don't count on an Intel and Micron joint venture to counter Samsung. Intel got out the DRAM business decades ago, and they've done very well despite it.
Intel/Micron may be counting on their 3D XPoint non-volatile RAM - NVRAM - to rewire the DRAM and flash markets in their favor. But, IMHO, their decision to tie 3D XPoint tightly to Intel CPUs dooms that hope. Besides, Samsung can buy or license competing NVRAM technologies from Crossbar or Nantero.
The Storage Bits take
The good news for us consumers is that in 12-18 months we should start seeing a return to ever-cheaper DRAM and flash as new capacity comes on line. But longer term we could face a world where Samsung dominates DRAM and flash production, slowing price declines, and a dearth of competition.
Ever cheaper memory and storage have driven our global conversion to a digital civilization. If we are to keep up the pace - and I'd argue it's essential for dealing with issues ranging from climate change to AI - we can't afford a permanent slowdown in their cost-effectiveness. This is bigger than a single industry.
Courteous comments welcome, of course.