A true cyclical market, the NAND flash business goes through periods of booms and periods of busts. Following a very profitable boom year in 2018, it looks like the market is in a down swing, as an oversupply is starting to impact the bottom lines of memory makers. To stem any potential for significant losses or an outright market crash, three major manufacturers of NAND memory — Intel, Micron, and SK Hynix — have announced that they will be taking measures to address the oversupply, such as reducing flash production, cutting down wafer starts, and/or slowing down ramp ups of new fabs. Furthermore it is highly likely thr another major manufacturer, Samsung, will follow suit.

The rapid transition to high-capacity 64-layer and 96-layer 3D NAND memory devices has enabled NAND flash manufacturers to increase their NAND supply (as measured in bits) and ultimately saturate the market with loads of flash memory. Meanwhile, demand for servers in the recent months has been weaker than expected, smartphone replacement cycles are getting longer, and other drivers of NAND demand have also disappointed. As a result, NAND supply has well exceeded demand, causing prices to fall by as much as 20% across multiple categories in Q1 2019, according to TrendForce. To ensure their short-term and long-term profitability, at various points in the last couple of months the three manufacturers have all announced that they are taking actions to minimize their exposure during this latest bust.

Micron said back in March that it was carefully managing its NAND bit supply growth (to tackle oversupply at least partially) and started to decrease its total NAND wafer starts by roughly 5% by cutting its legacy nodes. The company did not indicate plans to shrink its NAND bit supply, but reducing production of memory using older process technologies will likely lower its costs.

Meanwhile SK Hynix this week said that it had stopped production of 36-layer as well as 48-layer 3D NAND memory, which these days has a rather high per-bit cost relative to newer technologies. In the coming months the company plans to increase production of 72-layer 3D NAND and in the second half of the year it intends to release 96-layer 3D NAND solutions for the SSD and mobile markets. Furthermore, SK Hynix will slow down the ramp up of its M15 fab in Cheongju, South Korea. The company expects its NAND wafer output to decrease more than 10% compared to 2018. Just like Micron, SK Hynix does not seem to have plans to lower its NAND bit production, so it will still more memory than it did last year.

Intel, which has traditionally concentrated on the enterprise part of the SSD market, has also announced this week it will reduce its NAND output in 2019. Intel did not elaborate whether it intended to reduce the number of wafer starts, or do something more radical. But regardless, the company continues to expect challenges with prices of NAND memory going forward, and is acting accordingly.

Finally, while Samsung yet has to announce its Q1 2019 results, it has already warned investors that its profits for the quarter would be down 60% compared to Q1 2018. Analysts have been attributing this to multiple factors, including demand for flagship smartphones, lower prices of DRAM and NAND memory, and other weak markets. Given that the other major memory manufacturers are all taking steps to address the current oversupply, It is more than likely that Samsung will also adjust its NAND business this year; though how they'll do so remains to be seen.

Related Reading:

Sources: Intel, SK Hynix, TrendForce, Micron/SeekingAlpha, Samsung

Comments Locked

57 Comments

View All Comments

  • shabby - Friday, April 26, 2019 - link

    Do they really think people will buy if the price goes up?
  • smilingcrow - Friday, April 26, 2019 - link

    Looking at what people have paid for GPUs and RAM over the last 2 years will give you an idea.
  • willis936 - Friday, April 26, 2019 - link

    There is a negative correlation between GPU price and how much consumers spend on GPUs. So no, increasing price is not a good strategy.

    https://www.statista.com/chart/15843/nvidia-and-ra...
    https://www.statista.com/chart/17089/nvidia-revenu...
  • Opencg - Friday, April 26, 2019 - link

    it was show to be negative recently due to the price difference within a few month time period. when 10 series were overstocked the 1080 ti was 500. vs the similar performance 2080 for 800. however people still buy the cards and now that the deals have been crap for the past 6 months (outside of used) the profit is still comming in from these poorly priced cards.

    if price on nand flash goes up you can expect a couple months of lower sales then back to normal profit levels. its really complicated because economy of scale and perhaps one producer decideds to undercut but consumers overall are seeing good deals right now on these products so i would expect a slight swing backwards next year
  • Solandri - Friday, April 26, 2019 - link

    GPU prices haven't actually increased. Their price has tracked pretty closely with inflation. The only real blip is that Nvidia introduced the Titan as a high-price, high-performance, low-volume bragging rights product, where no similar product existed in the past.

    https://hexus.net/tech/news/graphics/103399-inflat...

    The historical time constant for NAND price cycles has been about 1-2 years. The current price slide began around 3Q of last year. So we more than likely have another quarter or two of price drops before it stabilizes. It could be shorter or longer though - you can never be sure.

    https://static.seekingalpha.com/uploads/2017/9/24/...
  • f4tali - Saturday, April 27, 2019 - link

    "The greatest trick the Devil ever pulled was convincing the world he didn’t exist."

    Nvidia outsmarted a lot of smart people with a supersly move:

    Nowadays a 1080 is a 104 chip and even a Titan (10 series) only a 102.

    Whoever Mark Tyson is.... he's wrong (sorry).
  • f4tali - Saturday, April 27, 2019 - link

    IOW GPU prices have more than DOUBLED since the 500 series...
  • f4tali - Saturday, April 27, 2019 - link

    In other OTHER words would you rather pay $554.93 for 520mm squared or $1200 for only 471mm squared?

    And yes, that's a rhetorical question :P
  • LordSojar - Saturday, April 27, 2019 - link

    Wafer size plays into pricing, no doubt.

    But it's not that simple. nVidia has sheer architectual superiority right now, and that shows with a 12nm (aka 16nm with improvements) chip outperforming a 7nm (a true generational improvement from 16nm+++), despite the size.

    Yes, the sizing is influencing price, as is greed... but sheer R&D costs of these chips is absurd. To shift technology as nVidia and MSFT are doing requires you to be big and spend a LOT of money.
  • HStewart - Sunday, April 28, 2019 - link

    Wafer size does mean smaller nm means cheaper but because of cost of manufacturing is so much more expensive when it gets smaller and not guarantee that same size wafer means more chips.

    If you look at cost to create new Fab's for chips - NVidia and MSFT prices is probably just a tip of Iceburg

    But one thing is the nm size does not mean performance. It takes the right architexture below it to make it happen. But architexture changes cost more R&D in some ways because it requires new ways of thinking.

Log in

Don't have an account? Sign up now