May 8, 2023
WD doesn’t have a DRAM business, and they sell hard drives in addition to NAND and SSDs, so I read their releases for insight into the NAND market.
- · Wafer starts have been reduced by 30%, beginning in January.
- · Bit shipments decreased 14% QoQ (increased 20% last quarter.)
- · ASPs decreased 12% quarter-over-quarter on a like-for-like basis. Absolute pricing decreased 10% sequentially (was down 20% last quarter.)
- · The company “saw signs of demand stabilizing across various end markets” during their third quarter. This is a vague statement that projects that seems meant to sound more positive than it is.
- · Non-GAAP gross margin was down to 10.6% in the quarter. They are guiding for 3.0% to 5.0% gross margin in their fiscal fourth quarter. Their GAAP gross margin guidance is positive also (2.4% to 4.4%). This overall gross margin is buoyed by their HDD business. Their flash gross margin in the quarter was negative 5%.
- · The last five quarters of flash gross margin, from oldest to most recent, are (1%), 2%, (22%), (20%), (10%). NAND ASPs have almost been cut in half in just three quarters.
Quotes are in italics.
David Goeckeler – Chief Executive Officer
… we are beginning to experience improved demand at certain customers in China. In flash, we are seeing signs of stabilization and content increase per unit. PC OEMs have emerged from inventory digestion and are now shipping closer to end demand. Gaming will remain strong while enterprise SSD for cloud applications will remain soft.
Wissam Jabre – Chief Financial Officer
Flash gross margin was -5.0%, down 19.5 percentage points sequentially and 40.6 percentage points year over year.
We expect flash revenue to decrease sequentially as modest growth in bit shipments is more than offset by ASP declines. We expect flash bit shipment growth to accelerate in the first half of fiscal year 2024. In the fiscal fourth quarter, total gross margin will be negatively impacted by underutilization charges and flash pricing.
Question and Answer
I mean nobody’s building inventory right now, but we think the inventory correction there is mostly behind us. Our channel business performed really well this past quarter, I think above what our expectation was. The inventory issue is still very much in data center and it’s very lumpy. I – you know, we have some – some big – big cloud customers that are, you know, consuming.
Clearly, we’re going through one of the – one of the most severe downturns in a while, but we think as we move through the second half of the year, the market will get – will come into balance.
On the buying behavior of different NAND customers: I guess I would say that customers that are different segments of the market that are through their inventory digestion are now, you know, more or less shipping to end demand, being lean on inventory. There are instances of people doing strategic buys in that – in some of those cases where they’ve – they’ve got their inventory to where they want it, but now they – you know, they’re making their own – they have their own view of the cycle. That’s a – that’s a very – that’s a very small number, I would say. And then, the other ones is just inventory where they’re at in their – the data center customers especially are just in – there’s a variability of the level of inventory at each customer and they’re so big that it can impact the entire market.
So, you know, like I said, ones that have heavy inventory are basically working that down in an aggressive way. And so, I think it’s going to – like I said, I think it’ll be lumpy for a couple of quarters until we get – we get through that.
Yeah, we’re kind of at the point where we’re – you know, I think it’s – I think probably the same ZIP code is probably a fair way to say it [in response to the question asking if NAND production cuts are still in the 30% ‘zip code’ from their second quarter call]
However, if we exclude that [underutilization charges], then we’re – we’re – we’re close to mid-teens in – in Q3. Now, to your question on BiCS8, I mean this is with – with the way our road map works and the timing of when BiCS8 starts ramping, it is pretty much designed in a way to allow us to continue that mid-teens reduction – cost reduction over time.
Management believes inventory across the industry peaked has passed. They went as far as to say the NAND market will come into balance later in calendar 2023. Memory makers are carrying so much inventory, this doesn’t mean pricing will start to recover. It may only mean that demand will be higher than the supply coming out of manufacturer’s fabs. Memory makers have well over half a year of sales on hand. With normal demand, it will take months for this inventory to come down to normal levels. WD is holding the line on their 30% production cut started at the beginning of the year. At least, they are in this regime of reduction. Cycle time – end to end – for NAND is probably close to three months for them. If they cut wafer starts in January, the effect of that on product output from fully assembled and tested product wasn’t seen until sometime in April. Here we are in mid-May. If you stood next to the huge pile of finished product inside WD right now and watched it for a couple days, it is shrinking, and probably has been for a few weeks. But it is still a huge pile. Some customers also have large piles of finished memory products. But we have – if you believe the memory executives, who will put the industry situation in the best light they can – turned the corner of this downturn. Barring a recession, later this year (my guess is in the late fall or early winter,) customer inventories will be depleted enough that they will begin to raise their purchase volumes. At this point, if the memory makers don’t misjudge demand and flood the market with their huge piles of inventory, pricing will start to rise. Once this happens, customers may accelerate their buying, to try and replenish their memory inventories at the lowest possible prices. This stimulates demand, which causes prices to rise further. And this feedback loop is the beginning of the next upturn. The memory makers, bloodied by several quarters of losses and cash outflows, may overreact to the stronger market and flood it with their inventory pile. Whenever the recovery does happen, I think DRAM will get better before NAND does. I think that because DRAM turned down first, and because the NAND marketplace is fundamentally less healthy, with five players instead of the three in DRAM.
–S. Hughes (cyclical long MU, no WD position)