8.5.24
Mark Murphy (CFO) and IR representative in attendance.
- · Overall, the NAND and DRAM markets are on an upward trajectory, on the way to a “substantial” revenue record in their FY-25.
- · Traditional server unit demand is recovering. AI server demand remains strong.
- · Some customers have and are buying memory ahead of anticipated price increases.
- · Automotive, industrial, and consumer markets are weak. China also continues to be weak.
- · The company forecasts both DRAM and NAND bit shipments to be flat in their November quarter (FQ1-25.)
- · They expect gross margin from the August to November quarter to increase sequentially “a couple hundred basis points.” This is surprisingly low.
- · FY-25 HBM revenue will be “multi-billion.” They are sticking by their forecast to match their HBM share to their overall DRAM share sometime in calendar 2025.
- · They believe DRAM prices will continue to go up in “25.” Where pricing is not what they want it to be, they are not taking the business. This means they are playing chicken with the market, holding inventory instead of selling, anticipating that DRAM prices will continue to rise and they will be able to sell that same memory for more later.
- · They believe their bit supply growth (implied to mean DRAM) in “25” (probably fiscal year) will be below demand. The 3-1 trade ratio of HBM memory is the major factor in this.
- · Inventory will be around 150 DIO through the first fiscal quarter (through November) and then it will start to decline from FQ2-25 through FQ4-25, then reaching their target inventory level of around 120 days. This is a major strategic bet Micron is placing here. They aren’t getting the pricing they want and are choosing to hold inventory for at least the next three-ish months in anticipation of higher pricing. At this point in the cycle, it is surprising and concerning that Micron isn’t seeing stronger demand and getting better pricing.
- · Sharp ramp of HBM happening now will affect cost downs in the first quarter, said cost will be flat to up in that quarter, seems to mean FQ1-25. This could explain the weak increase in gross margin he indicated earlier.
- · Murphy said that HBM is also a differentiated product and gets a price premium for that. New products will be introduced annually. As I’ve said before, I don’t believe HBM will not be a commodity in the long term, though it may be in these early stages of the lifecycle of this product.
- · On HBM, their product is “sold out” through “25” (assume this is FY-25) and pricing is “largely set.”
- · They believe their DRAM share is “low 20s percent.”
- · HBM market is 4% of industry bits this year (calendar 2024?) and will be 5% of industry bits next year. I thought the bits would be a higher share next year given the steepness of the ramp happening across suppliers. They believe the HBM TAM in 2025 will be over $20B.
- · Micron has “limited ability” to upside their HBM supply to customers. The ramp is limited by getting the assembly and test equipment needed for the HBM product.
- · Said again they are willing to hold inventory if pricing offered isn’t good enough, because they believe the DRAM market will continue to get stronger through FY-25.
- · EUV will be inserted in the 1-gamma DRAM node.
Summary
The bit bet being made by Micron is to hold inventory in anticipation of the DRAM market continuing to be healthy. The company is underway with a steep ramp of HBM, which is taking bit capacity away from other markets. Rather than adding wafer capacity to make up for it, Micron is holding inventory now with the belief they can sell it later to fill the coming shortfall of supply caused by the high HBM wafer trade ratio. This is a bold move by the company, one without precedent. I like their confidence and willingness to make a big, risky strategic bet.
– S. Hughes (cyclical long MU)