MU - Micron May Be Authoring One Of The Best Upcycles In History

2024-01-09 08:30:00 ET

Summary

• Micron's CapEx plans to structurally reduce bit output, and its latest call to prevent "pull-in" ordering mean this time is indeed different.
• Management has said it has visibility for the year on its leading-edge nodes, something unheard of in this industry.
• Between managing its orders and its strategic CapEx allocation, this time is different, just not how you've always been told it would be.

Micron ( MU ) has all but ushered in the beginning of the upcycle, but from the looks of the memory market, this is indeed the case. The stock has already rallied from its $48 lows to highs near $88, yet the fundamental upcycle has literally just begun. Are you asking yourself, "Is it too late to invest in the upcycle?" Perhaps you didn't get on board when I called the turnaround of the cycle around $49 and feel you're chasing now. I have some good news for you. Based on both the CapEx plans for this upcycle and the actions Micron's management described on its recent conference calls with orders, the stock may have quite a bit more to go after a period of consolidation.

If you're not new to the Micron cycle adventures, you'll know the big principles are book value and CapEx spend. Essentially, watch how much is spent on WFE (wafer fab equipment) relative to the prior years, and keep an eye on book value so you can "properly" value the stock on a price-to-book basis.

The problem with both is the nature of what's behind them. CapEx increasing doesn't necessarily mean higher bit supply, whereas predicting future book value and free cash flow have too many variables, with exponential change in probabilities each quarter further out.

Generally, most target a number, apply the ratio to the stock price, and call it a day.

But the problem with doing so today is we're entering an upcycle never before seen. It may not actually "be different this time," but the same factors of the status quo times may be burying the needle on the gauges this time around.

Different, But Not The Different You Were Told

However, there are a few things that are objectively different this time.

As a quick memory jog of the last few months, DRAM and NAND pricing has shot up higher and faster than any outlet has expected. The reason for this is relatively straightforward: supply has been reduced meaningfully and structurally in a manner it hasn't been approached before.

This is point number one.

The second point comes more recently from Micron's earnings call around organizing customer demand so prior upcycle patterns aren't repeated, namely, double ordering.

These two combined points - one firm and one yet to be proven but given credence due to the former - provide a materially different look to this upcycle, one capable of going on for longer while stronger, given the world doesn't fall into a massive recession.

Different CapEx Setup

First, I'll refresh your memory on the CapEx spend and then highlight the new aspect of what Micron's management is trying to do.

In October, I highlighted three areas of Micron's CapEx plans :

1. WFE Spend Isn't Up
2. Construction CapEx
3. Packaging And Testing Allow More Throughput, Not Output

Even though overall CapEx spending is set to increase in FY24 over FY23, it doesn't mean more bit supply. In fact, supply will be significantly down from the last upcycle to the start of this one. The reason is the company has done something it has never done before (and we now know the rest of the industry has followed suit), which was to take the idle, underutilized equipment and move it to undersupplied nodes and product areas. Doing this structurally reduces wafer starts and permanently removes the underutilized lines' idle capacity (unable to turn the machines on and produce the former supply at the flip of a switch). And while it moves the equipment to other lines, these other lines require more equipment to produce the same amount of wafers.

Furthermore, because of the shift in technology with DDR5 and HBM (high bandwidth memory), the amount of bits per wafer is roughly cut in half due to the doubled die size. Therefore, production in legacy nodes and technologies (like DDR4) is structurally reduced permanently to the cut downcycle output. At the same time, the repurposed equipment makes less than half of the output they were initially intended for.

What this means is there's a no-going-back type of situation in supply. Micron - and by following suit, the industry - has reset DRAM supply to a level from years ago. This is what I concluded in my October article:

This coming upcycle may be one of the best ones due to the shift in strategy coming out of a downcycle by removing supply completely from legacy lines. It is an excellent setup for Micron's business and stock over the next year.

So, while the headline number of higher CapEx appears to mean more supply, the investments are going toward bigger cleanrooms for the installation of this moved equipment but also quite a bit for back-end packaging and testing, which is a significant part of Micron's manufacturing chain. With HBM requiring sophisticated packaging, the CapEx investment is to enable this high-margin, high-demand product. Plus, back-end packaging and testing doesn't add supply to the fabs.

Therefore, CapEx isn't being spent on increasing bit supply; it's being used to structurally reduce bit supply and open up bottlenecks in packaging and testing.

Different Ordering Process?

So, how does this connect to the recent point I heard on FQ1 '24's conference call?

For one, when supply underruns demand, pricing moves up. When pricing moves up, there's a general run on orders from customers to fulfill their future needs. Ordering early gives them better pricing before it increases as the upcycle progresses. But then, they continue to order in case they need to get back "in line" of the order queue to get their next order to fulfill their end demand because, at that point, everyone has lined up to get memory, leaving them up a creek without a paddle.

If they don't order ahead of time, they may not have the components they need to fulfill their orders.

In the past, this has been an Achilles heel of the memory industry. It's otherwise known as double ordering.

But on the conference call, something caught my attention, and between it and the post-conference call analyst call, no one asked a pointed question about it even though I considered it a critical point.

Considering the industry is only just a few months in of pricing moving up after quarters of it dropping, it's interesting to hear Micron's management talk about tight inventory, pull-in, and sold-out production already.

Our leading-edge inventory is very tight, and we are also working to minimize pull-in of customer demand in response to higher pricing. As a result, our sequential growth in the near term will be driven primarily by pricing rather than a sequential increase in bit shipments. Both DRAM and NAND bit shipments are expected to decline somewhat in the fiscal second quarter.

- Mark Murphy, Micron CFO, FQ1 '24 Earnings Call

We expect supply-demand balance to tighten in both DRAM and NAND throughout 2024. Our leading-edge DRAM and NAND nodes are oversubscribed for the full year.

- Mark Murphy, Micron CFO, FQ1 '24 Earnings Call

I'll tackle each quote individually and provide the context for how they go together.

This is the first time I've heard about active anti-pull-in ordering. Management also said bit shipments will be down slightly for FQ2 and FQ3 after a 20% rise in FQ1. This limits bits shipped (i.e., not sending everything it has out the door) while letting pricing take its course higher these coming months. Management is calling it seasonal and mix shifts, but there's more to it than that. With the choices it made with CapEx for this upcycle I discussed earlier, the strategy is coming to fruition - essentially "delaying" orders. Management is literally saying it's working to minimize pull-in, and as a result, sequential bit shipments will be down.

Another word I heard also led me to believe there's a shift in order management. I noticed a word not typically used, one I haven't heard before in Micron's context: subscribed. It talked about its leading nodes being fully subscribed and/or oversubscribed. This goes hand in hand with the talk of preventing pull-in by selling out of bits for leading nodes. In other words, it's telling its customers, "Sorry, that's all we have for you this year; order more when our backlog can afford it."

The beauty of this situation is Micron can't do anything about it now that it has made its bed with its CapEx decisions. Some might say this is bad for business since it can't fulfill all the demand it needs to rake in sales and get profits pumping. However, in this industry, selling fewer bits at much higher prices with a buffer for demand to ebb and flow is better than selling a lot of bits at somewhat lower prices, but demand can't dip, or else the pricing environment turns unfavorable.

How This Will Drive "Different"

Limiting negative free cash flow and EPS is the most significant part of navigating a downcycle. Losing the least amount of money is the goal. However, getting out of the hole but not rushing the process is more beneficial in the long term. By underutilizing equipment early on and having the industry follow, the memory market has turned favorably into the producers' hands. But it's not just because they cut production; it's how they have restarted production.

By reallocating equipment formerly producing oversupplied nodes and products to leading nodes requiring more equipment with larger dies per wafer, the company has found a way to truly limit upcycle production with less evidence of collusion. It's running at 100% (or will be once the migrations are complete) and doing everything possible to produce in-demand chips.

Therefore, it will run at "full capacity" of the new structure faster. Otherwise, in the past, it would have to wait quite a few quarters before it could fully bring online the idle equipment if demand had only marginally moved up, resulting in underutilization charges for quarters on end. Today's situation has become a supplier's market without demand having to do anything. Combined with not having to grow CapEx more than marginally, it'll cost less to begin this upcycle, meaning a return to free cash flow quicker.

The main idea is setting up for the upcycle and not merely getting out of the downcycle in any way possible.

Again, the key wasn't cutting production; it was how to return to production, and Micron has shown this different path has put it in the driver's seat. Even with negative free cash flow predicted for another two quarters, pricing and the memory outlook have never looked better at this stage of a cycle. Never before has the memory industry taken this approach, so it is different this time, just not how everyone expected it to be, where demand drivers didn't have lulls in demand ever again.

This has allowed management to then work toward organizing its orders against its supply, providing visibility for the next year - something it has never been able to do in prior cycles.

The bottom line is Micron is taking control of how this upcycle forms. It's actively reducing wafer starts as it shifts WFE to other lines and shipping fewer bits from the last quarter as prices rise dramatically. The increase in supply expected as this upcycle turns will be one of the lowest we'll see in the previous two decades - mainly because there will be an overall decrease combined with strategic output.

Does this mean the stock will go straight up into triple digits and never return? No, I expect the stock to consolidate its run from $48 to $88 before heading higher. And as long as supports hold, the stock is a buy on dips as it formulates the biggest advantage to an upcycle against customers ever.

For further details see: