My regular clients who get quarterly in-person market updates and those readers who have attended my annual Flash Market Update session at the Flash Memory Summit have often been told how Objective Analysis models the memory market based on supply/demand dynamics and the fact that memory chip prices tend to follow a very predictable pattern. In this post The Memory Guy blog will explain that pattern to those unfamiliar with it and will show where we are, leading to an outlook on what to expect from here.
Typical Pricing Behavior
Since this is an abbreviated explanation I won’t go into much depth here, but will simply state that for the past 49 years DRAM prices have followed a repeated pattern that is also followed by NAND flash. Since these two technologies account for 94%-95% of the memory market, and since they are the two commodity memories that undergo the strongest cycles, they are the technologies that determine the memory cycle, and even the semiconductor cycle.
If you model the cost for either of these technologies as Objective Analysis does, and compare that cost to the current market price you will find that market prices follow three different phases, depending on the state of oversupply or shortage:
- Shortage: With rare exceptions, when a shortage begins prices level off or may even (in extreme circumstances) increase by more than 10%. For our forecast we model that shortage prices will remain flat. Meanwhile, costs steadily decline, resulting in increasing profits for memory manufacturers the longer the shortage lasts.
- Oversupply Begins: Prices fall precipitously, typically at a rate of 60% in two quarters, until they reach cost.
- Continuing Oversupply: Once prices have reached cost their decline continues to match costs until the next shortage begins.
This is illustrated in the idealized chart below:
Current Market Status
Today’s market is toward the beginning of the phase where prices follow costs. Manufacturers have experienced steep price declines, and have reached a point at which they have taken inventory write-downs. Often these are expressed as “Lower of cost to market” since the market price has dropped below the production cost of some of their inventory.
With the exception of the write-downs, chips are selling at their production cost. Profits have disappeared.
The curve above shows that the rate of decline of chip prices slows once cost is reached. How does this compare to actual market prices today?
The chart below gives the lowest price per gigabyte of DRAM and NAND flash on the spot market. I want for you to look at the shape of the curves, rather than to try to read the DRAM prices on the left vertical axis and the NAND prices on the right.
You can see that NAND flash (the red line) slowed its collapse in September, and the DRAM price collapse (black) slowed starting around February. Keep in mind that these are spot prices, so they will react earlier and more dramatically than will contract prices.
This means that we are still relatively early in the cost-following phase. It’s hard to tell when this phase will end, and when the market will re-enter a shortage, which will cause prices to flatten once more. Demand in the first half of 2022 was higher than normal, and manufacturers found a way to satisfy that overblown demand, so they can currently produce a lot more DRAM and NAND than the market needs. The puzzle is to determine when real demand will rise to the level of early-2022 production.
Here’s a rough explanation of the findings of our in-depth analysis.
What Does the Future Hold?
We expect for demand to grow steadily from its current depressed state, and might meet production capacity by the middle of 2024.
We are being cautious, because it’s hard to determine when demand will again match supply, especially when a demand change is the cause of the cycle rather than overinvestment. Most semiconductor cycles are driven by overinvestment. This one is not.
When a shortage returns, then profits will return.
But revenues will turn around much sooner than that. The most recent WSTS statistics tell us that April semiconductor revenues were higher than those of March. We might already have passed the revenue trough. A revenue turnaround, though, doesn’t mean a return to profitability.
Revenue growth results from price declines that are slower than gigabyte growth. That’s a pretty easy goal to reach once prices start to follow costs.
So expect for semiconductor revenue growth for the remainder of 2023 and the first half of 2024, but don’t look for profitability until the middle of 2024.
Make Your Company Prosper
Both manufacturers and buyers of semiconductors take advantage of this kind of knowledge to ensure that their companies perform better than their competition both in good times and in bad. Objective Analysis’ clients understand this and use it to their advantage. You can benefit too. Please visit our website and call or write us to explore ways that we can work together to make your company overcome the current market’s obstacles.