No Line of Sight

On 30 September, in his opening remarks on the company’s fiscal fourth-quarter call, Micron’s chief executive said: “we expect memory and storage supply-demand conditions to be much tighter in fiscal 2027 and 2028 than they were in 2026.” The prepared remarks repeat it in the market-outlook section for calendar years, then add the line I keep returning to: “we do not have line of sight to when supply and demand will return to balance.”

A forecast from a seller of the scarce thing is a company statement and should be read as one. What interests me is how little of the document is the forecast. Nearly every other claim in it has a number attached. “Much tighter” does not. So I went through the remarks looking for the numbers that would let me say what “tighter” means, and found that they point in a few directions at once.

What the document quantifies

Start with supply, because Micron gives it as growth in industry bit shipments. For DRAM, it expects growth “in the mid-20s percentage range” in calendar 2026, then “approximately in the low-20s percentage range” in 2027 and 2028, and says the industry will “remain supply constrained in both years.” For NAND the order is reversed, low-20s in 2026 and mid-20s in the next two. DRAM supply keeps growing at around a fifth a year, so the case for tightness rests on demand outrunning it, and Micron attributes the gap to a “structural gap between DRAM supply and demand growth rates.”

The remarks give no demand growth rate for memory overall. The nearest figure is for servers: unit growth “in the high-teens percentage range” in both 2026 and 2027, supported by “a modestly lower rate of content growth than prior expectations, amid tight memory supply.” Read plainly, scarcity is already slowing the growth of memory per server relative to what Micron expected. I’d call that the most concrete evidence of tightness in the document, and it shows up on the demand side.

Then there is when supply can respond. The dates below are from the remarks. Singapore HBM packaging begins output in early calendar 2027, Idaho’s ID1 fab starts wafer output in mid-2027, the Tongluo facility in Taiwan ships “meaningful” product in the same period, the new Singapore NAND facility starts in the second half of 2028, and Idaho’s ID2 fab and the Japan DRAM expansion follow in late 2028. Production from new facilities, Micron says, “takes time to ramp” and becomes meaningful “a few quarters after initial output.” The capex discussion adds that a majority of the planned capex increase is for construction, “most of which is to help accelerate clean room space availability in late calendar 2028 and beyond.” Read together, the DRAM capacity that starts in the window is mostly ID1, and the larger additions land near its end, which fits Micron’s claim that it sees no line of sight. That is my reading of the dates, not a figure Micron gives.

When six Micron facilities begin output, against the window it calls much tighter A timeline from mid-2026 to mid-2029 with a shaded band for calendar 2027 and 2028. Six dots mark dates stated by Micron: Singapore HBM packaging initial output in early 2027, Idaho ID1 wafer output in mid-2027, Taiwan Tongluo meaningful product shipments in mid-2027, Singapore NAND output in the second half of 2028, Idaho ID2 wafer output in late 2028, and Japan DRAM expansion initial output in late 2028. NEW CAPACITY DATES · MICRON FQ4 2026 PREPARED REMARKS, 30 SEPT 2026 Micron's "much tighter" window, calendar 2027 and 2028 202720282029 Singapore HBM advanced packaging: initial output early calendar 2027 — Micron prepared remarks, 30 Sept 2026 Idaho ID1 fab: wafer output mid-calendar 2027 — Micron prepared remarks, 30 Sept 2026 Taiwan Tongluo: meaningful product shipments mid-calendar 2027 — Micron prepared remarks, 30 Sept 2026 Singapore NAND facility: output begins second half of calendar 2028 — Micron prepared remarks, 30 Sept 2026 Idaho ID2 fab: wafer output late calendar 2028 — Micron prepared remarks, 30 Sept 2026 Japan DRAM fab expansion: initial output late calendar 2028 — Micron prepared remarks, 30 Sept 2026 Singapore HBM packaging · early 2027 Idaho ID1 wafer output · mid-2027 Taiwan Tongluo shipments · mid-2027 Singapore NAND · 2H 2028 Idaho ID2 wafer output · late 2028 Japan DRAM expansion · late 2028 Dots mark initial or wafer output, except Tongluo, which marks "meaningful" shipments.
Idaho ID1 puts DRAM wafers in the window by mid-2027. The two late-2028 DRAM starts, ID2 and the Japan expansion, arrive as the window closes.

Price increases are expected to slow

The quarter that prompted the forecast was mostly a price story. Micron reported fiscal Q4 revenue of $54.2 billion, up 379% on the year. DRAM revenue was $39.8 billion, up 27% on the quarter, on bit shipments up in the mid-single digits and prices up in the “high-teens percentage range.” NAND bits rose about 10% and prices about 30%.

The rest of fiscal 2027 is described differently. The CFO expects “higher gross margins beyond fiscal Q1 for the remainder of fiscal 2027, with a more moderate rate of price increases,” and calls fiscal Q1 the floor for gross margin. TrendForce, the research firm, published its fourth-quarter contract forecast the same day: conventional DRAM up 10% to 15% on the quarter and NAND up 15% to 20%, with “the pace of contract price increases is expected to moderate.” The measures are not the same (Micron’s is realized sequential change, TrendForce’s is a forecast of contract prices), so the figure below should be read as a direction and not a like-for-like comparison.

Micron's realized price change last quarter versus TrendForce's contract-price forecast for the next A dumbbell chart on a 0 to 30 percent axis. Conventional DRAM: Micron fiscal Q4 realized price increase in the high-teens, plotted at 18, against TrendForce's fourth-quarter 2026 contract forecast of 10 to 15 percent. NAND: Micron approximately 30 percent against TrendForce's 15 to 20 percent. QUARTER-ON-QUARTER PRICE CHANGE, % · MICRON 30 SEPT 2026, TRENDFORCE 30 SEPT 2026 0102030 Conventional DRAM High-teens percent (plotted at 18) — Micron fiscal Q4 2026 prepared remarks, 30 Sept 2026 Micron FQ4: high-teens 10% — TrendForce 4Q26 forecast, 30 Sept 2026 15% — TrendForce 4Q26 forecast, 30 Sept 2026 TrendForce 4Q26: 10–15 NAND Approximately 30% — Micron fiscal Q4 2026 prepared remarks, 30 Sept 2026 Micron FQ4: about 30 15% — TrendForce 4Q26 forecast, 30 Sept 2026 20% — TrendForce 4Q26 forecast, 30 Sept 2026 TrendForce 4Q26: 15–20 Filled dot: Micron's realized change. Rings: TrendForce's forecast range. Different periods and measures. "High-teens" is plotted at 18 and "approximately 30%" at 30.
Prices are still rising in both sources, only more slowly. "Tighter" describes the gap between supply and demand, and a gap can keep widening while the rate of price increase falls.

TrendForce’s note also says that “increases for some suppliers are expected to lag the market average due to pricing mechanisms stipulated in long-term agreements.” That detail leads into the part of the remarks I found most telling.

The contract book

Micron has signed 26 “strategic customer agreements,” multi-year take-or-pay contracts it estimates at “over 35% of our revenue through 2030.” Three-quarters of that revenue “has a defined pricing framework, a majority of which have pricing bands with floor and ceiling prices.” The remaining quarter is repriced periodically with the market. Customers’ financial commitments are $32 billion, “the vast majority of which are cash deposits,” and remaining performance obligations are about $150 billion. Micron adds that “even at floor prices, we expect margins meaningfully above any prior cycle peak margins.”

How pricing is set in Micron's strategic customer agreements A single bar split in two. Three-quarters of estimated agreement revenue has a defined pricing framework, a majority of it with floor and ceiling prices. One quarter is priced periodically to market. PRICING OF SCA REVENUE · MICRON FQ4 2026 PREPARED REMARKS 26 agreements, estimated at over 35% of Micron's revenue through 2030 75% — defined pricing framework, a majority with floor and ceiling — Micron, 30 Sept 2026 25% — pricing negotiated periodically on market prices — Micron, 30 Sept 2026 75% defined pricing framework 25% market A majority of the 75% carries floor and ceiling prices. Customer commitments: $32 billion, mostly cash deposits. Shares are of estimated agreement revenue, not of all Micron revenue.
Roughly a quarter of the contracted revenue floats with the market. The rest is priced inside a framework, which cuts both ways.

A floor protects Micron if the shortage ends early. A ceiling protects the customer if it deepens. The deposits are not payment: Micron says they are unrestricted but “will be returned to customers over time, toward the latter half of each agreement’s term, assuming minimum purchase requirements are met.” My reading, and it is only a reading, is that a buyer expecting a glut would be unlikely to post cash to lock in supply, and a seller expecting prices far above every ceiling would have less reason to accept one. That would make the contracts evidence that both sides believe in “tighter” and neither wants to be exposed to the alternative. Micron also gives a plainer motive: it says the agreements give it “greater visibility from SCAs into our demand through the end of the decade and beyond,” which it cites as a reason to raise capital spending in fiscal 2027. And it says it expects revenue to “well exceed” the remaining performance obligations, which implies it expects prices above the floors or volumes above the minimums.

HBM, the stacked memory that sits beside AI accelerators, is where the contracts look firmest. Micron says it has “completed agreements for the vast majority of our calendar 2027 HBM bit supply with significant price increases year over year.” TrendForce raised its 2027 HBM price outlook the day before, projecting a blended average selling price up 121% on the year. That is the research firm’s forecast, not a contract figure from any vendor.

One number in the remarks does not obviously fit the word “tight.” Inventory stood at 129 days at quarter end, up nine days on the quarter. Micron attributes the rise to build-ahead ahead of node end-of-life and to incentive pay absorbed into inventory cost, and says it expects days of inventory to fall in coming quarters. That is a plausible explanation, and I have no way to check it from outside. It is also a reminder that a company can describe its market as “extremely tight” while carrying more than four months of stock, and both statements can be true.

Micron has seen the opposite

Micron has lived through the opposite condition. In a 1988 article on DRAM prices, the journalist Simson Garfinkel reported that chips “that had cost $5 suddenly cost 25 cents” when Japanese capacity flooded the market in the mid-1980s, and that by the end of 1985 five of the seven American makers of high-capacity DRAMs had left the business, “leaving Texas Instruments and Micron Technology as the only American sources.” By 1988 a shortage had come, and chips that cost $2 two years earlier cost $15 to $18.

In that same article Micron’s then-chief executive, Joe Parkinson, said American buyers should build long-term strategic partnerships with their suppliers and stop hunting for the lowest spot price. For companies that did, Garfinkel wrote, he said “there never has been a shortage and there won’t be one in the future.” The 2026 contract book, 26 agreements with deposits and price floors, echoes that idea at a much larger scale. The analogy is mine: Garfinkel’s Parkinson was talking about securing supply, not about floors or deposits. In 1988 the argument came after a collapse and a shortage. This time it comes with the market at a record.

What would change my mind

I can’t tell from the outside whether “much tighter” is right, and Micron doesn’t say what number it would take for the claim to be wrong. A few figures would show it. The DRAM contract-price series should keep rising next quarter, but at a falling rate, as both Micron and TrendForce expect. Micron’s days of inventory should fall back from 129. Industry DRAM bit growth should stay in the low-20s. If demand disappointed, I’d expect the first sign in inventory days, a build alongside those slowing price increases, before any price cut. The deposits make that slower to show up.

There is also a demand-side counterweight I can only gesture at. Memory per token is something engineers are working to shrink, a point I made in an earlier essay on a claimed 437x cache reduction. Micron’s remarks point the other way, citing “growing model parameter size, longer context lengths and higher concurrency.” Both can be true over different horizons. A statement about 2028 is a bet on which one wins.

References

  1. Micron Technology (2026). Fiscal Q4 2026 Earnings Call Prepared Remarks. 30 September 2026. Source for every Micron quotation and figure in this essay, including supply-growth ranges, facility timing, price and bit changes, strategic customer agreement terms, HBM commitments, inventory days and capex.
  2. TrendForce (2026). AI Server Demand Sustains Memory Contract Price Increases in 4Q26, While Consumer-Side Pressure Persists. Press release, 30 September 2026.
  3. TrendForce (2026). HBM Supply Constraints Persist, 2027 Price Outlook Revised Upward with Blended ASP Forecast to Rise 121% YoY. Press release, 29 September 2026.
  4. Garfinkel, S. (1988). The Story Behind the High Price of DRAMs. Computer Update, September/October 1988, pp. 16–17. Background; scan of the author’s copy.