Micron Sold Its Own Ceiling — And a 7x P/E Is the Price Tag

The most immediate milestone is three days away — September 30. The headline numbers Wall Street will watch at Q4 FY2026 earnings are whether revenue hits the $50B ± $1B guidance and whether gross margins land around 86%. But in my view, the real numbers are elsewhere. The SCA disclosures are what matter. Specifically: how much has the RPO moved from the prior quarter's $5B, how many new SCAs have been signed, and what is the cumulative deposit collection status. If Q4 shows meaningful DRAM coverage expansion from the current roughly 20%, that reads as a signal the band model is growing. If RPO stagnates and no new contracts appear, it could be an early warning that customers are diversifying volume to ceiling-free competitors rather than committing to Micron's band structure.

Near-term supply dynamics remain strongly in Micron's favor. With the company able to meet less than 50% of data center customer demand, DRAM is functioning as customers' number-one operational constraint ahead of power, land, and logic wafers. Server DRAM contract prices rising 13~18% quarter-over-quarter in Q3 is direct evidence demand is overwhelming supply. In this environment, the cost of SCA ceilings grows every quarter. Every quarter market prices stay above the ceiling, the absolute dollar amount of revenue Micron foregoes increases. But simultaneously, this same supply shortage keeps pushing the scenario where SCA floors are tested further into the future. Near-term, ceiling costs are real and immediate; floor protection remains potential and deferred — that asymmetry will persist for the foreseeable future.

HBM4 shipment momentum also shapes the near-term outlook in meaningful ways. As of Q3, HBM4 (1-beta) is in mass shipment for a lead customer's platform, with HBM4E (1-gamma) volume production scheduled for 2027. HBM requires far fewer bits per wafer than standard DRAM, so as HBM's share of DRAM wafer starts rises from 22% in 2026 to 30% in 2027, the bit supply share climbs from 9% to 13% — but the gap between wafer share and bit share structurally crowds out conventional DRAM supply. This crowding-out effect reinforces conventional DRAM price floors, further reducing the odds that SCA floor pricing needs to activate in the next six months. As long as AI demand continues pulling HBM capacity expansion, the structural DRAM supply shortage remains intact and the floor stays theoretical.

The most critical medium-term variable is the 2027 DRAM supply outlook. TrendForce analysis projects global server shipments will grow more than 17% year-over-year in 2027, with DRAM fulfillment rates expected to remain in deficit. With new fab ramps delayed to 2028, supply won't catch up to demand for at least another one to two years. During this window, the opportunity cost of SCA ceilings keeps accumulating. The longer market prices stay above the ceiling, the larger the cumulative revenue gap between Micron and SK Hynix grows. But this same period gives Micron time to sign additional SCAs and expand band coverage on favorable terms — supply deficits weaken customers' negotiating position, and Micron can use that window to close more contracts before market dynamics shift.

The expansion path of SCA coverage is the central fork in the medium-term outlook. If coverage climbs from the current DRAM roughly 20%, NAND roughly 33% toward the planned approximately 40% of total revenue, Micron's earnings volatility structurally declines. If it pushes further toward more than half of revenue under SCAs as suggested in earnings call slides, the company becomes roughly half cyclical and half subscription in character. But wider coverage simultaneously amplifies the ceiling's constraint. If half of revenue is ceiling-capped while memory prices keep rising, the quarterly revenue divergence versus SK Hynix becomes increasingly visible in reported numbers. That trade-off will almost certainly spark shareholder debate, and management will face growing pressure to prove the SCA strategy's value in concrete, verifiable terms rather than qualitative characterizations.

Another medium-term variable worth watching is whether the US manufacturing premium factors into SCA pricing. According to TrendForce reporting, the premium for US-manufactured memory is already being reflected in SCA price-setting. Micron is executing large fab investments in Idaho and New York, and US production costs running higher than Asian alternatives is a well-established reality. If this premium is already baked into SCA ceiling prices, the effective ceiling could sit higher than Asian-market pricing — reducing the practical cost of the constraint. But from the customer's perspective, paying a US manufacturing premium within an SCA creates incentive to compare against Asian-sourced alternatives. Whether this premium accelerates SCA expansion or becomes a limiting factor will need to reveal itself in the next several quarters of results.

Competitive dynamics will also produce subtle but meaningful shifts over the medium term. Reports indicate SK Hynix has removed price caps from its LTAs, and Samsung is reportedly moving from one-year contracts toward three-to-five-year long-term agreements as well. All three memory giants are shifting toward longer-term contract structures while adopting structurally different pricing frameworks. Customers now evaluate Micron's ceiling-capped structure against SK Hynix's ceiling-free structure, allocating volumes based on stability versus flexibility preferences. For Micron to remain competitive, the differentiating value of floor protection and $18B in deposits needs to register as genuinely meaningful to customers — not just as an abstract hedge but as a reason to accept a ceiling. If customers don't see tangible value in a ceiling-capped contract, SCA expansion stalls and the band model fails to scale.

The 2028 to 2030 period is when the SCA band model faces its real test. Two scenarios diverge sharply. If AI demand continues on its current trajectory, memory prices even after supply normalizes in 2028 will likely remain above historical averages — in that case, SCA ceiling costs accumulate but floor protection is never activated, and Micron ends up paying insurance premiums without ever filing a claim. Alternatively, if the AI investment cycle decelerates around 2029, or alternative architectures like CXL or near-memory compute begin displacing HBM demand, memory prices could fall toward or through floor levels for the first time. That second scenario is when SCA's real value finally surfaces — and when the floor pricing commitment from customers gets its first genuine test against adversarial market conditions.

Looking at the widest frame, Micron's SCA experiment could signal a structural transformation for the entire memory sector. Memory semiconductors have historically been defined by high growth and high volatility — the cyclical characteristics that have permanently discounted their valuations. If all three major players move toward long-term contract structures that improve revenue predictability, the market may reclassify memory semiconductors alongside predictable infrastructure sectors like semiconductor foundries or industrial gases. That reclassification would produce a structural rerating — taking P/E multiples from single digits to double digits across the sector. Micron, as the first mover with the most developed band architecture, would be the primary beneficiary if this shift materializes.

Pulling the scenarios together: in the bull case, AI demand exceeds expectations and supply shortages extend beyond 2028, meaning SCA floor pricing never needs to activate. The $18B in deposits and take-or-pay security get recognized by the market as structurally eliminating downside risk, and the multiple expands as the bond-like earnings profile becomes undeniable. In the base case, supply and demand reach equilibrium around 2028, memory prices soften, and SCA floors activate for the first time — Mehrotra's claim that floors deliver margins "well above" the 2018 peak of around 61% gross margin gets tested against actual results. If it checks out, that represents a historically dramatic improvement versus past troughs and validates the band model. In the bear case, AI investment contracts rapidly and customers push for take-or-pay renegotiation, putting SCA's legal enforceability on trial in ways that will define or destroy the band model's credibility permanently. Investors watching September 30 earnings should track three specific SCA figures beyond the headline numbers: RPO movement, cumulative deposit collections, and SCA coverage ratio. These three data points will show whether Micron's band model is on track, stalling, or accelerating into the next phase of its existence.