After 87% Gross Margins, How Much Cycle Is Left for Micron?

The headline results:

The last two points are the interesting ones, and they need to be read together. Wall Street analysts kept raising their models in the weeks before the report, and the newest numbers had already pushed next-quarter revenue to $62.2 billion. Micron's $61.5 billion guide came in just under that, yet nearly 8% above the older forecasts. The guidance is hardly disappointing. It just didn't hand any extra surprise to people buying the rally. That gap probably explains the muted 3% move the next day.

The quarter-by-quarter trend says more about how steep this cycle is than any single quarter does:

Quarter-over-quarter revenue growth went from 75% to 74%, then to 31%, and the next-quarter guide implies 13%. The dollar gains are still large, but the curve is clearly flattening. Gross margin climbed from 57% all the way to 87%, but the gain each quarter shrank from 18 points to 2. To put 87% in perspective: it's like a restaurant that spends just $13 on ingredients for every $100 it takes in. That is close to the cost structure of a software company. Memory chips, by contrast, have always been a capital-heavy, highly cyclical commodity business. There is a ceiling on gross margin, and there isn't much room left to squeeze out more.

Broken down by product line, the contribution from price increases is plain to see:

By business unit, data center grew fastest and phones and PCs grew slowest:

(Current-quarter figures from the company 8-K, 2026-09-30; prior-quarter figures from the 10-Q, 2026-06-25)

MCBU revenue rose even as shipments fell. That means phone and PC makers are paying bigger bills for fewer bits, and consumers' ability to absorb higher prices is being tested. We take this thread on directly in the supply chain section below.

Micron gave no full-year guidance, saying only that "we expect an even stronger fiscal 2027" (8-K, 2026-09-30). That bar is so low it barely counts. Next quarter's $61.5 billion guide alone is already close to 46% of fiscal 2026 revenue, so even if the remaining three quarters came in at just $24 billion each, fiscal 2027 would still beat this year. The bar that actually tells you something comes from splitting growth into volume and price. Management expects DRAM bit shipments in calendar 2027 to grow in the low 20s percent range and NAND in the mid 20s (earnings call, 2026-09-30). That works out to roughly 5% DRAM bit growth per quarter. Next quarter's revenue guide is up 13.4%. Take out about 5% from volume, and the remaining ~8% can only come from higher prices and a richer product mix. By the same logic, if the next report guides the following quarter to a midpoint below roughly $64.6 billion (5% sequential growth), prices have stopped rising and growth is down to volume alone. Only a guide above $65 billion would show prices still going up. The one number to watch in the next report is which side of $65 billion the following quarter's revenue guide lands on. Second is whether the 86.25% gross margin holds. Management itself has called next quarter's margin the fiscal 2027 low, and if even that low gets broken, the market will turn those words against the company.

The evidence on demand is firmer than any price quote. Micron has signed 26 multi-year Strategic Customer Agreements covering more than 35% of revenue through 2030, 10 of them new this quarter. Customers have committed $32 billion to these deals, "mostly in the form of cash deposits." Remaining performance obligations under pricing frameworks total about $150 billion. And the "vast majority" of Micron's HBM bit supply for calendar 2027 has already been negotiated, at sharply higher prices than a year earlier (all from the earnings call, 2026-09-30). On when supply and demand will come back into balance, management said: "we really don't have line of sight to when supply and demand balances." How much weight that sentence deserves becomes clear in the supply chain and cycle sections.

Memory is a capital-heavy industry, so normally this section would be about a funding gap. Right now Micron is in the opposite position: it has so much cash that capital spending is almost a rounding error.

The quality of earnings holds up too. GAAP and non-GAAP net income differed by just $0.7 billion in Q4 and by $1.79 billion for the year, about 2% of non-GAAP net income, mostly stock-based compensation. For a semiconductor company that's quite restrained. In next quarter's guidance, the gap between GAAP and non-GAAP operating expenses is $250 million, consistent with that pattern. Full-year operating cash flow was 1.03 times non-GAAP net income, and about 72% of profit converted to free cash flow, which still leaves a huge cash pile after capex. On dilution, Q4 diluted shares were about 1.147 billion, and dilution from stock compensation runs within about 1% a year. Existing shareholders will barely notice it.

One piece of the cash flow needs a closer look, though. Customers' $32 billion of commitments arrive "mostly in the form of cash deposits." In accounting terms these are customer prepayments or deposits. They boost operating cash flow, but they're tied to products Micron must deliver later on agreed terms. In effect, they're down payments customers make to lock in supply. As of publication, we couldn't find the quarter's deposit inflows broken out in the press release, so we can't separate them out precisely. What we can say is that Q4 operating cash flow exceeded GAAP net income by about $6.3 billion, and part of that came from these prepayments and other working-capital changes. In an upswing this money is a bonus. In a downturn, if customers push to renegotiate prices under their agreements, it could turn into a liability that has to be negotiated.

Capital spending is the variable to watch most closely next year. Management guided to about $25 billion of capex in the first half of fiscal 2027, about $11.5 billion of it in Q1, with construction spending rising further in the second half (earnings call, 2026-09-30). Half a year's spending alone is already close to the $27.37 billion Micron spent in all of fiscal 2026. The money goes mainly to cleanrooms in Idaho and Singapore, aimed at capacity for late 2028 and beyond. On top of that, Micron has about $6.165 billion in finalized direct grants under the U.S. CHIPS Act (U.S. Department of Commerce, 2024-12), paid out as projects hit milestones. With free cash flow currently above $30 billion a quarter, Micron can fund all of this itself. The issue lies elsewhere: when the most profitable company in an industry doubles its capital spending, that spending becomes the whole industry's supply a few years later. We run those numbers in the cycle section.

As for spending only $0.65 billion on buybacks while sitting on $68.3 billion of net cash, that choice is a signal in itself. Management would rather hoard cash and expand capacity on the back of customer deposits than rush to buy back its own stock at these levels. That suggests they are more cautious about the cycle than their wording on the call lets on.

Upstream, the bottlenecks sit in three places, and none of them can be fixed quickly with money.

The first is equipment and manufacturing technology. Micron's 1-gamma DRAM process already accounts for most of its bit output, and the next generation, 1-delta, is scheduled to ramp in the second half of 2027 (earnings call, 2026-09-30). Both generations depend on EUV lithography and more complex deposition and etch steps, so equipment lead times set the pace of expansion. Management blames tight supply on "diminishing bit gains from technology migration" and "longer fab construction timelines." In plain terms, shrinking the chips squeezes out less and less extra capacity, so the only way to grow is to build new fabs, and a new fab takes two to three years from groundbreaking to first shipments.

The second is HBM packaging and the logic base die. HBM (high-bandwidth memory) stacks many layers of DRAM and sits right next to the GPU. The same wafer yields far fewer bits when made into HBM than into ordinary DRAM. That's the root cause of HBM eating into regular DRAM capacity and pushing prices up across the board. Micron is working with NVIDIA on a customized HBM4E (NV-HBM). A custom logic base makes the product harder to replace, but it also ties Micron more tightly to a single large customer's roadmap.

The third is power and data center space. Cloud providers' data centers face physical limits on power, cooling and networking, which can slow server shipments. Management expects server unit shipments to grow in the high teens percent in calendar 2026 and 2027 (earnings call, 2026-09-30). That's not explosive in itself. What really magnifies demand is how much memory goes into each AI server.

Who pays the bill is clearest in the business-unit data. The data-center-related CDBU and CMBU together brought in $34.3 billion, 63% of this quarter's revenue. Phones and PCs (MCBU) made up 24%, and automotive and embedded 13%. NVIDIA is the main buyer of HBM, and the big cloud providers are the main buyers of server DRAM and enterprise SSDs. China is expected to fall to a single-digit percentage of revenue in fiscal 2027 (earnings call, 2026-09-30). That shrinks the geopolitical exposure but raises customer concentration. Revenue share for individual customers won't be disclosed until the annual 10-K, and as of publication we hadn't found figures for this fiscal year.

The contracts need to be weighed in three layers, because not all of them are equally solid:

In other words, "sold out" means about $32 billion of prepaid cash under the strictest reading, and only reaches 75% of shipments under the loosest. When you see figures like these quoted, it's worth knowing which layer they refer to.

The weakest link is the consumer end. MCBU revenue grew this quarter while bit shipments fell, and TrendForce data explains why. Global notebook shipments are expected to fall 9.4% in 2026, and brands remain under heavy cost pressure (TrendForce, 2026-09-04). Memory has also become a much bigger share of the bill of materials for high-end phones (TrendForce, 2026-09-04). In data centers, higher memory prices can be absorbed by the returns on AI investment. In consumer electronics, they can only be absorbed by cutting specs and shipping fewer units. Once phone and PC makers start systematically cutting the memory in each device, this part of demand could shrink faster than the market expects.

Memory is an oligopoly of three players, so the closest comparisons are SK hynix and Samsung, plus SanDisk on the NAND side. Their fiscal quarters are staggered. Micron's fiscal Q4 ended in late August, while the other two companies' latest quarters ended in late June. When comparing, remember that Micron's numbers cover two more months.

All three are moving in exactly the same direction, with margins of the same magnitude. Micron's operating margin is even 6 points higher than SK hynix's, which lays to rest the old worry that Micron would miss out on the boom because it lagged in HBM share. But when comparing prices across companies, the trend in how fast prices are rising matters more. SK hynix saw DRAM ASP rise about 30% and NAND about 55% in the April–June quarter. Micron saw DRAM ASP rise in the high teens and NAND about 30% in the June–August quarter. The two companies measure differently and sell different product mixes, so you can't just subtract one from the other, but the direction is clear: prices are still rising, but the pace of increases is slowing. That's why revenue growth in section 1 dropped from 74% to 31% and then to the 13% guide.

Every memory, packaging and interconnect company in this peer group had reported by 2026-10-01. On the direction of demand, both Micron and ALAB read as up, and interconnect chipmaker CRDO's revenue growth read 114.7% (peer disclosure summary, data as of 2026-10-01). Demand across the whole AI hardware chain is still expanding.

On how the market prices this kind of report, only one peer offers a comparison this season. CRDO reported after the close on 2026-09-01 with EPS 16.1% below expectations. The stock fell 20.0% the next day, while options had implied only a ±11.0% move beforehand, so the drop was nearly double what was priced in. That shows the market has almost zero tolerance for misses among AI supply chain companies (peer earnings-reaction data, 2026-09-30). That said, CRDO makes high-speed interconnect chips, and its miss came from its own shipment timing, with no direct link to memory prices. Over the past 8 earnings reactions, CRDO and Micron moved in the same direction only 4 times, with a correlation of -0.46. With a sample of only 8 quarters and a negative sign, this reading has limited value. It doesn't show any spillover. It only shows that the market is currently punishing "great demand, but missed the numbers" very hard.

Back to Micron: options implied a ±6.3% move for this report. In the past 8 quarters, the actual move exceeded the implied move 4 times, with a median implied move of ±8.5% and a median actual move of 9.1% (peer earnings-reaction data, 2026-09-30). This time the stock rose 3.0% the next day, inside the implied range and noticeably smaller. Revenue and EPS both beat by about 6%, but guidance only matched the latest consensus, so the market priced it as "in line."

For a cyclical stock, the P/E ratio looks cheapest at the top of the cycle and most expensive at the bottom. That's well known, so the usual approach of multiplying current earnings by a multiple only partly works for Micron. A better yardstick is this: how many years of "normal" profits does the market value imply?

A forward P/E of 7 would look strangely cheap for any growth stock. For a cyclical stock, it's a clear statement. The semiconductor index has typically traded at a forward P/E somewhere between the mid-teens and low twenties in recent years, and Micron is priced at roughly a third of that. The market plainly doesn't believe $152 of annualized EPS can last. Work it backward: apply a through-the-cycle multiple of 12 to 15x, and today's market value implies "normal" EPS of roughly $73 to $91. That's right around fiscal 2026's full-year level, or about half to 60% of next quarter's run rate.

So what this market value has paid for is this: profits fall back by about half from today's peak, then stabilize there. It isn't paying for a "supercycle that lasts until 2028," and it isn't priced for "collapse next year" either. Bulls can argue that if prices keep rising for another two or three quarters, the cash piling up will make the valuation look even cheaper. Bears will point out that at the top of the 2018 cycle, Micron's P/E also dipped to just 3 or 4 times, and then profits were cut in half and the stock fell by nearly half. A low P/E on a cyclical stock has never protected against the downside. Of the 58 analysts covering the stock, 53 rate it bullish (Wall Street ratings consensus). Agreement that broad means there's limited room left for new bullish money to come in.

Memory is the purest cyclical business in semiconductors. A cycle runs roughly three to four years: demand surges, prices rise, makers add capacity, the new capacity arrives all at once two to three years later, and prices collapse. Three kinds of readings help locate where we are.

Prices are rising across the board. Over the past 180 days, DRAM price announcements included 15 increases and 0 cuts, NAND 13 increases and 3 cuts, and HBM 3 increases and 0 cuts (industry price-announcement tally, as of 2026-09-30). The latest few: AI server demand is supporting further increases in fourth-quarter memory contract prices, though pressure on the consumer side persists (TrendForce, 2026-09-30). HBM supply remains tight, and the 2027 blended ASP forecast was raised to +121% year over year (TrendForce, 2026-09-29). North American demand is keeping enterprise SSD prices rising in the fourth quarter (TrendForce, 2026-09-21). Second-quarter DRAM industry revenue rose 59.5% quarter over quarter, with supply growth still trailing demand (TrendForce, 2026-09-07).

The shortage is still severe. Micron's management says it can't see when supply and demand will balance. HBM bits for 2027 are mostly negotiated, and DRAM bit supply growth for 2027 to 2028 is only in the low 20s percent. Meanwhile gross margin is 87% and SK hynix's operating margin is 76%, with all three major makers running at full capacity. Margins like these only show up in a serious shortage.

Payback periods and new capacity are the only readings pointing the other way. Micron's capex in the first half of fiscal 2027 is about $25 billion, almost as much as all of fiscal 2026, and it's going into cleanrooms for late 2028 and beyond. SK hynix and Samsung are expanding too. At the top of every past memory cycle, manufacturers doubled capital spending while they were most profitable, and the capacity landed all at once two years later.

Putting the three together, our view is that the memory cycle is in the middle-to-late stage of its upswing. Prices are still rising and the shortage hasn't eased, but the rate of price increases has started to slow, and supply expansion is already under way. The real turning point will most likely come around 2028, when new capacity arrives, and stocks typically move two to three quarters ahead of that.

Two early warning signs of a peak worth tracking:

This section covers week-to-week signals about when to get in or out over the next few weeks. Readers planning to hold for a year or more can skip it.

Money had already started flowing back into the memory sub-sector before the report. As of 2026-10-01, the sub-sector's money-flow score was 59, in net-inflow territory, with 10 straight days of net inflows. Estimated from price and volume, net inflows were about 1.1 billion shares over the past 5 days and about 1.6 billion shares over the past 20 days (these are share counts, not dollar amounts), while the stock price rose 2.2% over the 5 days (sector money flow data, 2026-10-01). On 2026-10-01 the sub-sector was upgraded from moderately strong to strong, putting it in the "leading" quadrant on the relative-strength chart. Its parent sector, technology, is in the strongest tier, while the other ten major sectors, including materials, energy, financials, healthcare and utilities, all sit in the weakest tier (sector strength data, 2026-10-01). Money is clearly piling into tech.

The broader market backdrop is chilly, though. As of 2026-10-01, the market regime score was -25, in bear-market territory, and SPY at 763.99 was slightly below its 20-day moving average of 765.00. The 10-year Treasury yield was 5.29%, and net liquidity was $5.78 trillion, basically flat at +0.2% week over week, a level that calls for caution (market rotation data, 2026-10-01). The broad rotation signal is risk-on. The strongest pair of money moves is from energy (XLE) into technology (XLK), with a strength of 3.14. The sub-sectors drawing the most money are AI drug discovery (score 69), deposition and etch equipment (score 64) and the semiconductor index (score 63). The biggest outflows are from gaming (-85), cell tower REITs (-78) and water utilities (-75). Developed overseas markets, high-yield bonds and small caps are all being trimmed, with outflow shares of 13%, 12% and 14% respectively (market rotation data, 2026-10-01).

In a trader's terms: the overall market is weak, and money is leaving energy, defensives and small caps to crowd into one direction, semiconductors. Deposition and etch equipment scores higher on money flow than memory itself, which suggests the market prefers to bet on the "memory makers expand capacity" story. Memory isn't currently part of any established sector rotation chain. The money coming in is mostly following tech as a whole, with no independent rotation behind it.

Has the stock run ahead of the money? Micron's current price is about 15% above its 50-day moving average and about 63% above its 200-day moving average, while money has flowed into the sub-sector for just 10 days and it has been in the strong tier for just 1 day. Historically, after similar upgrades, the median return over the following 20 trading days was 0.0%, with a 58% chance of a gain (1,384 cases, covering 2021-04-13 to 2026-09-02). An upgrade by itself gives almost no directional edge. The stock has already made a big move, and the money is only now starting to catch up.

Long term (quarters to years): worth following. On fundamentals, Micron is in the strongest earnings period in its history: an 87% gross margin, $33.2 billion of free cash flow in a single quarter, $68.3 billion of net cash, and $32 billion of customer prepayments that turn part of its demand into firm commitments. Volume and price still point the same way next quarter. On valuation, a forward P/E of around 7 already builds in a scenario where profits fall by half, and the market isn't paying extra for the supercycle to continue. The costs of owning it are just as clear. This stock sits in the middle-to-late stage of a cyclical upswing, price increases are slowing, and capacity expansion has begun. Over the holding period you should expect to sit through a drawdown on the order of 30% at some point, and when the peak comes, the P/E will offer no protection at all. The way to hold it is to follow the checkpoints. With each quarterly report, check whether the following quarter's revenue guide is still above what volume alone can explain, and whether gross margin holds at the low point management promised. If both pass, keep holding. If either one misses for two quarters in a row, treat that as the first warning that the cycle is turning, and reassess the position before it's confirmed.

Short term (days to weeks): don't chase. Wait for a pullback, or for money flows to catch up. On flows, the memory sub-sector was just upgraded and has had 10 straight days of net inflows, so the direction is right. But the broad market is in bear-market territory, long-term rates are at 5.29%, liquidity is tight, and similar upgrades have historically won only 58% of the time with a median 20-day return of zero. On positioning, the stock is already about 15% above its 50-day average, and the post-earnings move stayed inside the options-implied range, meaning this report card is already priced in. With that combination, chasing the stock offers a poor risk/reward. The more sensible approach is to wait for the price to drift back toward its pre-earnings range, or for inflows to continue long enough to justify the stock's lead.

Exactly which conditions to wait for, which price levels to act at, and which numbers to check every day are laid out in the action plan below.

There are two entry conditions. Meeting either one justifies starting a small starter position. Meeting both justifies building up to the full planned position.

Condition 1 (money flows, checkable in a daily review): The memory sub-sector's streak of net inflows extends from 10 days as of 2026-10-01 to 15 or more trading days, with its money-flow score staying above 55 and not slipping back to the moderately strong tier. At the same time, either the market regime score recovers from -25 to above 0, or the 10-year Treasury yield falls from 5.29% to below 5.1%. The first part confirms the money is staying. The second confirms the broad market is no longer a headwind. If only the first part is met, use half the position size.

Condition 2 (fundamentals, tied to the next quarterly report): Fiscal 2027 Q1 revenue comes in at or above the $61.5 billion guidance midpoint, non-GAAP gross margin is at or above 86.25%, and the company guides fiscal Q2 revenue to a midpoint above $65 billion. This bar comes from the volume-versus-price breakdown in section 1. About 5% bit growth per quarter corresponds to about $64.6 billion. Only a guide above $65 billion shows prices are still rising and the cycle's upswing hasn't ended early.

Three reference price levels:

For stop-losses, short-term positions use the $952 early warning, while long-term positions use $673 as the last line of defense. Whatever sits in between is kept or cut based on how Condition 2 plays out in the next quarterly report.

Check daily:

Check weekly:

Check quarterly:

What triggers a follow-up update: either condition is met or fails; the close drops below $952 or holds above $1132; DRAM contract prices see their first quarterly cut; SK hynix or Samsung signals a clear slowdown in price increases in its quarterly report; or Micron announces a large buyback, a financing, or renegotiated customer agreements.

Catalyst calendar:

We watch the two entry conditions and all the daily, weekly and quarterly numbers listed above every day. Whenever one is met or fails, we'll write it up as a follow-up update and publish it with that day's daily review, so you don't have to monitor the data yourself. Every earnings figure, money-flow reading and industry price announcement cited here comes with its source and date, and you can check each one against the matching data page on our website. When a memory cycle reaches its middle-to-late stage, the hardest call is always when it will stop, and that kind of call is only reliable if you check it against the data quarter by quarter.

This article was produced on the NextPick research workbench; every money-flow, tier and valuation figure it cites can be verified on the corresponding data surfaces of the site, and the moment any tracking condition listed in the report triggers, an addendum will go out together with that day's daily review.