Micron stock was not celebrating on Friday morning. Nasdaq showed a premarket last sale of $1,107.30 at 5:36 a.m. Eastern on 2 October 2026, up $9.91, or 0.90%, from Thursday’s close. That print is above $1,100. It is not a 4% premarket jump, and it is not a new closing high. The fiscal fourth-quarter numbers, $54.23 billion of revenue and a $61.5 billion guide, were already public. What was new was Sanjay Mehrotra pushing the shortage past 2026.
Nothing on this page is a recommendation to buy or sell Micron, or any other stock.
Thursday was a wide day, not a gap
The cash close on Wednesday 30 September was $1,065.11, the last regular-session price before the news. Micron’s Form 8-K index on EDGAR is timestamped 20:02:22 GMT that day, after the 4:00 p.m. New York close. The call was set for 2:30 p.m. Mountain time, 4:30 p.m. in New York. Anyone who wanted to react with a market order had to wait for Thursday.
Thursday 1 October did not open like a surprise. Micron opened at $1,054.08, about 1% under the prior close, and traded as low as $1,022.90. That low was 3.96% below $1,065.11. The same session reached $1,098.90 and closed at $1,097.39, up $32.28, or 3.03%. From the open to the close the stock rose 4.1%. Volume was 45,735,430 shares, against 31,103,200 on Wednesday.
Friday’s $1,107.30 figure is a premarket last sale, with a bid of $1,107.44 and an ask of $1,108.53 at the same Nasdaq timestamp. It is not a close. The regular session had not opened. The chart stops at Thursday’s close on purpose.
From 1 April the move is the year’s, not the week’s. The 1 April close was $367.85. Thursday’s $1,097.39 is 198% above that, and still 9.6% under the 25 June closing high of $1,213.56. Nasdaq’s quote page on Friday morning listed a 52-week range of $165.50 to $1,255.00, separate from these closes. A stock this far off the spring base does not need a $54 billion quarter to re-rate. It needs a reason to think the tightness outlasts the quarter already in the model.
The quarter was the old fact
FinanceFeeds filed the results-day account on 1 October, while the stock was still around the Wednesday close: Micron after the $54.2 billion quarter and the $61.5 billion guide. That page has the bull and bear levels. This one does not repeat them.
The Exhibit 99.1 filed with the 8-K covers the fiscal fourth quarter ended 3 September 2026. That quarter-end is not the report date. Revenue was $54,229 million, against $41,456 million in the prior quarter and $11,315 million a year earlier. The year-on-year change is 379%. The highlights round the same figures to $54.23 billion and $11.32 billion. GAAP diluted earnings were $32.87. Non-GAAP diluted earnings were $33.42. GAAP gross margin was 86.8% of revenue. Non-GAAP gross margin was 87.0%. Full-year fiscal 2026 revenue was $133,188 million, which the release rounds to $133.19 billion.
The same exhibit guides fiscal first-quarter 2027 revenue to $61.5 billion, plus or minus $1.5 billion, so the published range is $60.0 billion to $63.0 billion. Non-GAAP gross margin is guided to about 86.25%, and non-GAAP diluted earnings to $38.15, plus or minus $1.00. GAAP gross margin is guided to about 85.95%, and GAAP earnings to $37.84, plus or minus $1.00. The board declared a quarterly dividend of $0.15, payable on 29 October to holders of record on 14 October. The dividend is not why the group moved.
Core Data Center revenue was $18,002 million and Cloud Memory was $16,283 million, together about 63% of the quarter. DRAM revenue was $39.8 billion, 73% of the total, and NAND was $14.1 billion. Those splits were in the release on Wednesday night. They were not new on Friday.
He said tighter twice, and he named 2028
The sentence that was not in the press release is in the prepared remarks for the 30 September call. Mehrotra, chairman and chief executive, said it early, then said a harder version in the market-outlook section.
As strong as fiscal 2026 was, we expect fiscal 2027 to be even better. Industry demand has strengthened since our last earnings call, and we expect memory and storage supply-demand conditions to be much tighter in fiscal 2027 and 2028 than they were in 2026.
Fiscal 2027 is the year Micron is in now. Fiscal 2028 is the year after. “Much tighter” than 2026 is not a claim that 2026 was loose. It is a claim that the shortage the market has been trading is worse, on the company’s comparison, across the next two fiscal years. Later he dropped the fiscal frame and used the calendar everyone else trades:
We expect memory and storage supply-demand conditions to be much tighter in calendar 2027 and 2028 than they were in 2026.
He then put numbers on the bits. For calendar 2027 and 2028 he expects industry NAND bit shipments to grow approximately in the mid-20s percent, and DRAM bit shipments in the low-20s percent, with the industry supply constrained in both years on both technologies. High-bandwidth memory bit shipments are expected to grow faster than conventional DRAM through calendar 2028. For calendar 2026 the script has DRAM bits in the mid-20s percent and NAND bits in the low-20s, Micron’s NAND supply growing less than the industry and its DRAM supply about in line.
The line that matters if a new fab announcement is being used as a reason to fade the group is still his: “Even with additional industry DRAM cleanroom space plans, with robust demand trends including new upside requests from customers, we do not have line of sight to when supply and demand will return to balance.”
The building schedule is why that is not a slogan. Idaho’s ID1 fab is on track for first wafer output in mid-calendar 2027. Initial output from the Singapore high-bandwidth-memory packaging site is expected in early calendar 2027. The first New York fab is a calendar 2030 wafer event. He also said output from a new fab becomes meaningful only a few quarters after the first wafers. A 2027 tool move does not close a shortage he has just run through calendar 2028. Most of the extra fiscal 2027 capital spending is construction, to accelerate cleanroom availability in late calendar 2028 and beyond.
Mark Murphy, the chief financial officer, put a near-term figure on the spending. Fiscal first-quarter capital expenditure, net of incentives, is projected around $11.5 billion. First-half fiscal 2027 spending is about $25 billion, and the second half is higher. That is the company’s answer to its own shortage call. The holes in the ground are for 2028 and after. The sold-out conversation is about 2027.
Why peers rose while Micron chopped around
A beat the stock already assumed did not require the rest of the memory tape to do anything. A longer shortage is an industry statement. SK Hynix, Samsung, SanDisk and Nanya do not collect Micron’s earnings. They do collect the price of a bit if Mehrotra is right that the industry is still short in 2028.
Micron’s Thursday close, up 3.03%, was not a laggard close. The chop was the path: down 3.96% at the low, then a finish 4.1% above the open, a session from $1,022.90 to $1,098.90. The peers did not need a clean Micron trend. They needed the date on the shortage to move.
SanDisk is the clean US comparison, and it is a NAND stock, which fits Micron’s $14.1 billion NAND quarter. The earlier account of that flash print is SanDisk after Micron’s NAND revenue. On Thursday SanDisk closed at $1,787.69, up $47.80, or 2.75%, from Wednesday’s $1,739.89 on Nasdaq. The call added a mix fact SanDisk holders did not have on Wednesday afternoon: data-center SSD revenue of nearly $10 billion in the quarter, more than ten times the year-ago quarter, and more than two-thirds of Micron’s NAND. That is not a SanDisk forecast.
Seoul does not offer one tick. Yahoo Finance’s regular-session series for SK Hynix, stamped at 3:30 p.m. Korea time on 2 October, shows 1,776,000 won on 30 September, 1,833,000 on 1 October and 1,841,000 on 2 October: 3.2% on Thursday, a further 0.4% on Friday. Naver’s table is several thousand won away, 1,783,000 and 1,828,000 on the first two dates, with a change field of plus 2.9% for 1 October. Direction agrees. The percentage does not, so the percents here are Yahoo’s regular closes, not an average. On either feed, Hynix rose after the call.
Samsung’s Friday price is the figure both feeds share: 276,000 won. Yahoo has Thursday already there, up 2.8% from 268,500, and Friday flat. Naver’s Thursday close is 274,500, change field plus 2.2%, and Friday 276,000. Same ending price, a small argument about Thursday. Three weeks ago the tape ran the other way: Samsung and SK Hynix slid about 5% as the KOSPI lost 7,000 after Chuseok. A two-day bounce is not a reversal of that. It is the 2027–28 calendar hitting names that had just sold off.
Nanya is the control. Taiwan’s exchange shows Nanya unchanged on 1 October at NT$519, then up NT$7, or 1.3%, to NT$526 on 2 October. Yahoo’s Taiwan series matches those closes. Nanya did not move on the first reaction day. A shortage comment is not an automatic bid under every DRAM name on the same afternoon.
Calendar 2027 HBM, and NVHBM
High-bandwidth memory is where Mehrotra was specific, and where the remarks stop short of a fantasy. He said Micron has completed agreements for the vast majority of its calendar 2027 HBM bit supply, at significant price increases from a year earlier, and that those prices narrow the gross-margin gap with conventional DRAM. HBM revenue in the quarter grew faster than company revenue. He did not give an HBM dollar total in the prepared script. This article will not invent one.
“Vast majority of calendar 2027 bit supply” is a volume statement about one product family for one calendar year. It is not a statement that fiscal 2027 revenue, across DRAM, NAND and every end market, is already sold. The 8-K and the prepared remarks do not say what portion of fiscal 2027 revenue is already committed. That figure, if it was said in the question period, is not in the script the company posted, and it is not used here.
What Murphy does say is that Micron has signed 26 strategic customer agreements, currently estimated at over 35% of revenue through 2030, and that remaining performance obligations are about $150 billion. Those obligations count only agreements with a determined pricing framework, using committed volumes and minimum prices. He called the figure inherently conservative and said revenue should well exceed it over the life of the contracts. Customer financial commitments on the 26 agreements and their extensions are $32 billion, the vast majority cash deposits. That book runs through 2030. It is not a fiscal-2027 sold-out ratio.
The product name that is new, rather than merely tight, is the one he tied to Nvidia:
We have a strong roadmap for future HBM products and are proud to be working with NVIDIA on the industry’s first custom-HBM4E implementation, NVHBM, to be adopted on next generation of GPUs and NVLink Fusion platforms.
It is a custom HBM4E part. The name in the remarks is NVHBM. It is “to be adopted” on a next generation of GPUs and on NVLink Fusion. It is not described as a chip already in a server this quarter. The HBM4 ramp, which he said is executing, is the current generation, separate from this custom program. The price-increase comment sits on the calendar 2027 supply agreements, not on the NVHBM label.
Nvidia is the customer in that sentence. Owning the memory stock and owning the GPU stock remain different bets, which is the point of Micron versus Nvidia. The competitive check is Samsung’s plan to more than double HBM4 output in 2027, reported here nine days before this print. Mehrotra still gave no date for when the market balances.
The 200GB line is a car, used as a benchmark
The other sentence that traveled was about humanoid robots and 200 gigabytes. It is real, and it is easy to overread. This is the passage, with the PDF page break removed and no words added. The apostrophe is the one in the script:
Memory content in Level 4 and higher autonomous vehicles typically exceeds 200GB, while storage content reaches multiple terabytes, each more than an order of magnitude greater than in today’s Level 2+ and Level 3 semi-autonomous vehicles. Humanoid robots are expected to have comparable memory and storage requirements to autonomous vehicles.
The 200GB figure describes Level 4 and higher cars. Memory content in that class typically exceeds 200GB, and storage is already in the terabytes. The humanoid sentence comes after, and it is a projection. Those robots are expected to need comparable memory and storage to those vehicles. Mehrotra did not say a humanoid on sale today carries 200GB of Micron DRAM. He did not read a bill of materials. He benchmarked a future machine on a car class he says is already above 200GB. “Typically exceeds” is not a minimum spec from a single platform.
The clock he put on it is the end of the decade, not fiscal 2027. Autonomous vehicles are “the first major deployment of physical AI,” and humanoids are where he thinks that goes next. Physical AI “can become a significant driver” by then if units and content both rise. Customers are sampling. Sampling is not an order, and 200GB times a robot forecast is a model the call did not provide.
The margin dip is what the CFO called the floor
The results-day objection was gross margin. Non-GAAP gross margin of about 86.25% for the fiscal first quarter is below the 87.0% just reported. Three-quarters of a point, at these levels, is a lot of dollars and a small amount of drama, and memory investors treat the direction as the cycle. The call does not deny the dip. Murphy dates it, and the cause he gives is not a rollover in the price of a bit.
We anticipate fiscal Q1 to be the floor for gross margins in fiscal 2027.
The mechanism is incentive pay. Micron raised fiscal 2026 incentives in the fourth quarter, including in manufacturing. Most of the manufacturing piece went into inventory. Those higher-cost chips are what get sold in the first quarter, which is what he says hits the margin. The second quarter has less of that fourth-quarter compensation left in cost, but higher fiscal 2027 incentive plans offset the relief. After the first quarter he expects higher gross margins for the rest of fiscal 2027, “with a more moderate rate of price increases.”
That is a forecast, not a turn that has already printed. Next to it: record revenue of $61.5 billion plus or minus $1.5 billion, and sequential revenue growth each quarter of fiscal 2027. Operating expenses, he projected, rise about $2.5 billion in fiscal 2027, mostly research and development and the higher incentives. The margin path and the revenue path can both be wrong.
Prices, on his account, are still rising, more slowly. In the quarter just reported, DRAM prices rose by a high-teens percentage and NAND prices by about 30%, while DRAM bits rose by a mid-single-digit percentage and NAND bits by about 10%. The revenue jump was price more than volume. If revenue keeps rising while price increases moderate, volume has to do more of the work.
What would make Friday’s quote look early
None of Friday’s figure is a close. The $1,107.30 sale was premarket at 5:36 a.m. Eastern, still under the 25 June closing high, and premarket quotes move before the open. A later close through $1,213.56 would say the longer shortage was not in the price. A close back through Thursday’s $1,022.90 low would say the peer move was a one-day reading of a script. Neither had happened at that timestamp.
The checks that come from the call are plainer. Whether fiscal first-quarter gross margin holds near the guided 86.25%, and whether the next quarter is actually higher, as a “floor” requires. Whether the calendar 2027 HBM agreements stay in place at the higher prices, or whether “vast majority” gets qualified later. Whether Samsung’s HBM4 additions, and Hynix shipments into the same Nvidia platforms, show up as a looser bit market before the “no line of sight” sentence gets a date. He did not put an expiry on it.
Server units are the demand line he quantified. He expects high-teens percent unit growth in both calendar 2026 and calendar 2027, with memory content per server growing more slowly than Micron used to expect, because supply is tight. Units up, content per box held back by the shortage itself. If those units slip, the 2027 and 2028 bit growth is the number that breaks.
Thursday’s close was $1,097.39. The $54.23 billion quarter was already public. The new sentences were the calendar, the NVHBM name, and a 200GB figure that belongs to Level 4 cars rather than to a robot on sale.