Applied Materials did not fall 35 percent from its June high because its business got worse. It fell because the market worked out how much of this year’s profit growth was not profit growth. AMAT closed at $468.85 on 9 September 2026, down 0.83 percent on the day and 35.2 percent below the $723.00 closing high of 30 June, on a market capitalisation of $372.1 billion (Nasdaq). The company’s fiscal third quarter, reported on 13 August, was a genuine record: revenue of $9.115 billion, up 25 percent, with GAAP earnings per share up 43 percent. But the Form 10-Q filed a week later, on 20 August, contains the number that explains the de-rating — and almost nobody has written about it. Applied’s effective tax rate for the first nine months of fiscal 2026 was 12.9 percent, against 27.2 percent a year earlier.
That single line does more work than the entire investment-gains argument the sell-side has been having. Run the arithmetic from the filing. Nine-month income before income taxes rose from $7,004 million to $8,460 million — up 20.8 percent. Nine-month net income rose from $5,101 million to $7,370 million — up 44.5 percent. The gap is the tax provision, which fell by $813 million in absolute dollars while pre-tax income rose by $1,456 million. Add the year-over-year swing in gains on strategic investments — $909 million this year against $270 million last year, per the cash flow statement — and $1,452 million of the $2,269 million increase in net income, 64 percent of it, came from the tax line and from marking a venture portfolio rather than from selling more equipment. The operating engine tells the honest story: Semiconductor Systems segment operating income rose 13 percent over the nine months, from $5,479 million to $6,176 million. Thirteen percent, against a headline of forty-five. That is the whole re-rating in one comparison, and the reason a bull case of $690 and a bear case of $340 are both defensible from the same filings.
Key facts
- Spot $468.85 (close, 9 September 2026), −$3.94 on the day; 52-week range $161.75–$739.67 — Nasdaq
- Nine-month effective tax rate 12.9 percent vs 27.2 percent a year earlier; the tax provision fell $813m while pre-tax income rose $1,456m — Form 10-Q, 20 August 2026
- Nine-month share repurchases $1,137 million, down from $4,037 million — a 71.8 percent cut, with $12.8 billion still authorised — 10-Q, Note on stock repurchases
- Accounts receivable $7,691 million at 26 July 2026 against $5,185 million at 26 October 2025, a $2,506 million build — 10-Q balance sheet
- Nine-month non-GAAP free cash flow $3,580 million, down 2.1 percent, while capital expenditure rose 34.8 percent to $1,988 million — Q3 FY2026 earnings release, 13 August 2026
- NAND fell to 7 percent of Semiconductor Systems revenue from 9 percent, the smallest slice in the mix — 10-Q, revenue by market
- Spot trades at 29.2 times the annualised run rate of the Q4 guide ($4.02 × 4 = $16.08), against 45.0 times at the June closing high — FinanceFeeds calculation
What the quarter actually delivered
Take the print on its own terms first, because operationally it was excellent. Revenue of $9,115 million against $7,302 million a year earlier. GAAP operating income of $3,075 million was 33.7 percent of revenue against 30.6 percent, and cash from operations was a record $3,037 million. Semiconductor Systems, the equipment business that is 77 percent of the company, produced a 55.3 percent gross margin and a 37.7 percent operating margin, with segment operating income up 45 percent to $2,657 million.
The guide was the friction point. Applied told the market to expect fourth-quarter revenue of $10,250 million plus or minus $500 million and non-GAAP diluted EPS of $4.02 plus or minus $0.20 — not weak, simply short of a bar peers reporting earlier in the season had already set. In a crowded sector, being merely very good triggered a rotation. We have tracked the same reflex in our work on TSMC’s bull and bear case and on how large allocators position across semiconductor equities.
What has changed since is the tape, not the fundamentals. On 29 July the shares closed at $436.45, the trough of the June–July drawdown; on 3 September they closed at $435.91, a new closing low, meaning the floor most desks were watching has already been broken once. The two sessions that followed produced gains of 4.31 and 3.98 percent. Sixty-session realised volatility runs at 78.8 percent annualised against 59.1 percent over the full year, on our calculation from Nasdaq closes.
Management’s framing has not wavered. “As the rapid global adoption of AI drives unprecedented demand for our materials engineering solutions, we are further raising our Semiconductor Systems revenue expectations for calendar 2026 and are confident we will grow faster than the market this year,” said Gary Dickerson, President and CEO, in the earnings release filed with the SEC. “Based on the increased demand visibility we are receiving from our customers, we expect another strong growth year for Applied Materials in 2027.”
The capital allocation flip nobody flagged
Here is the disclosure that should have moved the stock more than the guide did, and it appeared only in the 10-Q, two days after most of the coverage was written. Applied repurchased $1,137 million of its own stock in the first nine months of fiscal 2026, against $4,037 million a year earlier — a 71.8 percent cut, three million shares against twenty-five million, executed in a year when the stock spent most of the summer more than thirty percent below its high. It is not short of authorisation: approximately $12.8 billion remained available as of 26 July 2026. At the third quarter’s pace of roughly $400 million, that would take eight years to exhaust.
Where did the money go instead? Capital expenditure, which rose 34.8 percent to $1,988 million for the nine months. In fiscal 2025 Applied spent $1,475 million on capex and $4,037 million on buybacks — a ratio of 0.37 to one. This year it has spent $1,988 million on capex and $1,137 million on buybacks — 1.75 to one. The company has flipped from returning capital to building capacity, and it did so in the same year the multiple compressed. Dividends went the other way, rising to $1.52 per share declared for the nine months from $1.32.
Brice Hill, Senior Vice President and CFO, said as much in the earnings release: “We expect continued strong revenue growth in the second half of the calendar year, particularly in DRAM as well as leading-edge foundry-logic and advanced packaging. Looking further ahead, we are making additional manufacturing capacity investments to support projected demand through the end of the decade.”
That is a defensible decision for a capital-goods company facing a real demand cycle, but it has a price, and the price shows up in working capital. Accounts receivable stood at $7,691 million at 26 July 2026 against $5,185 million at the 26 October 2025 year end — a $2,506 million build, while nine-month revenue rose $2,469 million year over year. In absolute dollars, the receivables balance grew by more than revenue did. Receivables consumed $2,508 million of operating cash this year against $538 million last year, and free cash flow conversion fell from 71.7 percent to 48.6 percent.
The NAND option the mix is hiding
The consensus reading of Applied’s mix is that DRAM is the risk. Within Semiconductor Systems, DRAM rose to 26 percent of revenue from 22 percent in the quarter and to 29 percent from 25 percent over nine months, while foundry, logic and other slipped to 67 percent from 69 percent. Memory capex is the most violent line item in semiconductors, and a rising memory share makes the guide less durable than the headline implies — the same cycle risk we set out in our Micron bull and bear analysis.
But look at the other memory line. Flash memory — NAND — fell to 7 percent of Semiconductor Systems revenue from 9 percent in the quarter, and to 6 percent from 7 percent over the nine months. On $7,040 million of segment revenue, that is roughly $493 million in the quarter, or about $2.0 billion annualised. It is the smallest slice in the mix and it is still shrinking.
Now set that against the pricing signal. TrendForce forecast NAND flash contract prices rising 70 to 75 percent quarter over quarter in Q2 2026, after an 85 to 90 percent forecast in the first quarter, with enterprise SSDs absorbing 48 percent of global NAND bits against 26 percent a year earlier. Capacity relief, as that reporting made clear, is years away — because NAND makers cut capital spending through the downturn and have been diverting resources toward the more profitable DRAM market.
Equipment orders lag price recovery; they do not lead it. Applied’s NAND exposure therefore sits at a cycle low precisely when the NAND price signal is at a cycle high, and nothing in the guide or the mix table assumes it recovers. That asymmetry is the cleanest unpriced option in this stock: if capacity additions restart in calendar 2027, the 7 percent line has more percentage headroom than any other bucket in the mix, at incremental margins on a segment already running 55.3 percent gross. The same dynamic runs through the NAND makers, which is why SanDisk’s own bull and bear case hinges on whether tightness lasts long enough to justify new fabs.
Regulatory tension: the suspended denial order
The China story is more settled than the headlines suggest, and the export-control story is less settled than anyone admits.
On revenue, the adjustment has largely happened. China was $2,506 million in the quarter, 28 percent of revenue, down 2 percent year over year and down from 35 percent of the total — falling as a share purely because everything else grew. US revenue doubled to $1,367 million, Europe tripled to $483 million, Southeast Asia rose 92 percent. Over nine months, Korea ($4,551 million) and Taiwan ($5,902 million) together are 44 percent of revenue against China’s 28 percent, which makes the Korean memory cycle a bigger swing factor for Applied than Chinese policy is. The China exposure is real but has already stopped contributing growth — a dynamic visible from the other side in our reporting on CXMT’s HBM yields and the 2027 memory gap.
The under-covered risk is not revenue. It is the suspended denial order. Applied disclosed in the 10-Q that its February 2026 settlement with the U.S. Commerce Department’s Bureau of Industry and Security — under which it paid $253 million during the second fiscal quarter to resolve an inquiry into certain China customer shipments — “includes a denial order that is suspended and will be waived three years after the date of the order,” conditional on compliance. Applied is also required to conduct internal audits of its export-control compliance programme and maintain training and reporting mechanisms.
Read that carefully. For three years, a company whose entire business is shipping controlled technology across borders operates with a denial order suspended over its export privileges. The financial charge is behind it; the conditionality is not. That is a low-probability, high-severity contingency carrying no line in anyone’s model, and it does not lapse until 2029. Applied’s customers face their own version of the same policy risk, which is why the capacity build at Intel and at subsidised US fabs matters more to the 2027 order book than any single rule change.
Bull case: $690
The bull case is $690, some 47 percent above the 9 September close and — importantly — 4.6 percent below the $723.00 closing high the shares printed ten weeks ago. It does not require a new record. It requires the market to pay again what it was paying in June.
The build: guided fourth-quarter non-GAAP EPS of $4.02 annualises to $16.08. Dickerson has said 2027 will be “another strong growth year” and Applied has raised its advanced-packaging growth outlook for calendar 2026 to more than 70 percent. Twenty percent growth on the guided run rate puts fiscal 2027 non-GAAP EPS near $19.30. At $690 the shares would trade at 35.8 times that number, and 42.9 times the current run rate — rich, but below the 45.0 times the market willingly paid on 30 June.
Three conditions have to hold: capital expenditure has to peak, letting free cash flow converge back toward net income; the receivables balance has to stop growing faster than revenue; and the NAND line has to turn, the only part of the mix where the upside is genuinely unmodelled. Get all three and the cash flow the market is refusing to capitalise arrives at once. Nasdaq’s compiled one-year consensus target sits at $650, so $690 is a stretch case rather than a fantasy.
Bear case: $340
The bear case is $340, roughly 27.5 percent below spot — notably less than the 35.2 percent decline the shares have already suffered since June. It is a moderate scenario, not a catastrophe, and two independent routes land on it.
The first is the multiple. At $340 the shares trade at 21.1 times the annualised guided run rate of $16.08 — a mid-cycle multiple for a capital-equipment maker, not a distressed one. That price sits between the 2 January 2026 close of $268.87 and the 27 February close of $372.30, meaning most but not all of this year’s re-rating unwinds.
The second route is the tax, and it is the more interesting one. Apply last year’s 27.2 percent nine-month effective rate to this year’s $8,460 million of pre-tax income and net income becomes about $6,159 million — diluted EPS of roughly $7.71 rather than the $9.22 reported, a 16.4 percent haircut. Do the same to the guided run rate and $16.08 becomes about $13.44. At 25.3 times a tax-normalised earnings base, you get $340. The two methods agree, which is what makes the level worth taking seriously.
The path needs nothing exotic. A deferred-tax remeasurement is by nature a catch-up, not a permanent run-rate gift. Applied’s own non-GAAP reconciliation already shows a $220 million unrealised loss on strategic investments in the third quarter, so the mark-to-market tailwind has begun to reverse. And the conversion gap does not have to widen to do damage — it merely has to persist two or three more quarters for the earnings-quality view to become consensus. Semiconductor names reprice fast when that happens; our Wolfspeed analysis is a study in how little it takes.
Invalidation and what happens next
Invalidation level: a weekly close below $435.91, the 3 September 2026 closing low. That level has now been tested twice — $436.45 on 29 July and $435.91 on 3 September — and it is the floor separating the current range from the bear zone. A decisive weekly break below it retires the base case and puts $340 in play. In the other direction, the bear case is invalidated on a close above $548.15, the 12 August high and the last price at which this stock traded before the earnings-quality debate began.
Three expectations for the fourth-quarter print. First, the revenue number should land: the $10.25 billion guide reflects visibility management describes as customer-driven, and the ramp is funded. Second, the lines that move the stock will be neither revenue nor EPS but the effective tax rate and the receivables balance — if the rate normalises toward the mid-twenties, the headline EPS growth rate collapses on its own arithmetic. Third, watch the buyback. If Applied steps up repurchases from the $400 million quarterly pace against $12.8 billion of authorisation, that is management signalling the capex peak is in sight. If it does not, the capital build has further to run — and so does the discount.
Our base case is $520, or 32.3 times the guided run rate: the multiple the market was paying immediately after the August print, before the September flush.
Frequently asked questions
Why did AMAT fall 35 percent from its June high after a record quarter?
Because the composition of the profit growth changed the multiple investors were willing to pay. Nine-month net income rose 44.5 percent, but 64 percent of that increase came from an $813 million drop in the tax provision and a $639 million swing in gains on strategic investments. Semiconductor Systems segment operating income — the operating engine — grew 13 percent. Free cash flow fell 2.1 percent.
How big is the tax effect on AMAT’s reported earnings?
Applied’s nine-month effective tax rate was 12.9 percent against 27.2 percent a year earlier, driven primarily by a remeasurement of deferred tax assets arising from new tax incentive agreements in Singapore. Applying the prior-year rate to this year’s pre-tax income cuts nine-month diluted EPS from $9.22 to roughly $7.71 — a 16.4 percent difference, on FinanceFeeds arithmetic from the filed 10-Q.
Why has Applied Materials cut its share buyback?
Nine-month repurchases fell to $1,137 million from $4,037 million, a 71.8 percent cut, while capital expenditure rose 34.8 percent to $1,988 million. Approximately $12.8 billion remains authorised. The company has redirected cash from returning capital to building manufacturing capacity that the CFO says supports demand “through the end of the decade.”
Does the NAND price recovery help Applied Materials?
Eventually, and it is currently unpriced. NAND is only 7 percent of Semiconductor Systems revenue, down from 9 percent, even as TrendForce forecast NAND contract prices up 70 to 75 percent in Q2 2026. Equipment orders lag pricing, so a 2027 restart in NAND capacity additions would hit the smallest and most depressed line in Applied’s mix.
What are the bull and bear cases for AMAT stock?
The bull case is $690, which requires fiscal 2027 non-GAAP EPS near $19.30 and a 35.8 times multiple on it — below the 45 times paid at the June high. The bear case is $340, or 21.1 times the guided run rate, equivalently 25.3 times a tax-normalised base. The base case is $520. The invalidation level is a weekly close below $435.91.
Prices are as at the close on 9 September 2026, when AMAT last traded at $468.85. Financial figures are from Applied Materials’ fiscal Q3 2026 earnings release dated 13 August 2026 and its Form 10-Q for the quarter ended 26 July 2026, both filed with the SEC; price and market-capitalisation data are from Nasdaq. Bull, base and bear cases are analytical scenarios constructed by FinanceFeeds, not price targets, and nothing here is investment advice.