US Pre-Market Brief | 2026-08-26 (ET) Wednesday

• Related news: FY2027 Q2 results today at 16:20 ET, call at 17:00 ET (Kiplinger live page²). In-quarter consensus: revenue ≈ $92 billion, non-GAAP EPS $2.08–2.09, data-center revenue ≈ $85.4–85.7 billion; company guidance itself was $91 billion ±2% (Q1 FY27 press release¹⁸).

• Catalyst logic: the test is in the guidance, and the "in-quarter beat" carries almost no information. Consensus of $92 billion is only 1.29% above the midpoint of company guidance, while NVDA's actual beats versus guidance over the last four quarters were +3.87% / +5.57% / +4.81% / +4.63% (mean +4.72%, and even the worst was +3.87%). Extrapolating the same pattern, revenue of $94.5–96.1 billion this quarter is "business as usual" — a beat on the revenue line requires essentially no new good news, only that the company not do worse than its own worst quarter of the last four.

• 🔑 The real bar is next-quarter guidance, and it only means something once converted to a QoQ rate. FY27Q3 revenue consensus is reported at $103.5–104.2 billion across sources (Investing.com¹⁹ reports $103.96 billion; Yahoo's aggregation of 43 analysts gives $103.896 billion). But the absolute number is deceptive — the QoQ growth implied by NVDA's last four guides was +15.5% / +14.0% / +14.5% / +11.5%; if the current quarter comes in at $95 billion, then:

  Next-quarter guidance	Implied QoQ (based on Q2 actual of $95 billion)	Characterization
  $104.2 billion	+9.7%	The slowest of five quarters
  $106 billion	+11.6%	Barely matches last quarter's pace
  $108 billion	+13.7%	Genuinely "strong"
  $110 billion	+15.8%	Re-accelerating

  Table scrolls horizontally

  In other words: if the current quarter prints $95 billion, then "in-line guidance of $104.2 billion" is actually a deceleration on a QoQ basis. Look at the Q2 actual first, then do the division; don't read the headline absolute number.

• ⚠️ One intuition must be corrected: NVDA itself is not crowded — its supply chain is. The chain rallied together yesterday (AMD +4.91%, SMCI +9.35%, MRVL +4.84%, LITE +6.67%, SMH +1.65%), but pull the lens back to year-to-date:

  NVDA	SMH	AMD	MU	TSM	Nasdaq 100	S&P 500	AVGO
  +14.24%	+54.34%	+123.75%	+226.89%	+37.36%	+15.68%	+12.15%	+3.07%

  Table scrolls horizontally

  NVDA lags the semiconductor ETF by roughly 40 percentage points year-to-date, lags the Nasdaq 100, is only 2pp ahead of the S&P, and still sits −9.9% below its 52-week high. This is not a stock that money has chased for eight months; it is a stock that has been left alone for eight months. VIX at 15.68 is low and implied move is ~5.9% — the market is neither euphoric nor fearful about it. "Sell the news" presupposes crowded positioning, and that premise does not hold for NVDA itself (it does hold for MU/AMD).

Fundamental check (data from SEC 10-Q / 10-K / 8-K originals)

• Growth is accelerating, not decelerating (+55.6%→+85.2%, four straight quarters higher); data center in Q1 FY27 was $75.25 billion, +92% y/y, with Hyperscale +115%. The old-basis breakdown shows networking +199%, far faster than compute at +77% — consistent with the sell-side view that networking is growing faster than compute.

• China is already a zero base: Q1 FY27 data-center Hopper shipments to China were zero (versus $4.6 billion a year earlier), and Q2 guidance likewise contains no data-center compute revenue from China. All current growth rates are "ex-China" — no upside option from policy easing is embedded, and the marginal impact of further restrictions is close to zero.

• 🔴 Valuation-basis warning: the TTM P/E of 32.63x is flattered. In the Q1 FY27 income statement, "Other income (expense), net" was +$15.93 billion, against operating income of $53.54 billion in the same period. Investment gains over the last four quarters total $25.03 billion, or 15.7% of TTM GAAP net income. Stripping that out at a 16% tax rate gives an adjusted TTM P/E ≈ 37.6x (TTM non-GAAP P/E is 36.7x). Citing 32.63x to argue "cheap" is wrong. And those gains come from NVDA's equity investments in the AI ecosystem (non-public equity holdings grew from $3.24 billion to $43.36 billion in a year), which overlap heavily with its customer base — this profit line moves in the same direction as end demand and provides no cushion. ⚠️ But this must not be described as "two independent paths corroborating each other" (the prior version said so; corrected): NVDA's own non-GAAP definition already excludes "gains/losses from equity securities, net", the very same item being manually stripped out, so 36.7x and 37.6x being close is a structural necessity, not independent validation. The only thing worth remembering is this: using GAAP TTM 32.6x to argue "cheap" distorts the picture.

• ✅ But the forward P/E genuinely is at a three-year low: NTM forward P/E of 21.22x is the lowest of 12 quarter-end observations over the past 3 years (range 21.22–41.66x, median 31.95x), a 33.6% discount to the median; PEG 0.48. What is "expensive" is the absolute market cap ($5.16 trillion), not the multiple.

• Cash flow and balance sheet: TTM operating cash flow $125.65 billion, free cash flow $119.08 billion (FCF margin 47.0%); net cash $41.87 billion, debt/equity 0.07. ⚠️ The single-quarter OCF includes a $7.97 billion increase in income taxes payable, a tax-timing difference; excluding it, the figure is about $42.38 billion and cannot be extrapolated.

• 🔑 The hardest forward-looking signal is purchase commitments: manufacturing/supply/capacity commitments rose from $95.2 billion in 2026-01 to $119 billion in 2026-04 (QoQ +25.0%), of which $95 billion is payable during the remainder of FY2027, roughly 30% of consensus revenue for the rest of the year. This is management betting cash on the supply side — harder evidence than verbal guidance. ⚠️ Basis caveat: starting with the FY26 10-K, this line was renamed from "inventory purchase and long-term supply and capacity obligations" to "manufacturing, supply and capacity commitments," with the wording broadened from "contract manufacturers" to "suppliers," so the YoY +299% partly reflects a widened definition; QoQ +25.0% is the clean read.

• Inventory: $25.80 billion, +127.6% y/y, faster than revenue at +85.2%; but days on hand have been stable at 105–119 days for four straight quarters and were nearly flat sequentially (114.3→114.7), with raw materials at $6.647 billion, +74.6% QoQ — a raw-material-only surge is the classic signature of "stocking up for a new architecture," not unsold finished goods. Falsification condition: if days on hand break above roughly 125 days this quarter, or if sequential inventory growth materially exceeds the sequential revenue guide (+11.5%), the "stocking up" narrative fails. The company already took $1.1 billion of inventory-related charges this quarter, a 1.2pp drag on gross margin.

• 🔴 Gross margin has clearly topped, and this is the one number tonight where only one direction carries new information:

  	Q1FY26	Q2FY26	Q3FY26	Q4FY26	Q1FY27	Q2FY27 guidance
  GAAP gross margin	60.5%	72.4%	73.4%	75.0%	74.9%	74.9%
  Non-GAAP gross margin	60.8%	—	—	75.1%	75.0%	75.0%

  Table scrolls horizontally

  Non-GAAP gross margin has been pinned at 75.0–75.1% for three straight quarters, and the company's own Q2 guide is likewise 75.0% — zero improvement. Revenue is still +85% y/y while gross margin does not budge, meaning the incremental revenue no longer carries a higher marginal gross margin than the existing base. And the 10-Q states verbatim: "We expect our Rubin platform to start shipping in the second half of fiscal year 2027" — that is, the next two quarters starting now. The last architecture ramp (Blackwell) pushed gross margin down roughly 300bp and took three quarters to climb back.

• There is one piece of evidence on cost pressure the market has already voted on with prices: year-to-date, MU +226.89% versus NVDA +14.24%. In a world where "NVDA is the only winner," that ordering could not happen. HBM already accounts for 30–40% of an AI accelerator's manufacturing cost (under 20% two generations ago), and Rubin uses HBM4. The rent in the AI value chain is migrating from design to memory.

• ⚠️ Customer concentration and circular financing (the most substantive risk item on NVDA in this brief): the 10-Q "Direct Customers" disclosure shows that in Q1 FY27 the direct customers reaching 10% were 21% / 17% / 16%, with the top three totaling 54% (in Q3 FY26 it was briefly four customers totaling 61%). The same 10-Q also states: "one AI research and deployment company contributed to a meaningful amount of our revenue by purchasing cloud services from our customers" — indirect demand is further concentrated in a single unnamed AI company.

  More important is the 8-K dated 2026-08-17²⁰: NVDA issued a residual value guarantee to SB Energy covering roughly 4.25GW of IT load at the Portsmouth, Ohio campus, with cumulative payment obligations capped at $105 billion, classified under Item 2.03 "off-balance-sheet obligations," with leases expected to take effect progressively from 2028; the tenant is an OpenAI affiliate, and the trigger condition is precisely "OpenAI bankruptcy or non-payment of rent." Over the same period on the balance sheet, cash outflow for purchases of non-marketable securities was $18.582 billion this quarter versus only $649 million a year earlier (28x). This is not an accusation of fraud — but it does mean part of the demand was bought with the company's own money, and this risk moves in the same direction as end demand, providing no diversification.

• One piece of counter-evidence (unhelpful to bears): days sales outstanding improved from 51.4 days to 45.4 days. The classic "stuffing the channel to prop up revenue" objection is directly falsified by the data.

• ⚠️ An accounting-basis trap with the same origin as INTU's but far smaller in magnitude (a miscalculation in the prior version has been corrected here): NVDA also began including SBC in non-GAAP starting FY27 Q1, but the impact is far smaller than INTU's — Q1 SBC of $1.928 billion ÷ roughly 24.4 billion diluted shares = about $0.079/share (pre-tax), about $0.067/share (after tax), only about 4% of the $2.09 consensus (not the $0.35 / 17% written in the prior version of this brief, which mistook an annualized figure for a single-quarter one). More importantly: NVDA has already restated its historical series — the press release states "The historical non-GAAP financial information presented has been updated to include stock-based compensation expense" and provides restated figures of Q1 FY26 $0.78 and Q4 FY26 $1.59. So historical non-GAAP EPS is comparable across periods; cite the restated values directly and stop using the old-basis $1.05 / $1.30. This is the opposite of INTU's situation: INTU did not restate history and sell-side consensus is still on the old basis, which is what produced that −39% illusion.

• Pre-market and technicals: −0.10% on cumulative volume of 1.77 million shares (one of the largest pre-market volumes in the market, so the quote is credible). At $213.05 it is −9.9% from the 52-week high of $236.54 and sits at the 67.6th percentile of the 52-week range; 50-day MA 207.81 / 200-day MA 195.43 / RSI 49.61.

• IV and historical patterns (the most executable passage in this brief): options imply a single-day move of about ±5.4%~5.9% (the 8/28-expiry K=212.5 straddle measured $12.61 = 5.92%, which at a 0.85 factor works out to about 5.03%; ORATS's basis is ±5.4%, about $280 billion of market-cap swing).

  ⚠️ A widely circulated claim needs correcting here. The media often cite "an average realized move of ±7.4% over the past 12 quarters, so options are underpricing." But recomputing quarter by quarter from actual OHLC, the near-term sample does not support that conclusion at all: over the last 8 quarters the absolute day-after move averaged just 3.73%, and widening the sample to 9 quarters gives only 4.35%; in only 3 of the 8 quarters did the realized move exceed 5.03%. That ±7.4% figure includes the high-volatility 2023–2024 period and no longer applies today. On the near-term sample, options are expensive rather than cheap — buying the straddle has had a negative expected value over the past two years.

  The directional statistics are harder still:

  Statistic (last 8 quarters)	Value
  Next-day close change	2 up / 6 down, mean −2.78%; last 4 quarters 0 up / 4 down
  Next-day gap (prior close → next open)	Mean +1.35% (positive)
  🔴 Next-day "open → close"	Negative all 8 times, mean −4.04%
  T+5 cumulative	7/8 negative, mean −5.83%

  Table scrolls horizontally

  (Widening the sample to 9 quarters, "open → close" is negative 8/9 with a mean of −3.46% — the direction of the conclusion is unchanged.)

  Typical pattern: on 2025-02-27 the next day opened +2.83% but closed −8.48% (open→close −11.00%); on 2025-11-20 it opened +5.06% but closed −3.15% (open→close −7.81%). "Gap positive, intraday all negative" literally means overnight good news gets systematically sold after the open, with no exception in 8 quarters.

• And it is not because there was "no beat": NVDA has beaten for many consecutive quarters, but the size of the beat is decaying — EPS surprise has been compressed from +29.35% in 2023-08 down to +3.46% ~ +5.54% in recent quarters. When the probability of a beat approaches 100% and the size of the beat is stable in the single digits, the information content of the beat itself approaches zero. Pricing runs on variance, not on the number of beats.

• Quantified reference (base price $213.05):

  Scenario	Day-after move by historical analogy	Corresponding price range (historical analogy, not a target price and not advice)
  Q3 guidance ≥$108 billion (QoQ ≥+13.7%) + gross margin ≥75.0%	+3% ~ +5.4%	$220–225
  Q3 guidance $103.5–106 billion (in line but decelerating QoQ) + gross margin 75.0%	−1% ~ −3%	$207–211
  Q3 guidance $100–103.5 billion or gross margin guidance ≤74.5%	−4% ~ −5.5%	$201–205
  Q3 guidance <$100 billion and gross margin ≤73.5%	−5.5% ~ −8.5%	$195–201

  Table scrolls horizontally

  Note the second row: even "fully in line with consensus" has historically produced a small decline — 2 up / 6 down over the last 8 quarters and 0 up / 4 down over the last 4. The options-implied ±5.4%~5.9% falls entirely inside the 52-week range; the market is not pricing either a new high or a breakdown.

• Final judgment: Do not participate pre-market, but do not preset a direction using the "crowded stock dies on the news" template. The information only appears after the close; any pre-market position in either direction is pure speculation.

  • Both sides are using the same set of numbers: $119 billion of purchase commitments, the $105 billion guarantee, the 4.25GW contract, ACIE +31% QoQ — one side reads them as "a bet on the future," the other as "a stand-in for customer credit."
  • And gross margin is one of the few variables in this brief where the two interpretations give opposite predictions: genuine demand expansion should let 75% hold or even move up; demand self-funded off the balance sheet would first show up in gross margin via HBM4 costs and the Rubin ramp.
  • Fundamentals can support the current multiple (forward P/E of 21.22x is a three-year low), but only under the assumption that the FY27 consensus of $395.7 billion in revenue / $9.02 EPS is delivered. What is fragile is not the valuation multiple; it is the growth curve the valuation depends on.

This is the stock most likely to be misrepresented by the media today; it has to be unpacked.

• The media framing: FY27 Q1 EPS guidance of $2.44 vs. consensus $4.02, 39% below; the stock closed −3.37% yesterday and is −9.36% pre-market this morning (422,000 shares, quote credible).

• But the −39% does not hold, as verified word for word against the SEC original. In the same press release (8-K EX-99.01,2026-08-25²¹) the company states: "Effective August 1, 2026, share-based compensation expense will no longer be excluded from Intuit's non-GAAP financial measures." Sell-side consensus is still built on the old basis (SBC excluded). Comparing new-basis guidance against old-basis consensus is comparing net profit to gross profit.

• Restatement (the company itself provides the impact in footnote (1) of its guidance table; the original has been checked): full-year non-GAAP EPS includes an SBC impact of $5.81/share (corresponding to $2.020 billion of operating income), and Q1 includes $1.48/share (corresponding to $521 million).

  • ⚠️ $5.81 is an after-tax per-share figure, not pre-tax. Check: $2.020 billion × (1−21%) ÷ 277 million shares = $5.76, within 0.9% of the disclosed value — consistent. Adding it back as if it were pre-tax ($2.020 billion ÷ 277 million shares = $7.35) would restate FY27 to $30.2–30.5, overstating it by 5%. This is the easiest mistake to make when restating.
  • FY27 full year: $22.88–23.12 + $5.81 = $28.69–28.93 vs. old-basis consensus of about $27.30 → roughly 5.1%~6.0% above, not below.
  • Q1: $2.44–2.48 + $1.48 = $3.92–3.96 vs. $4.02 → only 1.5%~2.5% below.

• And FY26 Q4 was in fact a double beat: revenue $4.354 billion (+13.65%), about 1.8%~2.0% above consensus; non-GAAP EPS (old basis) $4.03 vs. consensus $3.54–3.58, a beat of 12.6%~13.8%. GAAP EPS was only $1.34 because the quarter included $293 million of restructuring charges (equivalent to $1.08/share), which is excluded from non-GAAP.

• 🔴 Debunking the accounting basis ≠ nothing is wrong here — the real negative is revenue, and the revenue basis is completely unaffected by SBC:

  Segment	FY27 guided growth	FY26 actual growth	Deceleration
  Total revenue	+8.5% ~ +9.6%	+13.90%	−4.3 ~ −5.4pp
  Global Business Solutions (excludes Mailchimp)	+13% ~ 14%	+16% (includes Mailchimp)	Bases not comparable; reference only
  TurboTax	+2% ~ 3%	+7%	−4 ~ −5pp
  Credit Karma	+11% ~ 13%	+20%	−7 ~ −9pp
  ProTax	+2%	+4%	−2pp
  Mailchimp (broken out starting FY27)	−1% ~ 0%	Not broken out in FY26	No y/y base; only the absolute level is visible

  Table scrolls horizontally

  All four lines in this table with a clean y/y base (total revenue, TurboTax, Credit Karma, ProTax) are decelerating. GBS is excluded from the "deceleration" verdict because Mailchimp was carved out starting FY27, making the two periods non-comparable; Mailchimp has no FY26 standalone base, so all we can see is that it is itself at zero growth. The conclusion is limited to those four lines; no universal "no exceptions across the board" claim is made. The harder point is this: the FY27 revenue guidance midpoint of $23.396 billion is about 1.45% below the $23.74 billion consensus — that is the real reason for yesterday's −3.37% and this morning's −9%. On the call, the CEO put the TurboTax problem bluntly: "price is now the number one reason customers leave TurboTax."

• The offset: margins are still expanding. FY27 comparable old-basis non-GAAP operating margin is 43.3% vs. FY26's 41.7% (+1.6pp); restructuring savings plus pricing/mix absorbed most of the EPS impact of the revenue deceleration, so same-basis EPS is still +18~19%. But that depends on the restructuring savings landing, and the 8-K does not disclose headcount reductions or annualized savings.

• ⚠️ A risk that has been structurally opened up but has not yet materialized: breaking out zero-growth Mailchimp as a reportable segment starting FY27 usually corresponds to "preparing for a divestiture" or "preparing to ring-fence a drag." On the balance sheet, goodwill is $13.981 billion, 74% of shareholders' equity of $18.992 billion, and Mailchimp was a roughly $12 billion acquisition in 2021. Finer segment granularity makes goodwill impairment testing easier to trigger.

• Cash-flow quality correction (important): FY26 operating cash flow was $8.838 billion, +42.4% y/y, which looks extremely strong. But "deferred income taxes" contributed +$1.279 billion, versus −$435 million in FY25 — a $1.714 billion y/y swing that is entirely non-cash (matched by long-term deferred tax assets collapsing from $1.222 billion to $172 million). Excluding it, FY26 free cash flow is about $7.338 billion — meaning roughly half of that "OCF +42%" is not sustainable.

• Shareholder returns are clean: FY26 buybacks of $5.5 billion (+96%), diluted share count −2.1%, so buybacks did cover SBC dilution; dividend +15%; net debt of only $469 million, close to zero net debt.

• Valuation position (two bases listed separately, never mixed):

  • On the 8/25 close of $357.46: −49.3% from the 52-week high of $705.08; TTM GAAP P/E 21.7x, at the 8.0th percentile of the last three years (three-year median 58.5x); P/S 4.62x, 7.7th percentile (median 10.28x).
  • On the pre-market trade of $323.25 (08:31 ET): −54.2% from the 52-week high; P/S about 4.19x. On either basis, this is a valuation-regime shift, not ordinary volatility.

• ⚠️ But a low percentile ≠ support: the three-year P/S low of 3.38x occurred just two months ago (2026-06-25, close $255.07). At yesterday's close 4.62x is 37% above that low; even at the pre-market 4.19x it is still about 24% above. Monthly path: 2025-07 $785 → 2026-06 $261 (52-week low) → 2026-08 $357 — this is a second leg down after a rebound, not a first decline.

• The sell-side pattern is worth noting: yesterday JPMorgan downgraded with a target cut of $605→$331, BofA cut to Neutral, Wells Fargo $360→$300, Morgan Stanley $335→$315; meanwhile the bears are raising — Stifel $275→$300 (Hold), Piper Sandler $250→$290 (Underweight). Bulls cutting and bears raising is the classic pattern of converging expectations.

• Final judgment: Watch closely, do not chase the short. The core pillar of the bear case (the −39% guidance blowup) is a false proposition per primary verification, and once the sell-side publishes restated-basis models today there is short-squeeze risk; but the revenue deceleration across the board and the 1.45% shortfall to consensus are real, and do not constitute a reason to buy. This is a spot where both sides' arguments are weaker than they look — the real disagreement is not in the numbers but in whether the market is pricing "AI's eventual replacement of tax/accounting software." If it is, an 8th-percentile valuation is not a margin of safety, and the $255 from two months ago is the evidence.

On the surface it is a blowout report; unpacked, it is a good report but not a blowout.

All data are taken from the SEC 8-K original (filed today at 07:38 ET)⁷:

• Excluding the refund, EPS of $2.42 still exceeds the top end of guidance by 21% ($2.42 / $2.00 − 1 = 21.0%) — relative to the company's own May guidance this is a genuine beat, unlike the earlier HD case (where excluding the refund turned a 4.0% beat into a 7.0% miss). The $1.75 is likewise an after-tax figure (the 8-K footnote specifies a 26% tax rate), on the same basis as EPS, so it can be subtracted directly.

• 🔴 But "beating guidance" and "beating last year" are two different things, and on a like-for-like basis the year-over-year comparison is down (prior-year figures likewise taken from the comparison column of this 8-K):

  Excluding the tariff refund	2026 Q2	2025 Q2 (adjusted non-GAAP)	YoY
  Operating income	$153 million	$168.1 million	−9.0%
  Operating margin	12.1%	13.9%	−182bps
  Earnings per share	$2.42	$2.32	only +4.3%

  Table scrolls horizontally

  Over the same period the company repurchased 7% of its beginning-of-year share count ($282 million). In other words: excluding the one-off refund, ANF's operating profit is down year over year, and that +4.3% in EPS is carried mainly by buybacks, not by operations. Layer on flat comparable sales and Hollister comps of −3%, and this report "beats the low guidance the company gave itself in May" — it does not beat last year.

• But the magnitude of the full-year guidance raise is being badly overstated:

Excluding the refund: operating margin of 12.3%–12.8% vs. the prior 12.0%–12.5%, only +30bps; EPS of $11.00–11.50 (midpoint $11.25) vs. the prior midpoint of $10.60, +6.1%, not the apparent +24%.

🔴 And how much of that +6.1% is operating improvement is something this brief cannot attribute. At least three non-operating contributors are known: ① the effective income tax rate falling from about 30% to about 29% (worth about +1.4% on its own); ② buybacks stepped up from about $450 million to ≥$500 million; ③ 8-K footnote (3) states explicitly that the new guidance lowers the assumed effective tariff rate from the prior 15% to 10%–12.5% — that is a change in an external policy assumption, not the business getting better. The 8-K does not break out the contribution of each of these three, and until they are broken out, the +6.1% should not be read as operating improvement; after deducting all three, the residual operating contribution could be close to zero.

• The real soft spot is comparable sales: total comps were flat (0%), with Hollister comps −3% and EMEA comps −4%; only the Abercrombie brand had comps of +4%. The +5% in net sales comes mainly from store openings (about 30 net new stores for the year). "Revenue growing but same-store sales not growing" is the most classic quality flaw in a retail stock.
• Q3 guidance: revenue +5%~6%, operating margin 13.0%–14.0% (including about 160bps of refund → 11.4%–12.4% excluding), EPS $2.90–3.20 (including $0.35 → $2.55–2.85 excluding).
• Pre-market and technicals: 08:08 ET +10.13% (316,000 shares) → 08:27 ET +11.11%($121.00,366,000 shares), the gap is widening with volume accumulating in step, so the quote is credible. At the current price it is still 10.0% below the 52-week high of $133.11. It has already gapped up substantially, and gap-fill risk rises in step with the gain.
• Final judgment: Watch closely, do not chase. Excluding the one-off it is still a beat, but the magnitude shrinks from "a 24% raise" to "a 6% raise," while the stock has already gapped 10% pricing in the former. The probability of a gap fill is higher than usual — that is a risk note, not an entry instruction.

Data taken from the SEC 8-K original (filed today at 08:03 ET)⁸:

• Q2 comparable brand revenue +6.2%, total revenue +6.7%; GAAP diluted EPS $2.84, non-GAAP $2.10 (+5.0% y/y) vs. consensus of $2.05 → a 2.4% beat.
• Key contrast: WSM and ANF treated the same tariff refund in opposite ways. WSM excluded as non-GAAP adjustments the entire $167.8 million refund credited to COGS, $6.3 million of interest income, a $47.5 million supplier rebate reserve, and a $10 million one-time employee 401(k) contribution. As a result:
  • GAAP gross margin 51.6% (+450bps, of which the refund contributed a net +610bps)
  • Non-GAAP gross margin 45.5%, −160bps y/y (merchandise margin −230bps, chiefly tariff costs)
  • GAAP operating margin 22.9% vs. non-GAAP 17.3% (−60bps y/y) In other words: WSM's non-GAAP numbers are clean, while ANF's GAAP EPS of $4.17 includes the refund. Cross-company comparisons that do not first align the basis will produce a completely wrong ranking.
• Full-year guidance raised: revenue +4.7%+7.2%, comps +4.0%+6.5%, non-GAAP operating margin 17.8%–18.2% (prior 17.5%–18.1%).
• There is a trackable divergence buried in the guidance: the company explicitly assumes "oil prices will stay elevated in the second half." Yet oil is falling for a third straight day today. If Hormuz diplomacy keeps advancing, WSM's own guidance assumption is conservative and has room to be raised — a real chain linking the macro main line to single-stock fundamentals.
• Inventory and cash flow: inventory +1.0% to $1.45 billion (healthy); operating cash flow $696 million (including $200.2 million of refund received).
• Evolution of the pre-market quote (updated): at 08:00 ET it showed +2.67% ($241.00) and at 08:05 ET −2.87% ($228.00), but cumulative volume at both timestamps was an unchanged 3,165 shares (quote frozen, unusable); at 08:27 ET updating resumed, printing −4.68%($223.76,7,308 shares). Direction is now confirmed as down, but volume is still under 50,000 shares, so it is usable for direction only, not for magnitude.
• Final judgment: Watch closely — and this is the single most worthwhile "fundamentals vs. price divergence" to monitor today. The fundamentals are the cleanest of this batch of retail prints: comps +6.2%, honest accounting (all tariff refunds adjusted out of non-GAAP), guidance raised, inventory +1.0% and healthy. Yet the stock is falling. Two explanations: ① what the market cares about is non-GAAP gross margin −160bps y/y and non-GAAP operating margin −60bps y/y — excluding the refund, margins are actually contracting, which discounts the quality of the "raised guidance"; ② a valuation constraint (some secondhand analysis claims its P/E is about 75% above its five-year average, but this brief did not obtain the primary source or the basis for that calculation and does not use it as a criterion). Today's 10:00 ET call and its explanation of merchandise margin and tariff costs are the key.

• 8/25 after hours: FY27Q2 revenue $341.9 million (a record, +33% y/y, vs. consensus of $328.6 million), adjusted EPS $0.71 vs. consensus $0.61.
• The guidance is what matters: Q3 revenue guidance of $410 million, about $50 million above consensus; adjusted operating margin guidance of 31.0%, +660bps QoQ. Management attributes it to AI networking demand.
• This is the only report today with all three of "in-quarter beat + large next-quarter guidance beat + margin expansion", a direct contrast to INTU/ZM's "beat but dull guidance."
• 🔴 But the pre-market has already digested the entire positive: 08:08 ET printed +4.10%($132.75,419,000 shares), and by 08:27 ET it had flipped red to −1.20% ($125.99, 536,000 shares). The giveback happened while volume kept expanding — not thin-volume noise. It had already closed +5.47% yesterday, meaning the market priced this in the day before the release, and the after-hours and pre-market sessions merely handed that pricing back.
• Risk: highly correlated with NVDA — if NVDA's guidance disappoints tonight, SMTC gives back a second time. At $125.99 it is still 40.8% below the 52-week high of $177.35.
• Final judgment: Downgraded from "watch closely" to "watch only." There is nothing wrong with the quality of the report (it remains the only one in this batch with all three of "in-quarter beat + large next-quarter guidance beat + margin expansion"), but the price no longer offers you an entry discount for that quality. If it can reclaim yesterday's close of $127.52 after the open, reassess.

Yesterday's most eye-catching large-cap advance — but on verification the catalyst contains no new information, which makes this the most counterintuitive item in this brief.

• Catalyst: Raymond James analyst Simon Leopold upgraded AMD from Outperform to Strong Buy (the firm's top rating) on 8/25, with a target of $565→$641, and raised the 2030 server CPU market size by $37 billion to $201 billion (about a 44% 5-year CAGR).

• 🔴 But AMD's own management gave the same metric as $220 billion, growing "more than 50%" annually, on the Q2'26 call on 8/11. That is, the sell-side TAM is below the number the company already publicly disclosed three weeks earlier. The essence of yesterday's +4.91% is one sell-side firm moving its model toward the company's existing figures — not the market receiving new information.

• $641 is not aggressive either: the consensus target averages $612.29 with a median of $600 (S&P Global, 54 analysts). $641 is only 6.8% above the median — a "slightly above consensus" follow. The genuinely aggressive figure is the $1,250 at the top of the range.

• The fundamentals themselves are genuinely accelerating (data from the company's 8-K originals):

  Quarter	Revenue $M	YoY	Data center $M	Non-GAAP gross margin	Non-GAAP EPS
  Q3'25	9,246	+35.6%	4,341	54%	$1.20
  Q4'25	10,270	+34.1%	5,380	57%	$1.53
  Q1'26	10,253	+37.9%	5,775	55%	$1.37
  Q2'26	11,536	+50.1%	6,718 (+107%)	56%	$1.66

  Table scrolls horizontally

  TTM revenue is $41.305 billion, with data center at 53.8%; non-GAAP operating margin improved from 12% to 27%; server CPU unit share rose from 27.3% to 34.5% (Mercury Research, Q2 2026). Q3'26 revenue guidance is approximately $13.0 billion ± $300 million (+41% YoY). ⚠️ The true y/y move in gross margin is 54%→56% (+2pp), not the +13pp seen in the news — the year-ago quarter included an $800 million inventory write-down from MI308 export controls, which is base noise.

• But valuation and cash flow are another matter:

  • P/S 18.94x, at the 91.3rd percentile of the last three years (three-year median 10.26x); TTM non-GAAP P/E 83.2x; forward P/E (NTM) 43.15x, falling to 31.0x only on FY2027 consensus EPS — and that FY27 consensus implies EPS +104.6% y/y.
  • FCF yield is only 1.07%; Q2'26 operating cash flow of $2.366 billion / non-GAAP net income of $2.760 billion = 0.86x, so profit-to-cash conversion is deteriorating (receivables −$1.246 billion in the quarter, capex +187% y/y).
  • 🔴 Buybacks went to zero while SBC is 4.5x that amount: H1'26 buybacks were only $221 million (versus $1.227 billion in H1'25, −82% y/y), with $0 in Q2'26; SBC over the same period was $990 million; diluted share count +1.78% y/y. The company stopped buying back stock while the shares are in a high range and it holds $8.84 billion of net cash.

• The implied second-half slope is steep: H1'26 actual of $21.79 billion + Q3 guidance of $13.0 billion = $34.79 billion for the first three quarters; full-year consensus of $50.81 billion → an implied Q4'26 of about $16.0 billion (+23% QoQ, +56% YoY, off the $10.27 billion base of Q4'25; the prior version of this brief erroneously wrote +109% y/y, having crossed wires with the FY27 EPS +104.6%, now corrected). The entire 43x forward P/E and the $641 target rest on that slope.

• Technical position: at $476.30 it is −18.5% from the 52-week high of $584.73 (equivalently: it needs to rise 22.8% to regain the high; this brief uses the drawdown convention throughout); but the monthly path was May $516 → June $581 (peak) → July $476 (−18.03%) → August $479 — peaked in June, pulled back in July, and has been ranging between 470 and 520 in August, with yesterday's +4.91% not yet breaking the top of the August range ($514.39).

• Final judgment: Watch only. The acceleration in fundamentals is real, but yesterday's advance priced information the company had already given three weeks earlier; with P/S at the 91st percentile, an FCF yield of 1.07%, and buybacks at zero, the risk is not "expectations aren't high enough" but "expectations are already high and depend heavily on whether data center doubles in 2027." Tonight's NVDA guidance will incidentally price AMD's 2027 assumptions too.