The Framework Went Up Again and the Stock Went Down 13.5%: Marvell Pays for Growth in Equity

Coverage Context Since the Q1 Print

We downgraded Marvell to Hold from Outperform on May 28 at $204.83, with a fair value range of $175 to $240 and a twelve-month target of $215. The argument was explicit and narrow: the operating thesis was stronger than when we upgraded in December at roughly $98, but the share price had moved further and faster than the earnings, and per-share growth was being taxed by a share count that had expanded 6.9% in two quarters. We wrote that we would revisit a positive rating on a pullback toward the high $160s to low $170s, on evidence that scale-up switching was converting to disclosed revenue inside the fiscal 2028 window, or on a fiscal 2028 gross margin path materially above the 56% to 57% our operating margin arithmetic implied.

Between then and now the stock did something the downgrade did not anticipate. It ran to a $316.43 closing high, came into this print at $241.45, and then gave back 13.5% over six sessions to close at $208.83 on September 3. Across the full interval it is up 1.9% from the level at which we downgraded, having been up 54% at the peak. The framework, meanwhile, went up twice more. This report grades the same nine thesis pillars we have carried since December, and the rating question this quarter is whether a 13.5% drawdown on a beat-and-raise has closed enough of the gap between the price and the framework. It has not, but it has closed some, and one of our three revisit conditions was met on the call.

Results vs. Consensus

Q2 FY2027 Scorecard

Year-over-Year Comparison

Sequential Comparison

Revenue assessment. The absolute number is less interesting than its shape. Data center is now 79% of revenue, up from 74% a year ago and 76% last quarter, and its growth accelerated on both measures: +45.7% year over year against +27.2% in Q1, and +18.5% sequentially against +11.0%. A business does not usually accelerate on a larger base, and the Q3 guide asks for more of it, at more than 20% sequential and roughly 75% year-over-year data center growth. Netting the guided low-to-mid-teens sequential decline in communications, the Q3 revenue guide implies data center of roughly $2.66B, or about 84% of total revenue. Marvell is, for practical purposes, becoming a data center company with a communications runoff attached, and the concentration in the revenue base is now structural rather than cyclical.

Margin assessment. There are two margin stories in this print and they point in opposite directions. GAAP gross margin rose 270 basis points year over year on better absorption, and non-GAAP operating margin rose 180 basis points, so at the operating line the leverage is working exactly as the model says it should. Non-GAAP gross margin, meanwhile, fell 50 basis points year over year and is guided down 90 more at the Q3 midpoint. The distinction matters because gross margin is the line that carries mix information and operating margin is the line that carries scale information. Marvell is winning on scale and losing on mix, and the mix loss is a direct function of the custom ramp that the entire bull case depends on. That is not a contradiction the company can resolve by executing better. It is the arithmetic of the business it has chosen.

EPS assessment. GAAP EPS of $0.33 landed four cents below the guidance midpoint even as non-GAAP landed a penny above, and the two largest identifiable drivers are both non-operational. Operating-expense stock-based compensation ran $310.3M against the roughly $282M contemplated in the May guide, and the Celestial earnout remeasurement net of the forward purchase contract put another $52.0M of pre-tax expense through Other expense, a line no guidance table forecasts. The larger point is the trend rather than the quarter: total SBC reached 11.9% of revenue against 8.6% last quarter and 7.7% a year ago, a 57.1% sequential increase on 13.3% more revenue. Marvell's GAAP operating margin is 16.8% and its non-GAAP operating margin is 36.6%, and readers who underwrite the non-GAAP number are underwriting a 19.8 point adjustment.

Segment Performance

Percentage changes in the table above are the company's own rounded figures from the revenue-by-end-market disclosure. Computed from the reported dollars, data center grew 45.7% year over year and 18.5% sequentially, communications and other grew 10.1% and declined 3.0%, and total revenue grew 36.5% and 13.3%.

Data Center

Data center delivered $2,171.5M, a record, and the growth rate accelerated on both comparisons at the same time. Year over year it went from +27.2% in Q1 to +45.7% in Q2. Sequentially it went from +11.0% to +18.5%. Management is guiding for the acceleration to continue, and the Q3 framing of more than 20% sequential and roughly 75% year-over-year would be the third consecutive quarter of a rising growth rate on a base that has grown 46% in twelve months.

"Now looking ahead to the third fiscal quarter, we expect this acceleration to continue with data center revenue forecasted to grow more than 20% sequentially and roughly 75% year-over-year. The drivers of this growth remain very broad-based as AI demand for our products continues to rise."
— Matthew Murphy, Chairman and CEO

The composition matters as much as the rate. Connectivity carried the quarter and, on management's own account, carries the larger share of the fiscal 2028 raise. Within scale-out, 800-gig optical DSP demand remains strong while the 1.6T ramp accelerates, scale-out switching is on track to more than double this fiscal year on the 51.2T product across a broadening customer base, and the broadband analog TIA and driver franchise is described as exceeding expectations. Three of those businesses now carry an explicit run-rate milestone.

"Each of these 3 businesses is on or ahead of the trajectory towards the $1 billion annualized revenue run rate we highlighted last quarter."
— Matthew Murphy, Chairman and CEO

Custom, by contrast, is a second-half event. Management said the acceleration in custom demand begins in the second half of fiscal 2027 and that custom more than doubles in fiscal 2028, unchanged from the May framing but now attached to a named commercial agreement. The timing is the whole story for margins: custom carries below-corporate gross margin, so a business whose custom line inflects in Q3 is a business whose gross margin steps down in Q3. That is precisely what the guide does.

Assessment: Data center is executing above the framework on every measurable line, and the mix inside it is shifting toward the part that dilutes gross margin. Those two facts are the same fact. The segment is the reason to own the stock and the reason gross margin is capped, and no amount of additional demand resolves the tension because more custom demand makes the mix worse, not better.

Communications and Other

Communications and other produced $567.8M, down 3.0% sequentially and up 10.1% year over year, and management characterised the trajectory as "largely as expected" with fiscal 2027 growth expected to "approach our 10% target." The segment is now 21% of revenue, down from 26% a year ago, and it is shrinking as a share of the business by roughly five points a year through arithmetic rather than decline.

"For the third fiscal quarter, we expect revenue to decline in the low to mid-teens percentage range both sequentially and year-over-year, followed by a solid sequential recovery in the fourth quarter."
— Matthew Murphy, Chairman and CEO

That is a sharper decline than the segment's usual lumpiness and it goes unexplained beyond "the mix of businesses in this segment." A low-to-mid-teens sequential decline takes the segment to roughly $483M to $494M in Q3, which would be its weakest quarter since the automotive ethernet divestiture closed in August 2025, and the year-over-year decline of the same magnitude means this is not a comparison artifact. No analyst asked about it on either of the last two calls.

Assessment: The segment is now small enough that a 15% sequential miss costs roughly 3% of total revenue, so the market is right to treat it as noise for modelling purposes. It is not noise for what it says about disclosure discipline: a fifth of the revenue base is guided to a double-digit decline in consecutive directions with a one-clause explanation, and the guidance was accepted without a question.

Channel and Customer Mix

Assessment: This is the first quarter in the coverage window in which concentration stopped deteriorating. Distributor A ticked down a point, and the receivable concentration broadened from three customers at 75% to four at 72%. Neither is a resolution. Two relationships still carry 60% of revenue against 50% a year ago, distributor revenue has drawn level with direct revenue for the first time, and the concentration disclosure appears only in the 10-Q. It was not mentioned in the release, not raised on the call, and not asked about by any of the nine analysts who spoke. That silence is now three quarters old.

The Google Warrant

On August 19, eight days before the print, Marvell disclosed a commercial agreement entered into with Google on July 29 and an associated warrant issued on August 18. It is the largest single disclosure of the quarter and it received roughly two paragraphs of prepared remarks and three of the nine analyst questions. The mechanics are worth setting out precisely, because most of the discussion around the deal has worked from the headline number rather than the structure.

Three things follow from the structure that did not get said on the call.

First, the near-term economics are already in the guidance. Management was explicit that the agreement does not add to the fiscal 2028 number, so the incremental revenue this warrant is meant to unlock begins in fiscal 2029.

"In terms of the impact of the new warrant agreement, revenue from programs covered by the agreement through fiscal 2028 is already reflected in the overall custom revenue target we have previously provided."
— Matthew Murphy, Chairman and CEO

The warrant, however, is issued now, and its first tranches can vest from Q3 of this fiscal year. Shareholders take the dilution mechanism in fiscal 2027 and 2028 and receive the incremental revenue in fiscal 2029 and beyond. That asymmetry is the honest reading of why a beat-and-raise sold off.

Second, the fully loaded share count now clears one billion. Marvell has four separate contingent equity claims outstanding, and only one of them is normally discussed.

Third, and least discussed, Marvell's own accounting policy treats customer warrants as a reduction to revenue. This is not our inference. It is the policy disclosed in the 10-Q for the two precedent customer warrants Marvell has already issued.

"The Fiscal 2025 Warrant Shares vest primarily based on the customer's achievement of qualifying product revenue milestones and are recognized as a reduction to revenue as qualifying revenues are recognized during the vesting term. The grant date fair value of the warrant was determined to be $54.44 per share and a total fair value of $227.6 million using the Black-Scholes option pricing model."
— Form 10-Q for the quarter ended August 1, 2026, Note 3

Assessment: The commercial agreement is the strongest third-party validation the custom franchise has received, and the vesting structure is well designed from Marvell's side, because nothing vests unless the revenue arrives. Our concern is not that the trade is bad. It is that the reported economics of the trade have not been laid out for shareholders. The company disclosed the share count and the strike, the market computed the dilution, and nobody on the call connected the warrant to the revenue-recognition policy sitting in Marvell's own filing. If the contra-revenue treatment applies at anything like the scale the precedents imply, the reported revenue and gross margin of the custom business over the next seven years are lower than the gross bookings suggest, and the October 6 Investor Day is where that needs to be quantified.

Key Topics & Management Commentary

Overall Management Tone: Management was declarative on demand and deliberately unquantified on almost everything else, deferring almost every request to size a number to the October 6 Investor Day, which the call referenced fourteen times. The posture on the raise was matter-of-fact rather than promotional, which is consistent with the last two quarters, but the willingness to size anything fell sharply: the scale-up optics raise, the second XPU program, the fiscal 2029 custom number and the entire scale-up switching opportunity were all described qualitatively and left unsized. The one place the tone tightened was gross margin, where a new CFO gave a clearer forward framing in his first quarter than the company had offered in the prior three.

1. Both fiscal years moved up again, and the raise was broad-based

This is the third consecutive quarter in which the forward framework has been raised, and this raise matched the May increments on both fiscal years. Fiscal 2027 went to roughly $12B from roughly $11.5B, and fiscal 2028 to roughly $18B from $16.5B. The more notable feature is that the growth rate rose alongside the base.

"Putting it all together, we now expect fiscal 2028 revenue of approximately $18 billion, up $1.5 billion from the $16.5 billion outlook we provided just 1 quarter ago. And importantly, even as our revenue base becomes significantly larger, our growth rate is accelerating. We now expect fiscal 2028 revenue to grow approximately 50% year-over-year compared with approximately 45% in our prior outlook."
— Matthew Murphy, Chairman and CEO

Data center carries all of it. Fiscal 2027 data center growth went to approximately 60% from approximately 50%, and fiscal 2028 to more than 60%. Set against our December framework, when fiscal 2027 was a $10B year, the company has raised its own current-year outlook by 20% in nine months.

Assessment: Nothing in the raise is padded and nothing is inorganic. The pattern of serial upward revision is now four quarters long and is the strongest single argument for the stock. It is also the argument the market has heard four times, which is why the fifth delivery of it moved the shares down rather than up.

2. Grading the commitments made on the May call

The May print carried ten forward commitments. Eight were met or exceeded, one was raised against shareholders, and one changed shape.

Assessment: The revenue commitments were kept and beaten; the cost commitments moved against shareholders twice in one call. That combination is the fingerprint of a company that is genuinely capturing more demand than it forecast and is spending more than it forecast to capture it. Both facts are true and only one of them was framed as news.

3. Gross margin, and the first forward framing the company has given

The Q3 guide of 57.5% to 58.5% non-GAAP gross margin is a 90 basis point step down at the midpoint from the 58.9% just delivered, and the CFO named the cause without hedging, including a mid-sentence self-correction that made the point more clearly than the corrected version.

"Revenue levels and product mix remain key determinants of gross margin in any given quarter with the forecasted acceleration of our custom business creating the sequential headroom in the fiscal third quarter -- the sequential headwind in the fiscal third quarter. We currently expect to maintain gross margin in this range in the fourth fiscal quarter."
— Daniel Durn, Chief Financial Officer

Asked directly in Q&A to extend that into fiscal 2028, the CFO did, which is the first time Marvell has framed a forward gross margin beyond the guided quarter.

"Gross margin depends on ultimate mix. My preliminary view is gross margins next year are going to be in a similar range, same range as we're exiting this year. So back half of this year, same range for FY '28."
— Daniel Durn, Chief Financial Officer

Assessment: Read against our own May analysis this is good news that the market took as bad news, and both reactions are defensible. We derived a 56% to 57% fiscal 2028 gross margin from the operating margin target and the then-current operating expense commitment, and complained that management had never stated it. Management has now stated it, and the number is roughly 58%, a hundred to two hundred basis points better than our implied figure. What the market heard instead is that gross margin expansion has been removed from the story for two full fiscal years, in a stock priced for a business whose every other line is accelerating. The disclosure improved and the outlook it disclosed was worse than the multiple assumed.

4. The operating expense commitment changed shape twice in one call

Fiscal 2027 non-GAAP operating expenses went to approximately $2.55B from approximately $2.45B, which implies roughly $1.362B across the back half against $1.188B in the first half. Fiscal 2028 was reframed from a number into an algorithm.

"For fiscal 2027, we expect non-GAAP operating expenses of approximately $2.55 billion, slightly above our prior expectation of $2.45 billion, reflecting the significantly larger revenue opportunity we now see. Importantly, we expect revenue to continue growing substantially faster than operating expenses."
— Daniel Durn, Chief Financial Officer

On fiscal 2028: "we currently expect non-GAAP operating expenses to grow at roughly half the rate of revenue growth in percentage terms." At approximately 50% revenue growth that is roughly 25% expense growth, against the mid-to-high-teens figure given in May. The offsetting news is that the operating margin timeline pulled forward: the 38% to 40% target range is now expected to be entered in Q4 of this fiscal year rather than during fiscal 2028, and the high end reached through fiscal 2028.

Assessment: A ratio commitment is weaker than a dollar commitment because it self-adjusts with whatever revenue turns out to be. The company gave up a fixed target and replaced it with a formula, and did so in the same breath as raising the current-year number by $100M. The operating margin commitment survives only because revenue was raised more than expenses were. That is a real result, but it is a different kind of result from the one promised in May, and the substitution was not flagged.

5. Scale-up optics became the single largest driver of the fiscal 2028 raise

In May the framing was that Celestial's photonic fabric contributed roughly $150M in fiscal 2028 inside a scale-up optics category of roughly $300M. Three months later the category is described as accelerating "even further" and as a significant driver of the $1.5B raise, and the reason is that near-package optics has moved faster than co-packaged optics.

"And I think the magnitude of our scale-up optics opportunity next year is much larger than we thought just a quarter ago. The year after is going to be, I think, way larger than I sort of could have ever comprehended back even last year when we were looking at Celestial."
— Matthew Murphy, Chairman and CEO

The technical argument is that customers are pursuing several architectures at once rather than converging on one, and Marvell's breadth across modulator technologies (MZM, EAM and MRM), packaging options (NPO and CPO) and its silicon-germanium TIA and driver franchise lets it participate regardless of which wins. Management explicitly rejected the substitution framing: "we see this as an and, not an or."

Assessment: This is the most genuinely new information in the quarter and it is entirely unsized. A category described in May at roughly $300M for fiscal 2028 is now the largest component of a $1.5B raise, which would imply a number several multiples of the original. Management gave no revised figure, and the Celestial acquisition's own earnout, worth up to 22.4M shares, vests on cumulative revenue milestones through fiscal 2029 that would be directly informed by that number. Investors are being asked to credit the acceleration and price the earnout without either datum.

6. Scale-up switching remains excluded from the framework for a second consecutive quarter

Marvell claims a unique position across all three purpose-built scale-up protocols: internally developed UALink and ESUN switches, plus NVLink Fusion through the NVIDIA partnership. The engagements are described in terms that would be material if quantified.

"As a result, we are engaged in multiple deep discussions with Tier 1 customers across our scale-up switch portfolio, with each engagement representing a multibillion-dollar lifetime revenue opportunity given the expected size of the scale-up TAM."
— Matthew Murphy, Chairman and CEO

The strategic argument for why the switching position and the optics position reinforce each other was the most substantive technical answer of the call, and it explains why Marvell is investing ahead of revenue.

"What customers really want to see though ... is that you then need to show a compelling clear roadmap that you can execute with credibility on the optics side, both NPO and then all the way to full integration on CPO ... I'm even more excited about the optics attach once that gets going over the next few years, because it just effectively almost doubles the SAM of the switching in terms of the attach you can get."
— Matthew Murphy, Chairman and CEO

Assessment: We flagged in May that scale-up switching was excluded from the $16.5B and asked whether it would get quantified. It was not, and it is still excluded from the $18B. That is a genuine free option in the numbers, and it is also the second quarter in which a "multibillion-dollar" opportunity has been described without a single revenue dollar, design win count or customer name attached. The right way to hold it is as upside we do not model and do not pay for.

7. CXL and memory expansion turned into a real business on memory scarcity

Structera and the broader CXL investment were sold to investors five years ago as a server memory architecture play. Management now describes it as a beneficiary of AI inference and of DRAM scarcity, which is a different and larger market than the one it was underwritten for.

"It turns out all that investment we've made is just ideal for memory expansion and inferencing. And so we're seeing this technology now getting deployed at multiple hyperscalers with varying architectures by the way, in extremely high volumes. One is just the demand of inferencing require it. The other is what we're seeing is as a result of the scarcity that's out there in memory. Customers are modifying and adjusting their plans to actually use more of this type of technology. So this has continued to have a strong upward bias. We've actually secured additional design wins in this area in the last couple of quarters."
— Matthew Murphy, Chairman and CEO

Memory interface controllers and near-memory compute both appear in the Google agreement's product scope, so this is not a standalone story. It is one of the XPU attach categories the warrant covers.

Assessment: A product line whose demand is driven partly by a memory shortage carries a demand profile that reverses when the shortage does. That does not make the design wins less real, but it argues for treating the current upward bias as cyclical amplification of a secular trend rather than as the trend itself. Management drew no such distinction.

8. The custom acceleration is a second-half event, and it is the entire margin story

Custom demand accelerates significantly in the second half of fiscal 2027, more than doubles in fiscal 2028, and accelerates significantly again in fiscal 2029. Those three statements were all repeated from May and none was sized. What changed is that the Q3 gross margin guide now makes the timing visible in the numbers, and the CFO confirmed the causation directly when asked.

"Part of the acceleration story in Q3, as you point out, mix is the primary driver. We've got a strong ramp in custom. And so you can see that play out in the profiling of the margins. Not a surprise. We've been signaling that custom ramp for quite some time as we window into Q4, that strength, we're signaling a bigger step-up, much larger step-up from a revenue standpoint, if you roll in the $12 billion annual target."
— Daniel Durn, Chief Financial Officer

Assessment: The company has now put its own arithmetic on the record: more custom revenue means lower gross margin, and the fiscal 2028 doubling of custom means gross margin does not recover. The offset is operating leverage, which is real and which management has pulled forward by a quarter. But the offset is finite, because the 38% to 40% operating margin model is a ceiling management set and reaffirmed. Beyond that ceiling, additional custom revenue converts to earnings at a declining rate.

9. A new CFO, a new head of investor relations, and a model reset in three weeks

Willem Meintjes stepped down as CFO in mid-June after a decade with the company. Dan Durn, previously a Marvell director, took the role. Ashish Saran will retire from investor relations in April 2027 and Ross Seymore, a sell-side semiconductor analyst of more than twenty-five years, has taken over. Both changes landed inside one quarter, at the company's largest inflection.

"I recognize that the growth opportunity ahead must be matched by its quality, and I plan to remain intensely focused on expanding operating margins, growing cash flow and driving attractive stockholder returns."
— Daniel Durn, Chief Financial Officer

The new CFO also pre-announced that the long-term financial model itself is being rewritten: "we're going to reset that long-term target model here in the coming weeks at the Analyst Day."

Assessment: An incoming CFO gave a clearer forward gross margin framing in his first quarter than the company had given in the prior three, which is a disclosure improvement worth crediting. The flip side is that the operating model shareholders have been underwriting since 2021 is being replaced in five weeks by an executive who has held the role for ten. Every unsized answer on this call points at that same October 6 date, which makes the Investor Day a genuine binary rather than a routine update.

10. Capital intensity is rising faster than revenue

The cost of securing the supply behind the framework is visible in three places in the filings and was described on the call only as a positive.

Operating cash flow of $605.5M fell 5.2% sequentially from a record $638.8M despite non-GAAP net income rising 20.6%, and the CFO attributed the gap to "the higher capacity prepayments to suppliers in support of Marvell's future growth." The six-month working capital outflow was $660.3M, driven by prepaid expenses and a $333.6M reduction in accounts payable. Capital expenditure was $126.7M in the quarter against $47.5M a year ago.

Assessment: A company whose revenue is compounding at 50% and whose foundry commitments run $2.1B to $2.2B annually through fiscal 2030 is converting balance sheet into capacity, and the TSMC arrangement carries an explicit forfeiture provision if target wafer quantities are not purchased. That is the correct trade in a supply-constrained market and it is not free. Free cash flow was again not given a figure on the call, for the second consecutive quarter.

11. Stock-based compensation stepped up sharply and was not addressed

Total stock-based compensation was $326.2M in the quarter, 11.9% of revenue, against $207.6M and 8.6% in Q1 and $153.6M and 7.7% in the year-ago quarter. That is a 57.1% sequential increase on 13.3% more revenue. The Q3 guide implies a partial reversal, with operating-expense SBC of $285M against $310.3M in Q2, so the step-up looks at least partly discrete rather than structural.

Assessment: Neither the release nor the call mentioned it, and no analyst asked. A three-point step in SBC as a percentage of revenue in one quarter is material to anyone underwriting the non-GAAP earnings stream, which is what the multiple is applied to. The guide suggests it normalises, but the absence of an explanation for a $118.6M sequential increase in a single expense line is a disclosure gap rather than an accounting one.

12. The Series A preferred and the Celestial earnout are still moving the GAAP numbers

The NVIDIA preferred issued in March at a $91.84 conversion price sits in the diluted share count on an as-converted basis and is now roughly 127% in the money. The Celestial earnout is the larger GAAP item: the contingent consideration liability reached $749.5M from nil at fiscal year end, with a $101.9M fair value increase in the quarter and $433.7M across the six months, partially hedged by a $300.0M notional cash-settled forward stock purchase contract that produced a $49.9M unrealised gain in the quarter.

Assessment: The mechanical consequence is that Marvell's GAAP net income now moves inversely with its own share price, because a rising stock increases the value of the shares owed under the earnout. That is why GAAP EPS of $0.33 came in below a $0.37 midpoint in a quarter when the operating business beat. The hedge covers $300M of notional against a liability of $749.5M and a maximum settlement of roughly $233.0M of cash plus 22.4M shares, so the exposure is partially covered rather than closed.

Guidance & Outlook

Multi-year framework

Implied Q4 ramp. The fiscal 2027 framework carries an internal tension worth surfacing. First-half revenue was $5,157.1M and Q3 is guided to $3,150M, so the implied Q4 depends on where "approximately 45% growth to roughly $12 billion" actually lands against a fiscal 2026 base of $8,195.1M. At exactly +45% the year is $11,883M and Q4 is $3,576M, up 13.5% sequentially. At $12,000M the year implies Q4 of $3,693M, up 17.2%. Management committed on the call that fourth-quarter growth accelerates "both sequentially and year-over-year" from Q3's +15.0% and +51.8%, and the sequential half of that promise requires Q4 revenue above roughly $3,623M, which in turn requires the full year to land at or above roughly $11.93B. In other words the "roughly $12 billion" is the operative number and the "approximately 45%" is slightly light. On the $12B figure, Q4 would be up roughly 66% year over year against $2,218.7M.

Street at. Both halves of the Q3 guide came in above consensus: revenue at $3,150M against $3.03B to $3.04B, and non-GAAP EPS at $1.10 against $1.07 to $1.08. Consensus for fiscal 2028 had not caught up to the $18B framework at the time of the print, and the sell-side response was to raise targets, with published increases on the reaction day reaching into the high $200s. Neither the guide nor the target revisions stopped the stock falling 10.3% that session.

Guidance style. Marvell has beaten the midpoint of its own revenue guide in every quarter of the coverage window, by $19.0M in Q4 of fiscal 2026, $18.0M in Q1 of this year and $39.0M in Q2, and has raised the forward framework in each of the last three. The beats are small and consistent, which is the signature of a company guiding to a number it is confident of clearing rather than sandbagging. The framework raises are large and repeated, which is the signature of a demand environment moving faster than the planning process. Both patterns argue that the next quarter's guide will be beaten modestly and the framework raised again. That is precisely the expectation now embedded in the price, which is why the raise stopped being a catalyst.

Analyst Q&A Highlights

Nine analysts asked nine questions across roughly thirty minutes. Three of the nine went directly at the Google warrant, and the dominant response pattern was deferral to the October 6 Investor Day. The one exchange that produced a genuinely new disclosure was the last one, on gross margin.

What the warrant actually covers, and in what proportion

The opening question tried to decompose the agreement between the inference accelerator and the attach products around it. Management confirmed the breadth of the product scope and declined the decomposition, describing the aggregate opportunity in terms of a six-and-a-half-year horizon rather than a near-term contribution. The answer is a good summary of the whole call: expansive on scope, silent on size.

Q: "I wanted to ask first on the warrants with Google. Obviously, a very robust relationship over a multiyear period of time. I would love if you could give a color on what's contributing to that revenue? Obviously, you have an inference accelerator, but you talked a lot about XPU attach, which you guys did kind of start the industry standard on in terms of defining. So maybe what XPU attach are you excited about there? And like any sort of color you can give us on the percentage of contribution between the inference accelerator and some of the attach that you're seeing there?"
— Thomas O'Malley, Barclays

A: "It includes inference accelerators as you mentioned, also storage controllers, NICs, memory interface controllers, near-memory compute, a whole bunch of different products. And you're right, we did define the -- what we call the XPU attach category a couple of years back, and actually we gave quite a detailed view of that in our June 2025 custom silicon event. And I think all of our projections to date have been under called, meaning that opportunity continues to get more and more significant. So, it's a broad set of products, Tom, that are covered here, and it's very broad-based in terms of where we can engage and where we're going to contribute. And I wouldn't call out any one of them at the moment specifically, but all of them in total, you can see represent -- if you look at the total envelope of the opportunity, it's just massive for Marvell and game changing at the sort of peak performance of what could be achieved now over the next 6, 6.5 years."
— Matthew Murphy, Chairman and CEO

Assessment: The refusal to break the agreement into inference accelerator versus attach is defensible on customer-confidentiality grounds and unhelpful for modelling. The substantive concession is the time horizon: "the next 6, 6.5 years" confirms that this is a fiscal 2033 story whose peak contribution sits well outside any window a current shareholder is underwriting.

Sizing $120 billion against the standing fiscal 2029 custom target

The sharpest question of the call did the arithmetic in public and asked management to reconcile it with the existing framework. The answer confirmed the math, confirmed the fiscal 2029 timing, and refused to update the number, which is the exact position that produced the sell-off.

Q: "Matt, given the Google commercial program you signed back in late July, a $120 billion in cumulative revs over 6 years, if you hit all of your milestones, right, that puts you if you just annualize it at about $18.5 billion per year in revenues just in Google XPU attach custom ASICs, right? I actually thought that, that would start to show up in fiscal '28 or your calendar '27, but given your guidance, it looks like custom is still going to be around $5 billion to $6 billion in calendar '27. ... The team has previously targeted $10 billion to $11 billion in custom revenues in fiscal '29 or calendar '28. What does that number look like now based on the Google commercial agreement and some of the new programs associated with the agreement?"
— Harlan Sur, JPMorgan

A: "If you look out to next year, we have comprehended already revenue that would come as part of this warrant in our numbers now. Now that's because some of these programs, obviously, are already in flight or have already started, but the programs ahead of us that are either in execution or just starting production, those will contribute much more significantly in fiscal '29. ... I'm not going to size it on the call here today, but you should assume with our Investor Day coming up Harlan, the Marvell team, we always do a thorough job in our investor -- in our analyst/investor days. And so you should expect a very robust detailed review of how we step through the revenue, not just through fiscal '29, but really out until the end of the decade. ... But clearly, there's a lot of upside bias in those numbers in fiscal '29 and beyond in custom."
— Matthew Murphy, Chairman and CEO

Assessment: The question established that fiscal 2028 custom is unchanged by the largest commercial agreement in company history, and the answer confirmed it. "A lot of upside bias" in fiscal 2029 is the most the company was willing to say. For a stock trading on a fiscal 2028 multiple, a fiscal 2029 upside bias is not a catalyst, and this exchange is the clearest evidence in the transcript for why the market treated a major positive as a timing disappointment.

Progress on the second large XPU program

A recurring line of questioning through this coverage window has been whether the new Tier-1 XPU socket is being underwritten conservatively. The answer confirmed continued progress across design, supply and commercial terms, and again declined to size it, while redirecting the emphasis to a different part of the raise.

Q: "Matt, I was hoping you could give us an update on the other large XPU program that you're planning to start in the next year. I think the assumptions there are quite modest for next year, $600 million or $700 million or so, and I was hoping you could give us an update on that? And then what is the opportunity for that program over time?"
— Vivek Arya, Bank of America Securities

A: "We've been judging that, I think, in a very conservative and practical manner as we've made progress, but we make -- we continue to make progress every quarter, not only on design execution, but also supply commercials and figuring out the -- sort of the envelope of that opportunity. And that's only gotten better. So without quantifying it, exactly what I would say is from the last call, if you look at the $1.5 billion raise we're doing for next year, just at a high level, it's broad-based. So some of that is from custom next year. There's also -- actually get another question on this one, probably the most meaningful part of the next year raise is actually from the scale-up optics switching in other areas, but customs definitely part of it, Vivek."
— Matthew Murphy, Chairman and CEO

Assessment: This answer relocated the fiscal 2028 raise away from custom and toward connectivity, which is analytically important and structurally favourable. Connectivity carries higher gross margin than custom, so a raise weighted toward optics and switching is a better-quality raise than one weighted toward XPUs. That point was made twice on the call and did not appear to register in the reaction.

CXL architecture and the scale-up optics trajectory

The longest answer of the call covered two topics that were separate a year ago and are now converging. On memory, a five-year-old organic investment is being pulled into AI inference by DRAM scarcity. On optics, the framing shifted decisively from a single acquired technology to a portfolio position across packaging and modulator choices.

Q: "Matt, if you can maybe talk a little bit about what you're architecturally seeing in CXL and how that's evolving, obviously, with the Structera product, and it seems to be a broadening ecosystem around that. And then similar on an architectural perspective, any update on -- you just touched on scale-up optics, the Celestial AI numbers that you've previously outlined. How have those progressed relative to your initial targets?"
— Aaron Rakers, Wells Fargo

A: "Last quarter, the way we framed it was we had the Celestial AI CPO photonic fabric solution, which for reference, was about $150 million for next year. We had actually said that overall scale-up optics inclusive of that was about $300 million as a category, that includes NPO, by the way. And as I said in my prepared remarks, all of that together, scale-up optics is accelerating even further from the number I gave you last quarter. And so, a significant driver of the $1.5 billion raise we have for next year, Aaron, is coming from that area. And it's not just a one-trick pony. ... So we see this as an and, not an or."
— Matthew Murphy, Chairman and CEO

Assessment: The answer restated the May baseline precisely ($150M Celestial inside a roughly $300M category) and then declined to give the new number, which makes the omission conspicuous rather than routine. If scale-up optics is "a significant driver" of a $1.5B raise, the category has multiplied from the baseline given eleven weeks earlier, and the Celestial earnout of up to 22.4 million shares vests against exactly this revenue.

How to model the out-years before the Investor Day

The most direct request for permission to raise out-year numbers came with the arithmetic already done and an acknowledgement that the answer would probably be deferred. Management validated the math without endorsing a number, and confirmed the margin profile would be consistent with the existing custom model.

Q: "Matt, so I want to go back to the Google deal. I think people are just wrestling with this. ... For FY, if you look at the $120 billion over the course of the years, that's about $18 billion a year, and at the $120 billion divided by 6.5. And that's like adding an FY '28, Marvell every year. And I realize that we don't know how much is incremental versus what's already in your guide. But, are we talking like FY '29 and FY '30 big step-ups as a result of this incrementally, because these numbers are huge. ... And is it at the comparable margin?"
— Benjamin Reitzes, Melius Research

A: "When you look at the scale of this, your math is not wrong. And what you can conclude from what I'm saying is, because most of this is comprehended already in next year, the big impact would be in '29 and beyond. ... On the margins, just in general, this is a custom business, and we've got a financial model for custom. We've got a financial model for our standard and merchant products. It would be in line with that, but it's significant. ... you should assume starting in FY '29 and beyond whatever you've modeled previously prior to the warrant for custom numbers definitely goes higher. How big, we'll be happy to show the range of outcomes. But in the context of where Marvell was, Ben, I mean, we were an $8 billion company last year. We just took everybody to $12 billion this year. We haven't even finished that year and $18 next year."
— Matthew Murphy, Chairman and CEO

Assessment: "Your math is not wrong" is the single most consequential sentence on the call, and it is also the one that cuts both ways. It endorses an out-year revenue opportunity of a size the sell-side has not modelled. It also confirms, in the same breath, that the margin profile is the custom model rather than the corporate model, which means the incremental revenue arrives at below-corporate gross margin on top of a mix that is already compressing. The question asked whether it was "at the comparable margin" and the answer was that it is comparable to custom, not to Marvell.

Where the connectivity growth rates now land

Asked to update the segment growth rates given in May, management declined to restate them and instead described the composition of the raise, naming connectivity as the largest single contributor across three sub-businesses.

Q: "I was wondering if you could maybe just give us a little bit of update in terms of the growth rate for connectivity, you expect. I think, Matt, you outlined those growth rates for both this fiscal year, next fiscal year last quarter. Can you maybe just give us an update on where those are landing, given the incremental strength. It sounds like most of that's being driven by that subsegment right now?"
— James Schneider, Goldman Sachs

A: "The way to think about it is the $500 million raise for this year and the $1.5 billion for next year is broad-based. Connectivity clearly is a driver, and I say connectivity at a broader level because underneath that, we spent time talking about scale-up optics, that's one, which is our NPO and CPO and SiGe products. But it's also just the transceiver market for scale-out and optical DSPs going into that segment, that's upsized versus the prior growth rates we talked about. ... And you're right. The connectivity is a bigger bucket, it is probably the largest driver net-net of the $1.5 billion raise. If I just look at all those pieces I mentioned underneath, scale-out, scale-up and switching."
— Matthew Murphy, Chairman and CEO

Assessment: The May framework carried fiscal 2027 interconnect growth of more than 70%, a figure raised twice inside one fiscal year. That number was not restated here, and the substitution of a qualitative "largest driver" for a quantitative growth rate is a step backwards in disclosure at exactly the point where the business is performing best. It is the second time in three questions that the answer traded a number for an adjective.

The 90 basis point gross margin step-down

The final question of the call was the one that mattered most to the tape. It put the margin guide directly against the data center guide and asked whether custom was the sole explanation. The answer confirmed it, and then volunteered the first forward gross margin framing the company has given for fiscal 2028.

Q: "Matt, just wanted to come back to the sort of the near-term guidance. Within data center, I think you guided up greater than 20% sequentially. But with margins coming down 90 basis points quarter-on-quarter, is custom the fastest-growing segment within data center that drives that lower margin? Or is there some other mix going on within data center kind of driving the lower 90 -- or sorry, that 90 basis point lower guidance for the October quarter?"
— Quinn Bolton, Needham & Company

A: "Part of the acceleration story in Q3, as you point out, mix is the primary driver. We've got a strong ramp in custom. ... Net it all out, we see gross -- Q4 gross margins in the same target range as Q3. If we use the back half of the year as a jumping-off point and think about how that looks going into FY '28, that broad-based strength continues. ... Gross margin depends on ultimate mix. My preliminary view is gross margins next year are going to be in a similar range, same range as we're exiting this year. So back half of this year, same range for FY '28. ... You're going to see us entering our target long-term model range, 38% to 40% exiting this year. You'll see us achieve the high end of that range as we progress through FY '28, and we're going to reset that long-term target model here in the coming weeks at the Analyst Day."
— Daniel Durn, Chief Financial Officer

Assessment: This is the most informative answer of the call and the most consequential. It removed gross margin expansion from the model for two fiscal years, in a single unprompted sentence, in a new CFO's first quarter. It also improved on our own May estimate of the fiscal 2028 path by a hundred to two hundred basis points, and pulled the operating margin timeline forward by a quarter. Whether it reads as good or bad news depends entirely on what the reader was previously assuming, and the tape's answer was unambiguous.

What They're NOT Saying

1. The accounting treatment of the Google warrant. Marvell's 10-Q states that its precedent customer warrants "are recognized as a reduction to revenue as qualifying revenues are recognized during the vesting term." The Google warrant is disclosed only as a subsequent event with no grant-date fair value and no mention of revenue recognition. Neither the release nor the call nor any of the nine analyst questions touched it. On the precedent ratio of grant-date fair value to strike price, the contra-revenue could run on the order of 6% of the qualifying revenue stream.
2. Whether the Q3 gross margin guide already contains warrant contra-revenue. The vesting window opens in Q3 of this fiscal year and revenue covered by the agreement is already inside the guidance. A single $500M tranche at the precedent fair-value ratio would carry roughly $31M of contra-revenue, against a guided sequential gross margin step-down of 90 basis points on $3.15B, which is roughly $28M. The company attributed the entire step-down to custom mix and was not asked whether any of it is warrant-related. We do not assert that it is. We note that nobody established that it is not.
3. The new scale-up optics number. Management restated the May baseline exactly, $150M of Celestial photonic fabric inside a roughly $300M fiscal 2028 category, said the category has accelerated "even further" and is "a significant driver" of a $1.5B raise, and gave no revised figure. The Celestial earnout of up to 22.4 million shares vests on cumulative revenue milestones through fiscal 2029 that this number directly determines.
4. The fiscal 2027 interconnect growth rate. The May framework put it above 70%, a figure raised twice inside one fiscal year and treated as a headline. Asked directly for an update, management substituted "probably the largest driver net-net of the $1.5 billion raise." A number that was disclosed twice has become an adjective in the quarter it performed best.
5. The $118.6 million sequential increase in stock-based compensation. Total SBC went from 8.6% of revenue to 11.9% in one quarter. It was not mentioned in the release, not addressed on the call, and not asked about. The Q3 guide implies a partial reversal, which suggests a discrete driver that was never named.
6. Customer concentration, for the third consecutive quarter. One distributor at 44% of revenue and one direct customer at 16% means two relationships carry 60% of the top line, against 50% a year ago. Distributor revenue has drawn level with direct revenue at 50% each, from 41% and 59%. Four customers hold 72% of gross receivables. Every one of those figures appears only in the 10-Q, and no analyst has raised the topic in three quarters.
7. Free cash flow, again. Operating cash flow of $605.5M was disclosed and characterised. Capital expenditure of $126.7M appears only in the cash flow statement. No free cash flow figure, no fiscal 2027 capital expenditure framing, and no discussion of the $8,518.9M of unconditional foundry and assembly commitments, of which $1,829.3M falls due in the remainder of this fiscal year alongside $94.1M of technology services fees, or of the $351.6M of capital expenditure commitments.
8. Why communications and other is guided down low-to-mid teens on both comparisons. A fifth of the revenue base is guided to a double-digit decline sequentially and year over year, explained as the "mix of businesses in this segment," followed by an unquantified "solid sequential recovery" in Q4. It has now gone two consecutive quarters without a question.
9. What replaces the 38% to 40% operating model. The CFO said the long-term target model will be reset at the October 6 Investor Day. Shareholders have underwritten the 38% to 40% framework since 2021, the current valuation depends on it, and no directional indication of the reset was given.

Market Reaction

• Pre-print setup: MRVL closed at $241.45 on August 27, entering the print up 184.1% year to date against the S&P 500's 12.9%, up 222.8% over the trailing twelve months, and up 38.4% over the trailing thirty days. The 52-week closing range into the print was $62.31 to $316.43, so the shares came in 23.7% below their highest close of the year and roughly four times their lowest.
• Reaction session (August 28): Opened at $225.26, a 6.7% gap down, traded a $215.55 to $228.88 range, and closed at $216.62, down 10.3% or $24.83. The S&P 500 fell 0.2% and the Philadelphia Semiconductor Index fell 3.5% on the same session.
• Volume: 49.0 million shares against a 30-day average of 23.5 million, or 2.1 times normal. The print session itself traded 41.0 million.
• Peer context: NVIDIA had reported on August 26 after the close and rose 8.7% on August 27 before falling 4.6% on the 28th; Broadcom fell 0.7%. Roughly a third of Marvell's decline is attributable to sector beta and roughly two thirds to the print.
• Subsequent sessions: The stock fell a further 2.3%, 0.6% and 1.9% over the next three sessions to close at $206.48 on September 2, then recovered 1.1% to $208.83 on September 3. From the print close, the cumulative decline is 13.5% against 4.5% for the Philadelphia Semiconductor Index and a flat S&P 500.

The proximate cause is not in dispute and is unusually easy to isolate. Revenue beat, EPS beat, the Q3 guide beat on both lines, and both fiscal-year frameworks were raised again, for the third consecutive quarter. Exactly one number in the release moved the wrong way: non-GAAP gross margin, guided to 57.5% to 58.5% against 58.9% delivered and 59.4% a year earlier. That is the only line on which the print disappointed, and it is the line the stock traded on.

The deeper cause is what the gross margin guide implied when the CFO extended it. A 90 basis point step-down in a single quarter is a mix event. The same range applied to Q4, and then a preliminary view that fiscal 2028 lands in that band, converts a mix event into a structural ceiling. At the $241.45 close going into the print, the shares carried an enterprise value of roughly 12.4 times the fiscal 2028 revenue the company went on to guide. That is not a multiple priced for a business whose gross margin is flat for two years while its revenue compounds at 50%, because that combination caps the rate at which revenue converts into per-share earnings. The market spent one session repricing that constraint and four more sessions confirming the reprice.

Positioning made the move larger than the news warranted. Marvell entered the print up 38.4% in thirty days and up 184.1% year to date, having roughly quadrupled from the 52-week low. A stock that has already absorbed three consecutive framework raises requires the fourth to be not merely large but larger than the market has extrapolated. This one was large and arrived alongside a constraint, and the constraint was the new information.

The Google warrant compounded the problem through timing rather than substance. Management confirmed that revenue covered by the agreement through fiscal 2028 is already inside existing guidance, so the incremental economics begin in fiscal 2029, while the 58,970,907 warrant shares were issued on August 18. A shareholder holding a fiscal 2028 multiple was handed the dilution mechanism now and the revenue later, and the deferral of any quantification to October 6 removed the possibility of resolving that in the near term.

Published sell-side targets moved higher on the reaction day, with increases reaching into the high $200s, on the same session the stock fell 10.3%. That divergence is the same pattern this coverage flagged in May. Targets are a mechanical function of the framework, which went up; the tape is a function of what marginal capital will pay for the framework, which went down. When those two move in opposite directions on the same information, the tape is the more informative signal.

Street Perspective

Debate: Is a flat gross margin through fiscal 2028 acceptable, or is it the thesis breaking?

Bull view: Gross margin is the wrong line to judge a custom silicon business on. What matters is gross profit dollars and operating margin, and both are compounding faster than revenue: non-GAAP gross profit grew 35.5% year over year and non-GAAP operating margin expanded 180 basis points. Management pulled the 38% to 40% operating model forward by a full quarter and committed to the high end through fiscal 2028. A flat 58% gross margin on $18B of revenue produces far more gross profit than an expanding 60% margin on $12B.

Bear view: Operating leverage is a finite offset with a stated ceiling. Management has capped operating margin at 38% to 40% and reaffirmed it, so once that ceiling is reached, every incremental dollar of custom revenue converts to earnings at the custom gross margin and nothing further. With gross margin flat and operating margin capped, per-share earnings growth reduces to revenue growth minus share count growth, and the share count is the part nobody is modelling.

Our take: The bulls have the better argument on the operating result and the bears have the better argument on the terminal value. We would add that the disclosure itself is the news here, and it is favourable: management framed a fiscal 2028 gross margin roughly a hundred to two hundred basis points above the 56% to 57% we derived in May from its own operating margin target. The stock fell anyway because the market was underwriting expansion, not stability. A business that grows revenue 50% with a stable 58% gross margin and a 40% operating margin is an excellent business. It is not a business that supports paying more for each incremental turn of revenue growth.

Debate: What is the Google warrant actually worth to shareholders?

Bull view: This is the strongest validation the custom franchise has received and the structure is favourable to Marvell in every dimension that matters. Nothing vests without revenue: 240 tranches, one per $500M, with a strike of $206.58 that the buyer must pay. Full vesting requires $120 billion of purchases and would deliver $12.18 billion of exercise proceeds. Trading 6.4% of the company for that revenue stream, with a customer that has every incentive to grow it, is a trade a rational board takes every time. It also removes the single largest bear objection to the custom story, which was that the fiscal 2029 target depended on programs that remained unnamed.

Bear view: The economics are back-ended and the cost is front-ended. Management confirmed the agreement adds nothing to fiscal 2028 revenue while the warrant was issued in August 2026. The company's own accounting policy treats customer warrants as a reduction to revenue, which means the reported top line and gross profit on this stream are lower than the gross bookings, by an amount nobody has quantified. And the deal deepens exactly the concentration the bear case has been pointing at: two relationships already carry 60% of revenue, and this agreement makes one of them structurally larger.

Our take: The trade is good and the disclosure around it is not yet adequate. The vesting structure genuinely aligns interests and the strike is a real payment, not a giveaway. But shareholders cannot currently model the reported economics, because the grant-date fair value has not been disclosed and therefore neither has the contra-revenue. On the precedent ratio from Marvell's own two prior customer warrants, the drag could approach 6% of the qualifying revenue and it lands entirely in gross profit on the lowest-margin product line. That is a question for October 6, and until it is answered the correct posture is to underwrite the revenue opportunity and discount the reported flow-through.

Debate: Has the drawdown created an entry point?

Bull view: The stock has fallen 13.5% in six sessions while the fiscal 2028 revenue outlook rose 9% and our own fiscal 2028 earnings estimate rose 8%. That is a genuine improvement in risk/reward: 32.6 times fiscal 2028 earnings and 24.6 times fiscal 2029 earnings on a company guiding 50% revenue growth next year is not a demanding multiple by AI infrastructure standards, and the October 6 Investor Day is a near-term catalyst at which management has explicitly promised to raise the out-year framework.

Bear view: The multiple has barely moved. At the May print the shares traded at 34 times our fiscal 2028 estimate and 11.4 times enterprise value to fiscal 2028 revenue; today they trade at 32.6 times and 10.7 times. A 13.5% price decline paired with an 8% estimate increase produces a 4% de-rating, which is noise. Meanwhile the share count is going up, the operating model is being rewritten by a CFO with one quarter in the seat, and the catalyst everyone is waiting for is the same event management used to avoid answering nine questions.

Our take: The bears have the arithmetic. The drawdown feels larger than it is because it followed a 38% thirty-day run, and the multiple is essentially where it was when we downgraded. What has genuinely improved is the quality of the disclosure on gross margin and the pull-forward of the operating margin timeline, which is why we are raising the target rather than the rating. Our May revisit level of the high $160s to low $170s sits roughly 18% to 20% below the current price. We would rather buy the Investor Day than the pre-announcement of it.

Model Update

Valuation impact: Raising the twelve-month target to $230 from $215 and holding the rating at Hold. The target is 35.9 times our fiscal 2028 non-GAAP EPS estimate of $6.40 and 27.1 times our fiscal 2029 estimate of $8.50, the same multiple discipline the $215 target applied to the prior estimates. Fair value range moves to $190 to $255 from $175 to $240. Against the September 3 close of $208.83 the target implies a 10.1% twelve-month return, which is consistent with a stock we expect to perform broadly in line with the market rather than ahead of it. The single largest unmodelled item is the Google warrant's contra-revenue, which we have deliberately left out of the estimates because the company has not disclosed the grant-date fair value; if it lands anywhere near the precedent ratio, our fiscal 2029 revenue and EPS estimates are too high.

Thesis Scorecard Post-Earnings

Overall: The operating thesis strengthened again on all four bull pillars, and one bear pillar improved for the first time since it was opened. Against that, gross margin compression moved from inference to guidance, the share count went past its own guide, and a ninth-figure equity claim was added to a capital structure that already carried three. The thesis is unchanged in direction and better documented in both halves: management disclosed more about margin than it ever has, and the filings disclosed more about dilution than the call did.

Action: Hold. One of the three conditions we set in May for revisiting a positive rating has been met, and it is the substantive one: the fiscal 2028 gross margin path was framed roughly 100 to 200 basis points above the 56% to 57% we had derived. That justifies a higher target and a better estimate, and it does not justify a higher rating at 32.6 times fiscal 2028 earnings after a de-rating of roughly 4%. We would move to a positive rating in the high $170s to $185, on the Google warrant's grant-date fair value being disclosed at a level that leaves the custom revenue stream materially intact after contra-revenue, or on scale-up switching being quantified and added to the framework rather than excluded from it for a third consecutive quarter. We would move negative if the reset long-term operating model announced for October 6 lowers the 38% to 40% range, if the diluted share count passes 950 million before the warrant begins vesting, or if fiscal 2028 gross margin is subsequently guided below the 57.5% floor the CFO set.