Marvell MRVL stock prediction: $315 bull vs $160 bear

The consensus reading of Marvell Technology is that the stock is cheap because the Google money has not arrived yet. That is only half right, and the half it gets wrong is the expensive half. Marvell closed at $235.01 on 9 September 2026, up 4.26% on the day and up roughly 252% over twelve months, yet still 28.8% below the $329.88 intraday high it set on 18 June. Our twelve-month Marvell stock prediction is a $315 bull case against a $160 bear case, with a $255 base. The spread is that wide because of something buried in Note 3 of the company’s own quarterly filing: the customer warrants that make the AI story possible are accounted for as a reduction to revenue. Marvell’s growth is real. It simply does not all show up on the top line.

Here is the part no competing analysis has run the arithmetic on. In the quarter ended 1 August 2026 Marvell disclosed two earlier customer warrants, both structured exactly like the Google one and both explicitly “recognized as a reduction to revenue as qualifying revenues are recognized.” The fiscal 2025 warrant carried a grant-date fair value of $54.44 per share against an $87.77 strike — 62.0% of strike. The fiscal 2026 warrant carried $53.02 against $87.00 — 60.9%. Apply that same ratio to the Google warrant’s $206.58 strike and each of its 240 vesting tranches implies roughly $30m of contra-revenue against every $500m Google spends. In other words, on Marvell’s own disclosed methodology, about six cents of every dollar of the headline $120bn Google programme is booked as negative revenue rather than revenue. Marvell has not published the Google warrant’s grant-date fair value; that figure is a FinanceFeeds estimate derived from the two warrants it has published. But the mechanism is not an estimate. It is in the filing.

Key facts

• Last close $235.01, +$9.60 (+4.26%), on volume of 20.48m shares; market capitalisation $206.08bn; 52-week range $65.19–$329.88 — Nasdaq, 9 September 2026
• Q2 FY2027 net revenue $2,739.3m, a record, up 37% year on year from $2,006.1m; GAAP gross margin 53.1%, non-GAAP 58.9% — Form 8-K Exhibit 99.1, 27 August 2026
• Data centre revenue $2,171.5m, or 79% of the total, up from 74% a year earlier; communications and other slipped to 21% — Form 10-Q for the quarter ended 1 August 2026
• A single distributor accounted for 44% of net revenue, up from 34% a year earlier; distributors in aggregate reached 50% of revenue, from 41% — same filing
• Shipments to China were 42% of revenue ($1,161.5m), up from 29%; shipments to the United States fell to 7% from 15% — same filing
• Google warrant over 58,970,907 shares at $206.58, of which 57,610,040 vest on “discretionary purchases” in 240 tranches of $500m — Form 8-K, filed 19 August 2026
• NVIDIA holds $2.0bn of Series A convertible preferred struck at roughly $91.84, convertible into about 21.8m common shares — worth about $5.1bn at Tuesday’s close, a paper gain of some $3.1bn — same 10-Q, FinanceFeeds calculation

What the quarter actually said, and why the stock still fell

Marvell’s second quarter of fiscal 2027 was, on the numbers, the best in the company’s history. Revenue of $2,739.3m came in $39.0m above the midpoint of its own guidance. GAAP net income more than doubled the prior year’s tally at $308.0m against $194.8m, and diluted GAAP earnings reached $0.33 from $0.22. Non-GAAP earnings were $0.94. Operating cash flow was $605.5m for the quarter and $1,244.3m for the half. Research and development ran at $741.1m, or 27.1% of revenue, up 42.8% year on year.

And the shares fell 10.28% the next session, from $241.45 to $216.62, then kept sliding to $206.48 by 2 September. That close matters more than it looks: it is ten cents below the $206.58 strike on the Google warrant. For one session the single largest piece of alignment between Marvell and its marquee customer was under water. The stock has since recovered to $235.01, but the market has now demonstrated that it will price this equity below the level at which the Google relationship is supposed to compound.

The drop was not about the print. Part of it was timing: on the results call Murphy told analysts that revenue from the programmes covered by the Google agreement, through fiscal 2028, “is already reflected in the overall custom revenue target” — the market had wanted the $120bn headline to be incremental to guidance, and was told it largely is not until fiscal 2029. The rest was the shape of the guide. Marvell told investors to expect third-quarter revenue of $3.150bn plus or minus 5%, with non-GAAP gross margin of 57.5% to 58.5% — below the 58.9% it had just delivered — and non-GAAP diluted earnings of $1.10. Run the incremental arithmetic and the mix shift becomes explicit. Between the first and second quarters, revenue grew $321.5m and GAAP gross profit grew $194.8m: an incremental gross margin of 60.6%, comfortably above the corporate average. Between the second quarter and the guided third, revenue grows $410.7m while non-GAAP gross profit grows roughly $213.5m at the midpoint — an incremental margin of 52.0% against a 58.9% base. Marvell’s next slice of growth is guided to arrive nearly nine points less profitably than the slice that came before it. That is the mix shift into custom silicon showing up in the outlook before it shows up in the results, and it is a large part of why a record quarter sold off. Our full breakdown of why the stock fell on a beat-and-raise covers the session in detail.

Chairman and chief executive Matt Murphy was explicit about where the payoff sits. “AI-related bookings remain exceptionally robust, and we expect our revenue growth to accelerate further through the remainder of fiscal 2027,” he said in the results release, pointing to “a significant acceleration in our Custom business beginning in the second half of fiscal 2027” and to the company’s investor day on 6 October 2026. The acceleration is a promise with a date attached. The margin compression is already in the guide.

The warrant structure, read properly

The Google agreement is the centre of the equity story, and the filing repays careful reading. Marvell and Google signed on 29 July 2026 covering “a comprehensive range of custom silicon programs that attach to the TPU ecosystem” — inference accelerators, storage controllers, network interface controllers, memory interface controllers and near-memory compute. On 18 August, Marvell issued Google a warrant over 58,970,907 shares at $206.58, exercisable to August 2033.

Only 1,360,867 of those shares are time-based, vesting in equal quarterly instalments over the first year. The other 57,610,040 vest, in the filing’s own words, “based on discretionary purchases” — 240 equal tranches, one for each $500m of custom-product revenue, running from the third quarter of fiscal 2027 to the end of fiscal 2033. That single adjective carries the whole risk. Google has bought an option on Marvell’s equity at a price it helped set, and has undertaken no minimum spend to earn it. Marvell’s own quarterly filing describes the warrant only as “eligible for vesting… upon meeting certain revenue milestone conditions or time-based conditions.” There is no floor.

Set that against the two disclosed predecessor warrants: fiscal 2025, over 4.2m shares at $87.77 (1.2m vested by 1 August 2026); fiscal 2026, over 1.0m shares at $87.00 (none vested). Both are recognised as reductions to revenue, and both carried grant-date fair values around 61–62% of strike. The Google warrant is 14 times the size of both combined and begins vesting in the quarter Marvell is guiding to right now. The first consequence is that reported custom revenue will be flattered less than gross purchase figures imply. The second is that the dilution accrues fastest precisely when the shares are working, because Google’s incentive to keep buying rises with the price.

Marvell is not alone in paying for demand with equity; the instrument has spread across the AI supply chain, as our coverage of OpenAI’s in-house accelerator programme and its Broadcom ties shows. What separates Marvell’s version is disclosure quality: 240 tranches at $500m apiece is an unusually precise statement of ambition, and of what shareholders give up to chase it.

The concentration nobody is discussing

Ask most investors where Marvell’s concentration risk sits and they will say Google, or Amazon, or hyperscalers generally. The filing says something different, and more uncomfortable.

Net revenue by customer type has flipped in twelve months. Direct customers fell from 59% of revenue to 50%; distributors rose from 41% to 50%. Within that, a single unnamed intermediary — “Distributor A” — accounted for 44% of Marvell’s net revenue in the quarter, up from 34% a year earlier, while the largest direct customer held flat at 16%. Five customers represented 72% of gross accounts receivable at quarter end. The ten largest represented 82% of revenue in fiscal 2026.

The geographic disclosure moves the same way. Shipments to China reached $1,161.5m, or 42% of net revenue, against 29% a year earlier. Taiwan was 17%, Singapore 12%, and the United States fell to 7% from 15%. Marvell attaches a fair caveat: destination of shipment “is not necessarily indicative of the geographic location of the Company’s end customers,” because most China shipments relate to non-China customers with contract manufacturing there. But the caveat does not neutralise the exposure. A thirteen-point swing toward China in the physical shipment of AI infrastructure silicon is export-control surface area that grew materially in one year, whoever ultimately owns the boxes.

Balance sheet, buybacks and the NVIDIA position

Marvell finished the half with $3,932.8m of cash against $4,962.9m of long-term debt and no short-term debt, having repaid $500m. It bought back $400.0m of stock and paid $107.7m of dividends in the half, while stock-based compensation ran to $533.8m — up from $295.7m a year earlier. The buyback did not cover the dilution.

Two acquisitions closed in February 2026: Celestial AI, a photonic-fabric interconnect specialist, and XConn Technologies, which brought PCIe and CXL switching silicon into the UALink scale-up effort. The cash-flow statement shows $1,270.9m of cash paid net of cash acquired, but the bulk of the consideration was stock: goodwill rose from $11,062.2m to $13,873.9m. A $433.7m increase in the fair value of contingent consideration ran through the half — earnouts on private AI assets reprice upward when the sector does.

The most striking line, though, is the preferred stock. On 31 March 2026 Marvell sold 2.0m shares of Series A convertible preferred to NVIDIA for $2.0bn, convertible at roughly $91.84 into about 21.8m common shares. At $235.01 that position is worth approximately $5.1bn — an unrealised gain of some $3.1bn, or about 156%, in a little over five months. NVIDIA is simultaneously the dominant merchant supplier of AI compute, the counterparty whose customers fund Marvell’s custom-silicon alternative, and now Marvell’s largest preferred holder. Marvell’s risk factors concede the point, warning that because the preferred “was issued to a single significant holder, the concentration of ownership could increase NVIDIA’s influence over matters submitted to our stockholders.” For a business that converts AI demand into GAAP profit at scale, see our breakdown of NVIDIA’s latest quarter; for the foundry layer beneath both, our TSMC bull and bear analysis.

The $315 bull case and the $160 bear case

Every scenario below is expressed on one consistent metric so they can be compared honestly: market capitalisation as a multiple of the annualised third-quarter revenue guide of $3.150bn, or $12.6bn. At $235.01 and roughly 877m shares outstanding, Marvell trades at 16.4 times that run-rate.

The bull case to $315 is 34.0% above the current price and implies a market capitalisation near $276bn. Critically, it does not require multiple expansion. Held at today’s 16.4 times, $315 requires revenue of roughly $16.9bn — about a third above the annualised Q3 guide, which is squarely what a custom-silicon acceleration through fiscal 2028 would produce if the Google programmes convert on the schedule the warrant contemplates and the connectivity business keeps compounding. It is also comfortably inside the sell-side range: broker data polled by S&P Global puts the consensus at “Strong Buy” with an average target of $284.80 across 44 analysts, a low of $143 and a high of $400 — KeyBanc’s John Vinh moved to $400 the day after the print, and B. Riley sits precisely on our $315. Management has since put a figure against it: on the results call, chief financial officer Daniel Durn guided fiscal 2028 revenue to roughly $18bn, about 50% growth, with data centre up more than 60%. Our bull case does not require that target to be hit in full. It needs execution, not re-rating.

The bear case to $160 is 31.9% below the current price, implies about $140bn of market value, and works out at 11.1 times the same run-rate. It requires no collapse in AI capital expenditure and no failure of the Google relationship. It requires only that the market decide a 52% incremental gross margin, a 44% distributor, 42% China shipment exposure and a contra-revenue drag on the flagship programme do not deserve a 16-times multiple. Marvell proved this year that it can do exactly that: the shares fell from a $316.43 close on 4 June to $163.40 on 29 July, a 48% decline in eight weeks with no change in the underlying business. Our $160 sits just below that trough — and is not the most bearish view on the street, since the lowest published target in the S&P Global poll is $143. The base case is $255, 8.5% above spot at 17.8 times — the stock grinding back toward its August high while the custom ramp is proved out quarter by quarter.

The invalidation levels are specific. The bull case is invalidated on a weekly close below $206.58, the Google warrant strike. Below that line the warrant is out of the money, Google’s incremental purchasing incentive weakens, and the market is explicitly refusing to capitalise the programme; the stock already closed at $206.48 on 2 September, so this is a tested level rather than a theoretical one. The bear case is invalidated on a weekly close above $255, which would put the shares back above their post-earnings range and signal that the 6 October investor day has been front-run. Between those two lines, this is a stock with no earnings floor to catch it — GAAP diluted EPS of $0.33 on a $235 share price is not a valuation anchor — and a shareholder register in which its largest competitor holds a $3.1bn paper gain.

Three predictions. First, the 6 October investor day, not the next earnings print, is the binary event: management has deferred the custom-silicon roadmap to it, and the shares have nothing else to trade on for four weeks. Second, the third quarter is the first to carry Google warrant vesting, so expect the gap between Google’s purchases and Marvell’s reported custom revenue to become a talking point once the contra-revenue charge is quantified. Third, if Distributor A’s share rises again next quarter it will overtake the Google narrative, because a channel intermediary at 44% and climbing is a working-capital and credit exposure, not merely a commercial one. As with AMD’s own bull and bear setup, the range is wide because the outcome is genuinely undetermined — an argument about position size, not direction.

Frequently asked questions

What is the Marvell (MRVL) share price today?

Marvell closed at $235.01 on 9 September 2026, up $9.60 or 4.26%, on volume of 20.48m shares. Market capitalisation was $206.08bn. The stock is up roughly 252% over twelve months but sits 28.8% below its 52-week intraday high of $329.88, set on 18 June 2026. The 52-week low is $65.19.

What is a realistic MRVL price prediction for the next 12 months?

Our range is $315 on the bull case, $255 on the base case and $160 on the bear case. At today’s 16.4 times the annualised third-quarter revenue guide, $315 requires roughly $16.9bn of revenue rather than a higher multiple. The $160 bear case assumes no revenue failure at all — only a de-rating to 11.1 times on margin and concentration concerns.

What did Marvell report in Q2 FY2027?

Record net revenue of $2,739.3m, up 37% year on year and $39.0m above the guidance midpoint. GAAP gross margin was 53.1% and non-GAAP 58.9%. GAAP diluted EPS was $0.33 and non-GAAP $0.94. Data centre revenue was $2,171.5m, or 79% of the total, up 46% year on year. Operating cash flow was $605.5m.

How does the Google warrant actually work?

Marvell issued Google a warrant over 58,970,907 shares at $206.58 on 18 August 2026, expiring August 2033. Only 1,360,867 shares vest on time. The remaining 57,610,040 vest on discretionary purchases in 240 equal tranches, one for every $500m of custom-product revenue, from the third quarter of fiscal 2027 through fiscal 2033. Full vesting implies $120bn of purchases. Google has no minimum spend obligation.

Why does the warrant reduce Marvell’s reported revenue?

Marvell’s two earlier customer warrants are, in the filing’s language, “recognized as a reduction to revenue as qualifying revenues are recognized.” Those carried grant-date fair values of roughly 61–62% of strike. Applying that ratio to the Google warrant implies about $30m of contra-revenue per $500m tranche, or roughly six cents in every dollar. Marvell has not disclosed the Google warrant’s fair value; that figure is a FinanceFeeds estimate.

Who is Marvell’s biggest customer?

Marvell does not name them. The 10-Q for the quarter ended 1 August 2026 discloses that one distributor accounted for 44% of net revenue, up from 34% a year earlier, and one direct customer accounted for 16%. Distributors as a whole were 50% of revenue, up from 41%. The ten largest customers were 82% of fiscal 2026 revenue.

This article is editorial analysis and is not investment advice. Financial figures are drawn from Marvell’s SEC filings — the Form 10-Q for the quarter ended 1 August 2026, the Form 8-K and Exhibit 99.1 of 27 August 2026, and the Form 8-K of 19 August 2026 — with market data from Nasdaq as of the close on 9 September 2026 and broker target aggregation polled by S&P Global via StockAnalysis. Incremental-margin and warrant fair-value figures are FinanceFeeds calculations from disclosed data and are identified as such. Price targets are FinanceFeeds estimates and may not be achieved. Featured image: the grid computing centre at Fermilab, photograph by ENERGY.GOV, public domain, via Wikimedia Commons.