Marvell Stock Price Forecast - MRVL at $78 AI Reset with Optical Upside Back Toward $170+

Marvell Stock Price Forecast - MRVL at $78 AI Reset with Optical Upside Back Toward $170+

Marvell Technology sits around $78.6 per share, well under its $112.5 peak, as investors weigh lost Amazon Trainium sockets against 75% AI data-center revenue, 20–25% growth to ~$10B sales, and the Celestial AI/XConn optical interconnect ramp | That's TradingNEWS

TradingNEWS Archive 2/14/2026 12:12:51 PM
Stocks MRVL AMD NVDA TSM

Marvell Technology (NASDAQ:MRVL) overview and current pricing

Marvell Technology Inc. trades around $78.61 with after-hours at $78.70 on NASDAQ:MRVL, inside a 52-week range of $47.09 to $112.50 and a market cap near $68.60B in USD terms. The trailing P/E is roughly 27.5x with a small 0.31% dividend yield, even after the stock lost more than 25% over the last year while other AI infrastructure names re-rated higher. At these levels the market is treating Marvell as a damaged growth story rather than as a core AI plumbing name despite its heavy exposure to data center and AI spending.

Business mix and AI data center dependency at Marvell Technology

Over the last twelve months roughly three quarters of Marvell’s revenue now comes from data center customers, making NASDAQ:MRVL an AI infrastructure play rather than a generic semiconductor name. Within that data center mix Marvell is exposed to custom AI and cloud ASICs, high speed optical components, retimers and switching silicon that sit directly in AI clusters. For the coming fiscal year management and analysts expect total revenue to move from about $8.20B in FY-2026 to roughly $10.00B in FY-2027, implying 20% to 25% growth in USD. Data center revenue is expected to grow around 25% while the rest of the portfolio, including enterprise and carrier infrastructure, is modeled closer to 10%, meaning the AI cycle drives the bulk of the expansion. Importantly, management already reports purchase order backed visibility into key custom data center programs through FY-2027, so the $10.00B revenue target is not a soft pipeline claim but supported by committed orders across AI networking, optics and custom silicon.

Valuation reset versus AI peers and Marvell’s own history

Despite that growth NASDAQ:MRVL now trades at only about 22x to 24x forward earnings, materially below Broadcom’s 32x range and below Marvell’s own ten year average forward multiple near 27x. On FY-2027 EPS estimates around $3.50 the current $78 to $79 price implies roughly 22x to 23x earnings in USD terms. On FY-2028 EPS scenarios between about $4.25 and $4.77 the implied multiple compresses into the high teens. Historically Marvell rarely traded below 20x forward earnings for long even when growth was slower and AI was not driving demand. Today the stock is priced more like a cyclical connectivity vendor than a company delivering around 38% year over year data center growth with non-GAAP gross margins near 60%. To justify a sustained move under 20x forward earnings from here you would need either a clear break in the AI spending story or a severe Marvell specific execution error. The current discount shows that the market is already pricing in meaningful risk around customer concentration and custom compute rather than giving full credit for the AI infrastructure exposure.

Custom compute exposure and Amazon Trainium risk for NASDAQ:MRVL

The main overhang on Marvell Technology stock comes from its custom compute and Amazon relationship. Marvell previously designed Trainium2 for Amazon Web Services, tying the company directly into Amazon’s AI ambitions. For Trainium3 and Trainium4 however reports indicate Marvell lost the socket to lower cost competitor Alchip after packaging, schedule and design issues on Trainium2 and a chiplet based Trainium3 proposal that did not win the bakeoff. External market share work now projects Marvell’s slice of the AI server compute ASIC market to slip from roughly 12% to about 8% by 2027 while Broadcom is expected to retain around 60%. With Amazon planning AI driven capital expenditures near $200B in 2026 the loss of future Trainium generations materially caps Marvell’s upside in custom XPUs. The structural risk is broader than one project. Hyperscalers are maturing their internal silicon teams, Taiwanese design houses undercut on price and only a handful of partners capture the highest value sockets with durable margins. If that trend extends beyond AWS Marvell’s role in custom compute could shift from core design partner to more commoditised physical design and packaging, which would compress margins and weaken the long term narrative the market once attached to NASDAQ:MRVL.

Optical networking pivot with Celestial AI and XConn as growth engine

Marvell’s response has been to pivot aggressively into optical interconnect and memory fabric, backed by two large acquisitions funded with about $3.79B in cash and stock. The Celestial AI deal, costing roughly $3.25B, brings Photonic Fabric technology designed for co-packaged optics and high bandwidth optical chiplets that attack the memory wall inside AI clusters by moving data optically rather than over copper. Management expects Celestial AI to reach a $500M annualized revenue run rate by the fourth quarter of FY-2028 and to scale toward $1.00B in FY-2029, all in USD terms. The XConn Technologies acquisition, around $540M, adds PCIe Gen6 and CXL switching silicon positioned at the heart of memory disaggregation and coherent interconnect. Together, Celestial’s photonic fabrics and XConn’s CXL capabilities let Marvell design an end to end optical scale up fabric that can decouple memory from compute across racks while maintaining near on package performance. That moves Marvell from selling discrete DSPs and switches to architecting the entire data movement topology of AI data centers. If hyperscalers adopt this architecture as they hit power and distance limits with copper, NASDAQ:MRVL shifts from a component supplier multiple toward an infrastructure apex multiple. Analysts building that scenario apply around 37.5x to FY-2028 EPS near $4.77 and reach indicative targets around $179 per share, more than 100% upside from the high $70s. That upside depends on Celestial AI and XConn hitting their volume timelines by 2027 to 2029 and on hyperscalers choosing Marvell’s optical roadmap at scale, but it explains why these deals are offensive moves into the bottleneck layer rather than purely defensive revenue patches.

Margins, earnings power and capital returns at Marvell Technology

On current numbers Marvell’s profitability remains healthy despite the stock’s de-rating. Non-GAAP gross margin sits just under 60%, supported by high value networking, optics and custom AI silicon. Non-GAAP operating margin has moved into the mid 30s, showing operating leverage as AI revenue scales faster than operating expenses. Stock based compensation still depresses operating margin by roughly eight percentage points, but share repurchases have more than offset this and net share count has fallen rather than exploded, which protects per share earnings power. With FY-2026 revenue around $8.20B and FY-2027 forecast near $10.00B EPS growth should run ahead of revenue growth as fixed costs spread over a larger base. The dividend yield of 0.31% is negligible, so the real capital return mechanism remains buybacks, which also help absorb dilution from the Celestial AI and XConn acquisitions. At the same time the acquisitions add about $75M per year of non-GAAP operating expenses, split roughly $50M from Celestial AI and $25M from XConn, which creates a near term drag until the revenue ramps land. The margin and earnings trajectory will therefore depend on management holding gross margin near or above 59% while the new optical businesses scale into their 2028 and 2029 targets.

 

Macro backdrop, dollar and rates context for NASDAQ:MRVL

The macro backdrop currently supports higher duration, AI heavy names like NASDAQ:MRVL if the company executes. Recent US data show headline CPI around 2.4% year on year with core inflation near 2.5%, and markets now price roughly 50 to 60 basis points of Federal Reserve rate cuts in 2026 rather than further hikes. The dollar index has softened after the latest inflation reports and US Treasury yields have eased from peak levels, both of which support valuations for growth stocks tied to long dated AI infrastructure demand. At the same time global AI and cloud capital expenditure is tracking near $700B in 2026 with realistic scenarios for cumulative spending in the trillions of USD by the end of the decade. Marvell sits directly in that flow through data center networking, custom ASICs and optical fabrics. If the Fed maintains a gradual easing path and AI capital budgets remain intact, the macro environment will act as a tailwind for Marvell’s earnings multiple rather than a headwind. The main macro risk would be a sharp slowdown in AI capex after the current buildout, which would force the market to revisit growth assumptions across the entire AI hardware stack, including Marvell Technology stock.

Key downside risks for Marvell Technology stock investors

The upside story around optics and AI networking is balanced by several hard risks that investors cannot ignore. The first is continued erosion in custom compute share. Losing Trainium3 and Trainium4 after issues on Trainium2 shows that Marvell can be out executed on high profile AI sockets and that hyperscalers are willing to move business to competitors like Alchip when timelines and packaging do not meet expectations. If similar losses occur at other cloud customers Marvell’s role in XPUs will shrink to a secondary, lower margin position. The second is customer and distributor concentration. One distributor accounts for about 38% of revenue and one hyperscaler client, widely understood to be Amazon through its “Client A” designation, provides roughly 11% of direct revenue. That level of concentration gives large customers bargaining power on pricing and roadmap decisions that can pressure margins or volumes if relationships deteriorate. The third risk is execution on Celestial AI and XConn. Together they add around $75M in annual operating expenses immediately, with meaningful revenue only expected to arrive closer to FY-2028 and FY-2029. Any slip in standards adoption for co-packaged optics or CXL, or any delay in customer deployments, would create a period where costs rise faster than revenue. The fourth risk is the AI capex cycle itself. Current plans assume more than 20% annual growth in AI infrastructure spending for several years. If hyperscalers enter a digestion phase after 2025 and 2026, both growth and multiples for AI hardware suppliers could compress sharply. In that scenario NASDAQ:MRVL could trade below 20x forward earnings for a sustained period and re-test price levels in the high $60s even without a fundamental collapse.

Buy, sell or hold – where NASDAQ:MRVL stands now

After the drawdown it is clear that NASDAQ:MRVL no longer trades at AI bubble levels. At around $78.61 per share and roughly 22x to 24x forward earnings, you are paying for a company with about $8.20B of current revenue, a credible path to roughly $10.00B next year, non-GAAP gross margin near 60% and data center growth close to 38% year over year. The stock has already absorbed the first wave of disappointment around Amazon’s Trainium roadmap and the initial fear that custom ASIC share could erode. At the same time Marvell has repositioned itself as a central player in the next bottleneck of AI infrastructure, namely optical interconnects and memory disaggregation, backed by Celestial AI and XConn. On realistic FY-2028 EPS in the mid-$4 range, current prices offer mid-teens implied multiples two to three years out, with upside to the $150 to $180 region if the market eventually awards Marvell a premium AI infrastructure multiple as those optical businesses scale. The trade off is volatility and execution risk. From a pure data and valuation standpoint the stock aligns more with a Buy than a Sell, but only for investors prepared to hold through potential drawdowns back into the low $70s or even high $60s if AI sentiment or Marvell execution missteps trigger another leg of de-rating.

That's TradingNEWS