Marvell Stock Price Forecast – Can NASDAQ:MRVL Climb From $80 to the $120–$175 Target Range?

Marvell Technology (NASDAQ:MRVL) – AI data-center pivot from $80.23

Current positioning of NASDAQ:MRVL – price, multiples and trading profile

Marvell Technology NASDAQ:MRVL last closed around $80.23, down $2.87 on the day for a 3.45% decline, with after-hours trading nudging it to about $80.31. The stock is sitting well above its 52-week low of $47.09 but meaningfully below its $127.20 high, giving a market cap of roughly $68.0B. On fundamentals, the equity trades at a P/E of 28.06x, with a token 0.30% dividend yield and heavy liquidity around 14.41M shares in average daily volume. The intraday range at $79.72–$82.02 shows buyers starting to defend the high-70s, but the broader picture is a name priced as a structural AI infrastructure play, not as a cheap cyclical.

Q3 FY2026 results – $2.07B revenue, $0.76 EPS and the start of a smoother earnings curve for NASDAQ:MRVL

In fiscal Q3 2026, NASDAQ:MRVL delivered the kind of quarter that justifies a premium multiple. Revenue reached $2.07B, up 36.18% year-over-year, while adjusted EPS climbed to $0.76, a surge of 76.7% YoY. Both lines beat consensus: top line by about 0.44%, bottom line by roughly 3.01%. The tape reaction was chaotic: shares initially dropped more than 13% on the print, then reversed and now trade around 14% above the pre-earnings level. That volatility reflects repositioning around the forward story rather than the quarter itself. The key change is structural: historically, Marvell’s earnings profile was lumpy, with roughly a coin-flip chance of positive EPS in some years. Now consensus expects FY2026 EPS to jump more than 80%, followed by a projected ~26.4% EPS CAGR from FY2027 to FY2032, turning NASDAQ:MRVL into a more predictable AI-driven earnings compounder instead of a boom-bust semi name.

Shift to data-center AI infrastructure – 73–75% of revenue tied to cloud and AI for NASDAQ:MRVL

Following the divestiture of its automotive business, NASDAQ:MRVL has essentially recast itself as an AI data-center infrastructure vendor. Management indicates that 73–75% of revenue now comes from data-center customers, with the rest from legacy segments. That concentration is intentional. The company sees the data-center franchise as its primary long-term growth engine, underpinned by AI training and inference demand, networking upgrades and memory disaggregation. Management’s internal roadmap points to data-center revenue growing more than 25% in FY2027, then accelerating further to about 40% growth in FY2028 as new AI programs ramp and recent acquisitions fully integrate. In other words, the company is now heavily levered to AI infrastructure capex cycles; as long as hyperscalers are spending to scale out and scale up AI clusters, Marvell’s core mix should continue to expand.

Celestial AI and Photonic Fabric – dismantling the “Memory Wall” and expanding NASDAQ:MRVL’s AI TAM beyond $10B

The acquisition of Celestial AI is the central 2026–2029 catalyst for NASDAQ:MRVL. Traditional AI clusters connected XPUs within a rack using copper links in a scale-out architecture. That model is hitting hard physical limits on bandwidth, power and thermal load. Hyperscalers are now shifting toward multi-rack “scale-up” fabrics that tie together hundreds of accelerators across racks, which copper cannot handle efficiently. Celestial AI’s Photonic Fabric addresses this directly. Its photonic chiplets replace electrical signals with light, delivering around a 10x increase in bandwidth density versus current state-of-the-art scale-out ports, while also offering nanosecond-class latency and meaningfully superior power efficiency. That architecture allows memory to be pooled and disaggregated, so hundreds of XPUs can access shared memory banks across racks, exactly what large-parameter LLM training demands. Industry forecasts see the merchant scale-up switch market approaching about $6B revenue by 2030, with optical interconnect generating similar content per deployment. That combination pushes the addressable interconnect and scale-up TAM beyond $10B, and Celestial’s Photonic Fabric is designed to sit at the center of that spend.

Celestial AI revenue path – from zero today to a $1B run-rate and margin-accretive growth for NASDAQ:MRVL

Management has put concrete numbers behind Celestial AI for NASDAQ:MRVL. The company expects Celestial to reach roughly a $500M annualized revenue run-rate by Q4 FY2028, then double to around $1B by Q4 FY2029. For a business with an enterprise value in the $70B area, $1B may not sound transformational at first glance. The real impact is that the revenues are expected to be margin-accretive and sit in a high-growth, underpenetrated TAM, strengthening Marvell’s AI data-center position and smoothing its earnings profile. Importantly, the regulatory process has not become a bottleneck. The FTC granted early termination of the Hart-Scott-Rodino waiting period for the roughly $3.25B transaction, enabling the deal to close ahead of original expectations. Celestial also arrives with tangible customer traction: it already holds a design win at a very large hyperscaler that plans to deploy Photonic Fabric chiplets in its custom XPUs and switches. That anchors the technology in real AI roadmaps instead of leaving it as speculative IP.

Custom silicon and XPU attach – quadrupling, doubling, and set to double again for NASDAQ:MRVL

Custom silicon is where AI dollars become highly visible for NASDAQ:MRVL. Over the last several years the growth trajectory has been steep: custom revenue quadrupled from calendar 2023 to 2024, then doubled again from 2024 to 2025. For the next fiscal year, management guides to custom growing around 20%, which is a normalization after the recent explosions. Beyond that, the plan is aggressive again: in FY2028, Marvell expects the custom business to double off FY2027 levels. That path is supported by a broader socket base. Management cites more than 20 distinct XPU and attach designs in the pipeline, including engagements at 2nm and A16 nodes, aligned with hyperscaler deployments. High-volume attach products such as SmartNICs and CXL controllers are ramping, and the company sizes the XPU opportunity at approximately $10B of TAM. If that roadmap holds, custom silicon becomes a multi-year AI compounder instead of a single-customer lever.

Amazon exposure – Trainium uncertainty and concentrated client risk for NASDAQ:MRVL

The main crack in the custom narrative is hyperscaler concentration, especially Amazon. External research has suggested that NASDAQ:MRVL may have lost design work for Amazon’s Trainium3 and Trainium4 AI chips to a competitor, implying that current guidance could be leaning on continued Trainium2 orders rather than a clean transition to newer generations. Management has publicly rejected that view, stating it has not lost any Amazon business and that its forward projections already assume next-generation XPU programs. The fact that such a dispute is being fought in public highlights the risk profile: a single large customer can materially move the revenue line for custom silicon. If Amazon or any other major hyperscaler delays a program, scales back orders, or re-allocates designs, growth in this segment can shift sharply. Given the valuation, the equity will not be forgiving if the custom curve deviates meaningfully from “plus 20% then another double in FY2028.”

XConn acquisition – building a full CXL value chain for NASDAQ:MRVL

The XConn Technologies transaction plugs a critical hole in the CXL stack for NASDAQ:MRVL. Marvell is paying about $540M, structured as 60% cash and 40% stock, with the stock portion equating to roughly 2.5M shares. XConn brings high-end CXL switches, which sit at the core of memory pooling and device-sharing topologies. Marvell already had CXL memory expansion controllers via its earlier Tanzanite acquisition and a broader PCIe/CXL interface portfolio. With XConn, the company can now offer both CXL controllers and CXL switches, effectively covering almost the entire CXL value chain from endpoints to the switching fabric. That aligns with Marvell’s 2022 positioning, where management described future cloud data centers as fully disaggregated architectures built on CXL, demanding far higher levels of high-speed interconnect integrated with compute, memory, and storage silicon.

CXL memory pooling economics – DRAM inflation, 20–30% power savings and adoption catalysts for NASDAQ:MRVL

The macro backdrop for CXL is rapidly turning favorable for NASDAQ:MRVL. DRAM prices surged by 50–60% in Q1 2026, and tight supply is expected to keep pricing elevated into 2027. Under that regime, traditional practice—over-provisioning local DRAM at each CPU or GPU for worst-case workloads—is becoming prohibitively expensive. CXL memory pooling allows CPUs, GPUs and accelerators to tap into a shared memory pool, requesting resources dynamically based on actual workload requirements. Simulation work indicates that CXL architectures can cut memory power consumption by about 20–30%, since memory is no longer powered at maximum levels per node around the clock. A few years ago, critics argued that the cost of CXL cabling and switching hardware would outweigh the benefits, but that view assumed much lower DRAM prices. With current memory inflation and AI workloads stressing capacity, the economics have shifted. If CXL becomes the default way to handle memory in AI data centers, the combination of Tanzanite controllers and XConn switches positions Marvell to be one of the principal beneficiaries.

Competitive field in CXL and interconnect – where NASDAQ:MRVL has to win and where it lags

The CXL and AI interconnect market is competitive. Players such as Astera Labs and Rambus are pursuing their own CXL controller and connectivity strategies, and initiatives like UALink are defining standards for pooled memory and accelerator fabrics. NASDAQ:MRVL is effectively betting that the combination of CXL controllers, CXL switches, photonic interconnect from Celestial AI, and established Ethernet and switch silicon can produce a compelling platform for hyperscalers. The company is not currently the default market leader across all of these areas; peers like Broadcom maintain strong positions in networking and AI connectivity. The upside case is straightforward: if Marvell manages to become a top-tier CXL solution provider by integrating these pieces effectively, it captures a central role in the future AI data-center stack. If execution lags and rivals secure the dominant share of CXL deployments, the recent acquisitions will not yield the returns implied in the current multiple.

China exposure and macro sensitivity – tariff and geopolitical risk for NASDAQ:MRVL

On the geographic side, risk concentration is significant. As of FY2025, more than 40% of revenue for NASDAQ:MRVL is associated with China and related channels, which dramatically increases sensitivity to tariffs, export controls and broader geopolitical decisions. Additional U.S. restrictions on advanced chipset exports, new tariffs on China-linked hardware, or local policy shifts that affect ordering patterns can all compress revenue and margins. If Marvell cannot push cost increases through to OEMs and cloud customers in the form of higher ASPs, gross margin will take the hit. The equity is currently priced on the assumption that data-center demand and margin expansion offset these macro frictions; any sustained deterioration in China-related revenue trends would force the market to cut both growth expectations and valuation multiples.

Valuation of NASDAQ:MRVL – high relative to the sector, reasonable versus its own growth curve

At around $80.23, NASDAQ:MRVL trades at a P/E near 28.06x current earnings and forward P/S of about 8.6x, with forward EV/Sales roughly 8.8x. That compares to a sector median P/S and EV/Sales around 3.5x, so Marvell commands more than 2x the sector revenue multiple. However, the growth profile is also significantly stronger. The market is underwriting 80%+ EPS growth in FY2026 and a forward EPS CAGR of roughly 26.4% through FY2032. On that basis, the PEG ratio stands around 0.74, which is not demanding for a high-growth AI infrastructure name if the company delivers on its roadmap. Several internal valuation exercises suggest reasonable upside from here. One framework assumes that FY2027 EPS gets revised about 10% higher as Celestial AI, custom silicon and CXL contributions come through. Applying a 30x P/E—roughly in line with the current level and below some AI leaders—to that FY2027 EPS yields a price around $118–$120, implying roughly 42% upside over the next 16–18 months. A more aggressive scenario, consistent with prior internal work, points to a FY28 target near $175, contingent on Celestial hitting a $1B run-rate, CXL adoption accelerating, and data-center growth tracking the >25% in FY2027 and ~40% in FY2028 range guided by management.

Valuation risk – what happens if NASDAQ:MRVL misses its aggressive AI roadmap

The premium valuation of NASDAQ:MRVL leaves little margin of safety if the AI roadmap slips. Revenue multiples at 8.6x P/S and 8.8x EV/Sales, against a 3.5x sector median, assume that Marvell’s transition into an AI infrastructure compounder is successful. If Celestial AI’s targets of $500M run-rate by Q4 FY2028 and $1B by Q4 FY2029 are delayed, if custom silicon growth stalls below the 20% + another double trajectory, or if CXL deployment timelines stretch beyond the current expectation, the market will cut its growth assumptions. When that happens on a premium multiple, the adjustment is not limited to earnings estimates; the multiple itself also compresses. That combination—lower EPS forecasts and a falling P/E—is how a stock like this delivers 30–40% drawdowns even if the long-term technology positioning remains intact. Any investor entering at current levels must accept that volatility as part of the trade.

M&A integration and execution – Celestial AI and XConn as high-stakes bets for NASDAQ:MRVL

Marvell is effectively re-engineering its portfolio through M&A, with Celestial AI at about $3.25B, XConn at $540M, and the earlier Tanzanite deal. Successfully integrating these assets is non-trivial. Celestial has to be tied into Marvell’s existing switch, optical and custom silicon franchises while preserving performance and cost advantages. XConn must be integrated into a coherent CXL solution stack and qualified across multiple hyperscaler deployments. Mis-steps in integration, delayed tape-outs, or slower-than-expected customer qualification cycles could push out revenue ramps by multiple quarters. That would directly challenge the consensus path that assumes data-center revenue grows >25% in FY2027 and about 40% in FY2028, and that Celestial’s revenue curve reaches $500M and $1B run-rates on schedule. Given the scale of these transactions, execution quality will be one of the main drivers of whether the stock tracks toward $118–$175 or moves sideways while the market waits for evidence.

Insider behavior and stock profile – monitoring conviction around NASDAQ:MRVL

Given the combination of aggressive growth assumptions and large acquisitions, insider behavior around NASDAQ:MRVL is an important signal. Persistent net insider buying on weakness would reinforce confidence that management believes in the Celestial, XConn and CXL roadmap. Heavy selling into rallies, particularly if it clusters around key execution windows, would raise questions about internal conviction.

Overall stance on NASDAQ:MRVL – Buy, with 40%+ medium-term upside and clear execution risk

On the numbers and the roadmap currently on the table, NASDAQ:MRVL justifies a Buy stance rather than a Hold or Sell. At roughly $80.23, the stock offers a credible path to around $115–$120 over 16–18 months if EPS revisions move up by about 10% for FY2027 and the market is willing to maintain a 30x earnings multiple. Under a stronger execution scenario in which Celestial AI reaches a $1B run-rate by Q4 FY2029, CXL becomes a standard memory pooling layer, custom silicon doubles again in FY2028, and data-center revenue growth tracks the >25% and ~40% trajectory, the longer-term FY28 target around $160–$175 remains in play. The risks are clear: heavy revenue exposure to China, high hyperscaler concentration, and substantial M&A integration risk in Celestial and XConn, all sitting on top of a premium revenue multiple. The stock is not a defensive compounder; it is a high-beta AI infrastructure bet. For investors comfortable with that risk profile, the upside embedded in the current data—both in terms of earnings growth and TAM expansion—supports treating NASDAQ:MRVL as a Buy rather than waiting for a cheaper entry that may only appear after a clear execution stumble.

That's TradingNEWS