Marvell Stock Price Forecast: MRVL $90 After Three Guidance Raises to $11B — Data Center at 74%, EPS Up 81%

Marvell Technology (NASDAQ: MRVL) Stock Analysis: $8.2 Billion FY2026, $15 Billion FY2028 Target, and Why $90 Is the Entry Point Into the AI Infrastructure Company Nobody Is Fully Pricing Correctly

MRVL at $90.51 — Down 1.14% Tuesday but 20% Above Where It Was Before the Q4 Print Changed Everything

Marvell Technology (MRVL) is trading at $90.51 on Tuesday, down 1.14% on the day within an intraday range of $90.35 to $93.44, carrying a market capitalization of $79.19 billion. The 52-week range of $47.09 to $102.77 tells the complete story of what fiscal year 2026 did to sentiment around this name — from a stock that was still being treated as a traditional semiconductor cyclical in the mid-$40s to a company that has now raised its FY2027 revenue guidance three separate times in six months, from $9.5 billion to $10 billion to the current $11 billion. The forward P/E of 29.50 against a trailing P/E that is compressing rapidly and a Wall Street consensus of Strong Buy at 4.53 out of 5 establishes the institutional view clearly. The Quant rating at Hold is the mechanical result of historical multiples screening expensive against sector peers — a framework that is simply the wrong tool for evaluating a company in the middle of a structural re-rating driven by an architectural shift in AI infrastructure.

The stock has gained approximately 20% since the Q4 FY2026 earnings print on March 5. It was down roughly 15% year-to-date entering that print, making the post-earnings surge a genuine sentiment reset rather than a momentum chase. The FY2027 guidance revision upward — from prior expectations of 25% growth to a new management target of 30% growth — was the catalyst. But the number that the market hasn't fully processed yet is FY2028 at $15 billion with EPS "well over $5" — a target that, if achieved, puts MRVL on a valuation trajectory that makes the current $90 look considerably more attractive than the absolute multiple suggests.

FY2026 by the Numbers — $8.2 Billion Revenue, 42% Growth, and the Operating Leverage That Separates MRVL From Every Other AI Beneficiary

Marvell Technology (MRVL) generated $8.2 billion in revenue for fiscal year 2026 — a 42% year-over-year increase that is the fastest top-line growth in the company's modern history. Data center revenue exceeded $6 billion and represented 74% of total revenue — a proportion that makes MRVL less a general semiconductor company and more a pure-play AI infrastructure proxy with some legacy carrier and enterprise business attached. The transformation from 2023, when data center was a minority of revenue, to 2026, when it represents three quarters of the entire business, is the structural story that the market is still calibrating to price.

The profitability metrics are where MRVL separates from the broader universe of AI-adjacent companies that are generating revenue excitement without converting it to earnings power. Non-GAAP EPS increased 81% to $2.84 in FY2026 — earnings growth nearly double the already-extraordinary revenue growth rate, driven by operating leverage that is compressing costs faster than revenue scales. Non-GAAP operating margin expanded 640 basis points to 35.3%. R&D spending increased just 6.4% — significantly slower than revenue even as absolute R&D spend reached $2.08 billion — demonstrating that the company is extracting more output per dollar of engineering investment as its product portfolio matures and scales.

Free cash flow reached $1.4 billion for the full year. Cash and equivalents finished at $2.64 billion. Net debt to EBITDA was just 0.57x at fiscal year-end — a balance sheet that is not stretched even after the Celestial AI and XConn acquisitions, which together will cost meaningful capital but are financed from a position of strength rather than desperation. The combination of 81% EPS growth, 640 basis point margin expansion, $1.4 billion in free cash flow, and a sub-1x leverage ratio in a single fiscal year is the operational profile that justifies re-rating conversations. This is not a company growing revenue while burning cash and hoping scale eventually produces profitability. It is already there, and it is accelerating.

The Q4 FY2026 Quarter That Triggered the Re-Rating — $2.22 Billion Revenue, $1.65 Billion Data Center, and the Guidance That Beat by $1 Billion

The Q4 FY2026 print delivered revenue of $2.22 billion — growth of 7% quarter-over-quarter versus estimates of $2.21 billion. The beat itself was marginal. The transformation came from the forward guidance. Management guided Q1 FY2027 revenue to $2.40 billion — 8% sequential growth and $120 million above the $2.28 billion consensus. Data center revenue within that Q1 guidance is expected to reach $1.82 billion — growing 10% sequentially from Q4's $1.65 billion.

Data center revenue in Q4 grew 9% quarter-over-quarter to $1.65 billion, better than expected. Non-AI-related revenue grew 2% sequentially to $567 million — the legacy carrier and enterprise business providing ballast without being the growth driver. Non-GAAP gross margins contracted 70 basis points quarter-over-quarter to 59%, weighed down by the lower-margin custom ASIC business that is scaling faster than the rest of the portfolio. Management guided gross margins to 58.25%–59.25% for Q1 FY2027 — another quarter of compression from the ASIC mix shift. This margin pressure is the one legitimate near-term headwind, but the market is correctly prioritizing the $2.40 billion Q1 revenue guide and the $11 billion FY2027 target over the 70 basis point gross margin headwind.

The guidance trajectory — FY2027 raised from $9.5 billion to $10 billion to $11 billion in six months — is the single most important data point in the MRVL thesis. Companies that raise guidance three times in six months are not managing to a plan. They are being surprised to the upside by demand that is exceeding their own models. CEO Matt Murphy's language on the call confirmed this: "I mean by design because the top four U.S. hyperscalers is spending the bulk of the CapEx, that's where the dollars are going to go." Murphy added that interconnect is expected to grow more than 50% year-over-year in FY2027 — "well above our prior expectation of 30% growth." When the CEO is beating his own prior guidance on the interconnect business by 20 percentage points, the market's next question is whether $11 billion for FY2027 has the same upside potential. The answer, based on three consecutive guidance raises, is yes.

Custom Silicon — From Zero to $1.5 Billion in FY2026, Doubling in FY2028, and Why MRVL Is Targeting 20% of a Market Broadcom Owns

Marvell's custom silicon business — the XPU and custom ASIC segment that designs application-specific chips for hyperscaler customers — scaled from effectively zero to $1.5 billion in FY2026. Management has guided that custom silicon will grow again in FY2027 before at least doubling in FY2028, with the addition of a new Tier 1 XPU customer alongside growing NIC and CXL attach revenue. A doubling of $1.5 billion is $3 billion of custom silicon revenue alone in FY2028 — nearly 20% of the $15 billion total revenue target coming from a business that didn't exist three years ago.

The competitive context matters here. Broadcom (AVGO) holds an estimated 70%+ share of the custom ASIC market. MRVL is targeting 20% market share by 2028. The total custom ASIC market is expected to grow at a 27% compound annual growth rate through 2033. In an expanding market growing at 27% CAGR, capturing 20% share doesn't require taking business from Broadcom — it requires being the alternative that hyperscalers deploy alongside Broadcom as their AI infrastructure diversification strategy. Amazon (AMZN), Google, and Microsoft are all actively developing proprietary custom silicon, and none of them want to depend entirely on a single external ASIC designer. MRVL is the structural beneficiary of that diversification imperative.

The December 2025 concern that MRVL lost the Amazon Trainium 3 and 4 contract — which triggered Benchmark's downgrade to Hold and a sharp pullback — appears to have been at minimum overstated and at maximum incorrect. The Q4 earnings call guidance confirming $11 billion in FY2027 with everyone "on board" and the ASIC opportunity "only expanding" is the company's answer to those concerns. Whether Amazon's exact program commitment was modified or not, the broader custom silicon pipeline is clearly generating enough aggregate demand to drive the guidance upward, not downward.

The Interconnect Business — 50% Growth in FY2027, Switches Exceeding $600 Million, AEC/Retimers Doubling

The part of Marvell's business that the market consistently underweights relative to the custom silicon headline is the interconnect portfolio — and it is precisely this segment that makes the long-term thesis more compelling than a pure ASIC play. Interconnect revenue is projected to grow over 50% in FY2027. Switch revenue is expected to exceed $600 million. AEC (Active Electrical Cable) and retimer revenue is projected to more than double.

Understanding why this matters requires understanding the AI data center architecture shift that is driving the numbers. As AI infrastructure transitions from 400G to 800G to 1.6T bandwidth speeds — a generational transition that is happening in the current build cycle — every network element inside a hyperscaler data center needs to be upgraded simultaneously. Port count increases drive unit volume. The shift from lower to higher bandwidth speeds drives ASP (average selling price) increases per unit. MRVL benefits from both dimensions simultaneously — more ports needed and higher price per port as bandwidth requirements scale.

CEO Murphy was specific on the call about this dynamic: "Even if port count doesn't increase, the industry itself is moving, and with that, higher ASP." This statement captures why interconnect is a structurally growing business independent of the macro AI capex cycle. The technology upgrade cycle — not just the AI infrastructure buildout — is generating demand. When the industry moves from 400G to 800G to 1.6T, there is no option to defer the transition. Every new switch rack, every new server deployment, every new optical module requires the higher-bandwidth components that MRVL designs. The 50% growth in FY2027 interconnect revenue is therefore not purely dependent on whether AI capex increases — it is partially locked in by the technology migration itself.

Celestial AI and XConn Acquisitions — $500 Million Annualized Run Rate Target by End of FY2028 and Why These Are Medium-Term Not Immediate Catalysts

Marvell completed acquisitions of Celestial AI — a photonic scale-up connectivity company — and XConn, which adds PCIe and CXL switching technologies to the portfolio. The combined revenue from these two acquisitions is expected to reach approximately $250 million within FY2028. Celestial AI alone is expected to reach an annualized revenue run rate of $500 million by the end of fiscal year 2028, with annual non-GAAP operating expense impact of approximately $50 million and 27 million diluted shares.

The strategic logic of both acquisitions is sound. Celestial AI extends MRVL's footprint into photonic scale-up connectivity — the optical layer that connects AI accelerators within and across racks in next-generation data centers. As AI models grow larger and distributed training requires faster interconnects between chips, photonic connectivity becomes the performance bottleneck that determines how fast clusters can train. MRVL positioning itself in the photonic layer — not just the electrical interconnect — is the architectural completeness play that makes the company harder for a hyperscaler to replace at any single layer.

XConn's PCIe and CXL switching technologies address the memory disaggregation challenge that is emerging as AI model sizes scale beyond what can fit in a single server's memory. CXL (Compute Express Link) allows multiple processors to share a common pool of memory across a high-speed fabric — a capability that becomes essential as AI inference workloads require access to terabytes of model parameters that no single chip can address directly. MRVL owning the CXL switching layer puts it inside the fundamental memory architecture of next-generation AI servers.

The honest assessment is that neither acquisition is a FY2027 revenue driver. The combined $250 million in FY2028 revenue contribution represents approximately 1.7% of the $15 billion target. Current valuation should not be based on these acquisitions — and it isn't. The $11 billion FY2027 and $15 billion FY2028 guidance is predominantly organic data-center and interconnect growth. The acquisitions are medium-term optionality, not near-term revenue justification.

The Lumentum Partnership — Optical Circuit Switching and Why MRVL Is Penetrating the AI Network Fabric Itself

Marvell's partnership with Lumentum (LITE) on optical circuit switching is one of the most strategically significant developments in the company's recent history that is receiving the least analytical attention. Optical circuit switching at the AI data center layer — where MRVL's silicon and Lumentum's photonic components combine — promises materially lower latency and more power-efficient AI cluster architecture than the current generation of electrical interconnects.

The significance is not the revenue from this partnership in the near term. The significance is that MRVL is positioning itself inside the AI network fabric at the optical layer — the deepest and most architecturally foundational layer of the infrastructure stack. Hyperscalers will eventually need optical scale-up networking at exabyte-scale AI cluster sizes. The company that owns the silicon layer of that optical fabric — not just the custom XPU but the connectivity fabric that binds the XPUs together — has a moat that is qualitatively different from a company that only designs chips. MRVL and Lumentum together at OFC (Optical Fiber Communication Conference) highlighting new data center solutions is the early-stage signal of a collaboration that could produce meaningful revenue contribution in FY2029 and beyond.

FY2027 Revenue at $11 Billion, FY2028 at $15 Billion — The Guidance Math That Makes the Valuation Defensible

Marvell (MRVL) at $90.51 carries an EV/Sales multiple of approximately 9.6 times — 200% higher than the sector median. The EV/EBITDA multiple of 30 times is also well above industry averages. The trailing P/E at 29.50 and forward P/E at approximately 23–24 times look stretched against a semiconductor sector that averages 21.46 times forward earnings. These are the numbers that generate the Quant Hold rating and the thesis from skeptics that the stock is "priced for perfection."

The bull case requires rejecting the premise that traditional multiples are the right framework. A company guiding to $11 billion in FY2027 — revenue growth of approximately 34% from FY2026's $8.2 billion — and $15 billion in FY2028 — approximately 36% growth — with EPS "well over $5" by FY2028 is not a company where static multiple comparisons against the semiconductor sector median are informative. The forward PEG ratio at 0.56 is the more relevant metric: it measures the price-to-earnings multiple relative to the earnings growth rate. A PEG below 1.0 indicates undervaluation relative to growth. At 0.56, MRVL is priced below fair value on a growth-adjusted basis.

The 12-month price target math: FY2027 consensus EPS at approximately $5 (management guided "well over $5" for FY2028, implying $4+ for FY2027) at a 23–25 times forward P/E produces a January 2027 target price of approximately $115–$125. The current price of $90.51 implies 27–38% upside over a 12-month horizon. The bull case with a 25 times forward multiple on $5 EPS is $125. The more conservative case with 23 times on $4.50 EPS is approximately $103. Neither requires multiple expansion — they simply require delivery of the guidance that management has now raised three times.

Customer Concentration Risk, Export Policy, and the 82% Revenue Concentration at the Top 10

The risk framework for MRVL is not abstract. One distributor alone accounted for 37–38% of revenue in FY2026. The top 10 customers represented 82% of total revenue. Customers with operations in Asia accounted for 77% of total revenue. The company carries $11.1 billion in goodwill against $22.3 billion in total assets — nearly half the balance sheet is intangible acquisition history. These concentration metrics are real, they are not immaterial, and they deserve explicit acknowledgment in any honest assessment of the stock.

The China export restriction evolution is the most acute concentration risk. As advanced computing restrictions on China tighten — and the policy direction has been consistently toward tighter controls, not looser ones — MRVL's 77% Asia revenue exposure creates a scenario where a single regulatory action could materially disrupt revenue for one or more quarters. This is not a theoretical risk. It is an ongoing regulatory reality that management must navigate with every new product generation.

The customer diversification within the concentration is the partial offset. CEO Murphy was specific: "within our portfolio…we're highly diversified within each of these customers." The same hyperscaler customer — Amazon, Google, Microsoft — may be buying custom XPUs, interconnect switches, optical components, and CXL solutions simultaneously. The revenue concentration is in a small number of customers but spread across an expanding product category footprint within each customer. That reduces the binary risk of losing a single program — losing one program from a customer that buys seven different product lines is dramatically less damaging than losing one program from a customer where that program is the entire relationship.

For insight into insider transaction activity at MRVL, the insider transactions page and full stock profile provide the complete ownership and management activity picture — worth reviewing in context of the guidance raises and the Celestial AI/XConn integration timeline.

MRVL Is a Buy at $90 — The Full Trade Decision With Price Targets

Marvell Technology (MRVL) is a buy at $90.51 with a 12-month price target of $115–$125. The investment case rests on five compounding factors that the static multiple skeptics are misweighting: 42% revenue growth that will sustain at 30%+ in FY2027 on management's own guidance; 81% EPS growth that is accelerating the P/E compression toward fair value; interconnect growing 50%+ in FY2027 from an undervalued and underanalyzed segment; custom silicon doubling in FY2028 with new Tier 1 customer additions; and the Celestial AI/XConn/Lumentum photonic architecture positioning that creates medium-term optionality not priced into any current model.

The RSI approaching overbought on the daily chart suggests the near-term entry is best managed by waiting for any pullback toward $87–$89 rather than chasing at Tuesday's $90.51. A 3–5% consolidation from the post-earnings surge is the technically expected behavior before the next leg higher. Any dip to the $87–$89 zone provides the entry with the 12-month target intact and a defined stop below $84 — the pre-earnings support level.

The bear case requires either a hyperscaler CapEx pause that is not currently signaled by any of the top four customers' guidance, or an export restriction action that materially disrupts the Asia revenue base. Neither is the base case. Both are worth monitoring and would warrant sizing reduction if evidence emerged. Absent those specific catalysts, MRVL at $90 with $15 billion in FY2028 revenue guidance and $5+ EPS is one of the clearest risk-reward setups in the AI infrastructure semiconductor space — more attractively priced than Broadcom (AVGO) on a growth-adjusted basis and with more product diversification within its AI architecture footprint than any direct comparison suggests.

That's TradingNEWS