Marvell Technology (MRVL) Q4 FY2026 Deep Dive: The AI Connectivity King's Defining Night

Price as of March 5, 2026: ~$78.09 | 52-Week Range: $47.09–$102.77 | Market Cap: ~$68B

The Setup

Marvell Technology reports Q4 FY2026 earnings tonight after the bell, and the timing couldn't be more loaded. This morning, the company unveiled the world's first 1.6T ZR/ZR+ pluggable optical module and the industry's first 2nm coherent DSP chips — a statement of technical intent arriving just hours before its biggest quarterly report of the year. Options traders are pricing in an 11–12% move in either direction. The stock sits at $78, roughly 24% below its 52-week high, in a zone where bulls and bears have been grinding for months.

The real question tonight isn't whether Marvell hits $2.2B in quarterly revenue. It almost certainly will. The question is whether management can answer the one thing the market actually fears: what happens after Amazon.

Company Overview: From Storage Controller to AI Infrastructure Backbone

Marvell was, not long ago, grinding through an identity crisis as a commodity storage controller supplier. CEO Matt Murphy changed everything with a series of high-stakes acquisitions — Cavium ($6B, 2018), Inphi ($10B, 2021), Innovium (2021) — that repositioned the company at the heart of cloud and AI infrastructure.

Today, its data center segment is 73% of total revenue, supplying custom AI chips (called XPUs) for hyperscalers, the optical DSPs that connect AI clusters, ethernet switching silicon, and increasingly, the photonic interconnects that will replace copper as data speeds outpace what metal wires can handle.

The financial transformation is just as striking. Trailing twelve months: $7.8B in revenue (+45% YoY), GAAP profitable with $2.47B in net income, and $2.71B in cash after collecting $2.5B from selling its automotive Ethernet division to Infineon last August. The balance sheet is in its best shape in years.

The Catalyst: What's Actually Happening Today

Q4 Expectations

Wall Street expects $2.20–2.21B in revenue (+21% YoY) and $0.79 in non-GAAP EPS, up from $0.60 a year ago. Those numbers are the floor, not the ceiling — Marvell hasn't missed its own quarterly guidance in recent memory. The more important number is data center revenue, expected around $1.63B, representing mid-to-high single-digit sequential growth from Q3's $1.52B.

The quarterly progression tells a consistent story: data center revenue went from $1.49B in Q2 to $1.52B in Q3, now projected at $1.63B in Q4. Every quarter up. And management already guided FY2027 data center growth of at least 25% on a path toward ~$10B in organic full-year revenue. If they deliver that, at any reasonable multiple, MRVL is significantly underpriced today.

The 2nm Optical Announcement That Changed the Morning

Just this morning, Marvell unveiled the COLORZ 1600 — the industry's first 1.6T ZR/ZR+ pluggable optical module — powered by "Electra," the world's first 2nm coherent DSP. They also launched "Libra," the first 2nm 800G coherent DSP. Both include built-in MACsec security, which matters as AI data centers become geopolitically sensitive infrastructure.

This is not a marketing stunt timed to earnings. It is a deliberate signal that Marvell is leading the transition to 1.6T optical speed, and doing it on the industry's most advanced node. As AI clusters scale toward millions of accelerators, the interconnect bottleneck becomes as critical as the compute bottleneck itself. Marvell is planting its flag at that exact intersection.

The $3.8 Billion Optical Bet

In December 2025, Marvell acquired Celestial AI for ~$3.25B — a startup with "Photonic Fabric" technology that puts optical interconnects directly alongside compute silicon, disaggregating memory at the system level. In January 2026, they added XConn Technologies for $540M to extend PCIe 8.0 and CXL switching capabilities. Both deals closed by early February 2026.

Total spend in two months: ~$3.8B, entirely on optical connectivity and next-generation switching. Management targets Celestial contributing $500M annualized run rate by Q4 FY2028 and $1B by Q4 FY2029. These are not near-term revenue drivers, but they are structural moat-builders at a time when Broadcom is competing for the same territory. Crucially, an expanded AWS warrant now includes Celestial AI's Photonic Fabric as a separate "swim lane" — meaning Amazon itself is co-developing the next generation of AI optical infrastructure with Marvell.

The Elephant in the Room: The Amazon Question

Here is the uncomfortable truth about Marvell's business: it is heavily reliant on Amazon Web Services. Amazon's Trainium chips — the custom AI accelerators AWS uses for machine learning — have been a cornerstone of Marvell's custom silicon revenue.

In December 2025, Benchmark analyst Cody Acree downgraded MRVL with what he called "high conviction" that Marvell had lost the Trainium 3 and Trainium 4 designs to Taiwan's Alchip. The stock dropped 6–7% that day. Amazon hasn't confirmed it. Marvell hasn't confirmed it. Counterpoint Research's supply chain work suggests Alchip is expanding as a premier AWS design partner for Trainium 3 on TSMC N3 process, ramping mid-2026.

My opinion: This risk is real but partially mis-framed by the market. Even if Alchip takes the lion's share of Trainium 3/4 silicon design work, Marvell still participates in XPU "attach" products — the ethernet switches, optical modules, and interconnect chips that wrap around those accelerators. Counterpoint still projects Marvell's shipments doubling between 2024 and 2027 despite the design share dip. And the expanded AWS Photonic Fabric warrant suggests Amazon is not walking away from Marvell — it's expanding the relationship into a new product line.

But the market will need to hear management's tone directly on this tonight. Any hedging, pushout language, or evasion about customer concentration could be the catalyst for the downside scenario.

Competitive Landscape

Broadcom is the obvious comparison, and the contrast in scale is stark. Broadcom commands a $1.8 trillion market cap versus Marvell's $68B. It has Google's TPU, Meta's MTIA engagements, and a massive VMware software business providing a high-margin floor that Marvell doesn't have. Since ChatGPT's launch in November 2022, Broadcom has returned +600%. Marvell +105%.

But this is not a winner-take-all market. Hyperscalers deliberately multi-vendor their silicon supply chains to control costs and avoid lock-in. Microsoft, Amazon, and Google all want credible alternatives to Broadcom. Marvell is the most capable #2 in the custom silicon space, and arguably the leader in optical networking silicon — which may matter more in the next phase of AI infrastructure buildout.

The more direct margin threat isn't Broadcom — it's Alchip and GUC, Taiwan-based design houses competing on price for XPU programs. This is a multi-year erosion risk that management needs to offset with wins elsewhere.

Financials and Valuation

At $78 per share, Marvell trades at ~24x forward FY2027 earnings and roughly 8.6x trailing revenue. For context, Broadcom trades at 41x forward earnings and is valued at ~22x revenue.

If Marvell achieves $10B in FY2027 revenue — management's stated organic target — at a 9x EV/Revenue multiple (conservative for an AI infrastructure name), you arrive at roughly $90B enterprise value, or approximately $100–103 per share. At a more generous 11x multiple (justified if Celestial AI optics ramp accelerates and margins expand), you reach roughly $110B EV, approaching $125 per share.

The 39-analyst consensus average price target is $112.89, with JPMorgan at $130, Needham and UBS around $118–120, Evercore ISI at $133, and Morgan Stanley the outlier at $95. The spread reflects genuine uncertainty about execution — but the weight of analyst opinion says the stock is 40%+ undervalued at current levels.

Key TTM financial metrics (data as of November 2025, ahead of Q4 tonight):
- Revenue: $7.8B (+45% YoY growth rate)
- Non-GAAP Gross Margin: ~59.7%
- Free Cash Flow: $1.58B
- Net Income (GAAP): $2.47B
- Forward P/E: ~24x

Technical Picture

MRVL has been in a consolidation pattern since its 52-week high of $102.77, settling near its 20/50/200-day moving average cluster around the $77–80 range. This convergence often precedes a directional resolution — and tonight's earnings are the trigger.

• Near-term support: $77.00–77.70 (well-tested pivot zone)
• Next support: $74.20 (recent swing close)
• Major support: $70–71 (significant technical level)
• Resistance: $88–90 (options-implied upside), then $95–100 (gap fill territory)

Options traders are pricing ±12.44% on the weekly expiry (into March 6) and ±15.64% on the monthly (into March 20). From $78, that maps to roughly $68–88 on the week and $66–90 on the month. The asymmetry is interesting: the stock is already 24% off its high, so the downside case is somewhat cushioned, while the upside case to $88–90 would represent a clean breakout.

What to Watch Tonight

The headline revenue beat will matter less than the following five signals:

1. Data center revenue vs. ~$1.63B estimate. Sequential growth from $1.52B confirms the AI infrastructure acceleration narrative.

2. FY2027 guidance specifics. Any step-up in the $10B revenue conviction, Q1 FY2027 guidance details, or data center growth above the 25% floor would be materially bullish.

3. Custom silicon ramp cadence. How confident is management about Trainium program continuity? Even indirect signals — language about "secured purchase orders," "lead hyperscaler engagement," or "multi-generational" programs — matter significantly.

4. Gross margin execution. The 58.5%–59.5% non-GAAP guided range must hold. Any compression raises concerns about competitive pricing pressure in XPU design.

5. Celestial AI / optical integration. Earlier-than-expected revenue contribution, hyperscaler pull-through on Photonic Fabric, or confirmation of the AWS Photonic Fabric warrant progressing would be a genuine positive surprise.

The Risk Picture

Four execution items need to go right simultaneously for the full bull case: AWS remains a meaningful long-term customer despite Alchip displacement; a second Tier-1 hyperscaler custom silicon win materializes by FY2028; Celestial AI optics hit their $500M/$1B annualized targets; and margins expand as revenue scales. That's a demanding set of conditions.

The Trainium loss uncertainty is the single biggest near-term overhang. If tonight's call reveals faster-than-expected content decline at Amazon without a credible offset, the stock could test $65–70.

Beyond company-specific risks, the macro backdrop is complicated. The Hormuz crisis is creating broad tech sector turbulence, and tomorrow's NFP print will generate additional market volatility regardless of what Marvell delivers tonight.

My Opinion

Marvell's real bull case is about light, not silicon.

Everyone focuses on the custom ASIC competition with Broadcom, the Trainium design wins, the XPU ramp. Those matter. But the structural opportunity most investors are underweighting is the photonic interconnect buildout — what is quietly becoming the defining bottleneck of second-wave AI scaling.

In the first wave of AI (2022–2024), the constraint was GPU availability. In the second wave (2025–2028), the constraint is bandwidth — moving data between accelerators fast enough to make clusters of a million chips function as a single machine. Copper has hard physical limits at this scale. The solution is light. Through Inphi (2021), Celestial AI (2026), and today's 2nm product launches, Marvell has been assembling the most complete optical networking portfolio in the industry, piece by piece.

The 1.6T announcement this morning is not incidental. It is Marvell saying: we've set the pace for coherent DSP technology for a decade, and we're not stopping at 2nm. When AI clusters deploy ZR/ZR+ optical interconnects at hyperscale — inside racks, between racks, between campuses — Marvell's DSPs and COLORZ pluggables will be the default infrastructure. That's a business that earns a premium multiple.

My recommendation: BUY below $80.

• Entry zone: $76–$80 (current range)
• 12-month price target: $100–$110 (base case: $10B FY2027 revenue, 9–10x EV/Revenue)
• Bull case target: $115–$125 (Celestial AI early revenue contribution + second hyperscaler win in FY2028 guidance visibility)
• Stop loss: $68 (options-implied weekly downside; major technical break level)
• Timeline: 12–18 months for the optical thesis to appear in reported numbers

Tonight is binary in the near term — the ±12% options pricing is honest about that. But for investors with a 12–18 month horizon, the combination of 40%+ revenue growth, GAAP profitability, $1.6B in free cash flow, industry-leading 2nm optical products, and a management team that has consistently been right about where AI infrastructure is headed makes MRVL one of the most compelling setups in semiconductors at current prices.

Size appropriately for the volatility. But don't let a noisy earnings night obscure what Marvell is building for 2028.

Data sources: Yahoo Finance, StockAnalysis, Futurum Research, Counterpoint Research, Marvell Investor Relations, Benzinga, MarketBeat, EBC Research, Moor Insights & Strategy, Trefis. Prices verified as of March 5, 2026 (intraday). This is opinion, not financial advice.