What They Do
Marvell Technology designs semiconductor solutions that power the infrastructure of the internet, focusing heavily on data center connectivity and networking chips. Their primary customers are cloud giants like Amazon, Microsoft, and Google who need specialized chips to connect servers, storage, and networking equipment in massive data centers. The company generates revenue by selling both standard semiconductor products and custom silicon solutions designed specifically for individual hyperscale customers.
Investment Thesis
**AI Infrastructure Goldmine** — Data center revenue hit $6+ billion in fiscal 2026 (46% growth) and is expected to drive total company revenue from $8.2B to nearly $15B by fiscal 2028, as AI workloads demand exponentially more connectivity**Custom Silicon Moat** — Their custom business doubled to $1.5B in 2026 and is expected to at least double again in 2028, creating sticky, high-margin relationships with hyperscale customers who can't easily switch vendors**Margin Expansion Machine** — Operating margins expanded 640 basis points to 35.3% in fiscal 2026 with EPS growing 81% to $2.84, showing tremendous operating leverage as revenue scalesKey Risks
**Customer Concentration** — Heavy dependence on a few hyperscale cloud customers means any spending pullback or design losses could significantly impact revenue**Cyclical Semiconductor Market** — Data center CapEx is notoriously volatile, and any broader economic slowdown could quickly reverse the current AI-driven spending boomGrowth Story
Marvell is riding the AI infrastructure wave with revenue expected to grow from $8.2B in fiscal 2026 to approaching $11B in 2027 (30%+ growth) and $15B in 2028 (close to 40% growth). The growth is driven by insatiable demand for data center connectivity as AI workloads require exponentially more bandwidth between processors, memory, and storage. Management expects to exit fiscal 2027 with quarterly revenue exceeding $3B, demonstrating the accelerating momentum.
Profitability Profile
Marvell has achieved impressive profitability with 59.5% gross margins and 35.3% operating margins in fiscal 2026, representing 640 basis points of operating margin expansion year-over-year. The company generated strong cash flows and returned $2.2B to shareholders through buybacks and dividends. Management expects non-GAAP EPS to grow from $2.84 in 2026 to "well over $5" in 2028, showcasing tremendous operating leverage as their high-margin custom silicon business scales.
Competitive Position
Marvell holds a strong position in data center interconnect solutions, competing with companies like Broadcom and Intel but differentiating through custom silicon capabilities and deep customer relationships. Recent acquisitions of Celestial AI and XConn strengthen their AI networking portfolio, particularly in emerging technologies like co-packaged optics. Their custom business creates switching costs for hyperscale customers, as these chips are designed specifically for each customer's unique architecture and requirements.
Recent Momentum
Q4 2026 delivered record revenue of $2.2B with strong beats on both revenue (+0.5%) and EPS (+$0.01), driven by robust data center demand representing 74% of total revenue. Management significantly raised fiscal 2027 guidance from $10B to approaching $11B and provided aggressive 2028 outlook of $15B revenue. The company has consistently beaten earnings expectations over the past four quarters, demonstrating strong execution in a demanding market.
What Wall Street Thinks
Analysts are increasingly bullish on Marvell's AI infrastructure exposure, with the company consistently raising guidance and beating expectations. The recent earnings call showed management's confidence in sustaining 30-40% annual growth rates through 2028, which has likely prompted upward estimate revisions. The key debate centers on sustainability of this growth trajectory and whether current valuations adequately reflect the cyclical nature of semiconductor demand.
The Bottom Line
Marvell is uniquely positioned to capitalize on the AI infrastructure buildout with revenue expected to nearly double from $8B to $15B over two years, driven by custom silicon solutions that create deep moats with hyperscale customers.
