Marvell reported a record quarterly revenue of $1.817 billion for Q4 FY2025, up 27% year-over-year and 20% sequentially, led by the data center end market which contributed 75% of total revenue and was driven by AI-related demand and execution. GAAP profitability was achieved in the quarter and non-GAAP earnings per diluted share rose 40% sequentially to $0.60, underscoring the substantial operating leverage in Marvell’s model. Management reinforced that AI remains the primary growth engine, with several custom silicon programs in high-volume production and a multi-year ramp for next-generation XPUs supported by ongoing design wins and technology leadership (including 2nm silicon IP). The company also highlighted robust cash generation, deleveraging progress, and capital returns, while signaling continued AI-driven data center expansion into fiscal 2026 and beyond.
For the full-year FY2025, Marvell delivered $5.767 billion in revenue, with data center revenue growing 88% year-over-year, and a record $1.68 billion in operating cash flow. Non-GAAP gross margin expanded to about 60%, and non-GAAP operating margin reached 28.9%. The balance sheet showed improved liquidity and credit metrics, including a Fitch upgrade to investment grade and a net debt to EBITDA ratio of 1.58x. Guidance for Q1 FY2026 points to revenue around $1.875 billion (±5%), GAAP gross margin ~50.5% and non-GAAP gross margin ~60%, with non-GAAP EPS in the $0.56–$0.66 range and a long-term non-GAAP operating margin target of 38–40%.
Investment thesis: Marvell’s AI data center momentum, breadth of custom ASIC programs, and expanding interconnect strategy (notably Co-Packaged Optics and 3D/2D packaging, HBM-related IP, and CXL/DPDK-enabled NICs) position the company to benefit from an expanding AI compute market. However, execution risk remains with a competitive custom silicon landscape and potential mix-driven gross margin volatility. The next 12–24 months will be critical to witness continued ramp progression, supply-chain discipline, and sustained AI-driven design wins across hyperscalers.