AI Demand Surges: Marvell Technology Forecasts Custom Chip Revenue to Top $10 Billion
Marvell Technology expects its custom chip business to exceed $10 billion in revenue by fiscal year 2029, driven by cloud companies expanding AI data centers and seeking alternatives to NVIDIA processors.
📋 Article Processing Timeline
- 📰 Published: May 28, 2026 at 08:49
- 🔍 Collected: May 31, 2026 at 23:48 (86h 59m after Published)
- 🤖 AI Analyzed: June 2, 2026 at 00:53 (25h 4m after Collected)
Marvell Technology predicted today that its custom chip business revenue will exceed $10 billion by fiscal year 2029, as cloud computing companies continue to expand AI data centers and invest in custom chip R&D to reduce reliance on NVIDIA processors. According to Reuters, the surge in AI applications is driving demand for special-purpose chips. These chips, combined with Marvell's interconnect technology, play a key role in advanced data centers, connecting thousands of processors used to train and run AI models. Marvell's stock price has more than doubled this year. Morningstar analyst William Kerwin stated that the custom chip revenue outlook implies that the business alone could generate $5 billion in revenue in fiscal 2028-2029, signaling strong growth in fiscal 2029. Marvell currently estimates that revenue for fiscal 2028 will increase to approximately $16.5 billion, up from its previous forecast of $15 billion. According to data compiled by the London Stock Exchange Group (LSEG), Marvell's second-quarter revenue is expected to reach $2.7 billion, plus or minus 5%, higher than the analyst average estimate of $2.6 billion. Adjusted earnings per share are estimated at 93 cents, plus or minus 5 cents, also better than the market estimate of 90 cents. Marvell, along with its larger competitor Broadcom, helps cloud computing companies design custom chips tailored to their specific data center needs, and this business has now become a massive source of revenue for both companies.
FAQ
Why are custom chips important?
They offer better efficiency and cost-performance for specific AI model training and execution compared to general-purpose chips.