Broadcom and Marvell are leading in custom AI chip development for 2025
From Nasdaq: 2025-01-12 06:50:00
In the early days of artificial intelligence (AI) infrastructure, graphic processing units (GPUs) have been crucial for training AI models. However, large tech companies are now turning to custom AI chips, or ASICs, for better performance and efficiency. These chips are designed for specific tasks but lack the flexibility of GPUs.
Broadcom, a leader in custom AI chips, has worked with Alphabet to design tensor-processing units (TPUs) for AI training. These custom chips offer features like matrix multiply units and SparseCores, making them superior to GPUs for AI tasks. Alphabet credits the use of TPUs for cost savings and reduced inference time in its cloud computing segment.
Broadcom’s AI revenue exceeded expectations, reaching $12 billion, with potential for significant growth. Marvell Technology is also making strides in custom AI chips, collaborating with Amazon on the Trainium chip for large language models. Marvell sees a $40 billion market opportunity in custom AI chips and aims to capture a 20% market share.
Investors have a second chance at a lucrative opportunity with “Double Down” stock recommendations for companies poised for success. Past recommendations like Nvidia, Apple, and Netflix have yielded impressive returns. The current “Double Down” alerts highlight three promising companies for potential growth.
Read more at Nasdaq: 2 Custom AI Chipmaker Stocks for 2025 (Hint: Not Nvidia or AMD)
