Marvell Rallies 5% as Google Plans to Diversify for AI Chips
The AI chip boom is reaching an all time high with tech giants splurging billions in the space. Latest reports from The Information indicate Alphabet’s Google planning on diversifying beyond Broadcom for designing its custom chips.
The leading search engine company is now in talks with American chip designer Marvell Technology to develop new chips, aiding the running of its AI models efficiently.
This could be two distinct chips: a memory processing unit to complement Google’s Tensor Processing Unit (TPU) and a new TPU built for running AI models.
Following this news, the Marvell (MRVL.O) shares climbed over 5% on Monday 20 April, a peak not seen in the last 52 weeks. With the rising interest in AI, Marvell shares have gained about 64% so far this year, contrasting its 23% rough decline in 2025.
Both Broadcom and its primary challenger Marvell help clients design custom chips as growing adoption of AI tools boost demand for specialised processors used in advanced data centres.
While NVIDIA remains the market leader, selling the most popular general tools, the other two are slowly emerging as strategic alternatives, providing more options at lower costs.
Getting a custom-built chip
Leading tech companies are slowly reducing their dependence on external chip suppliers like NVIDIA by investing heavily on the development of custom AI chips.
This is to handle the huge daily costs of running AI as custom chips are often significantly cheaper and more energy-efficient, especially for day-to-day repetitive tasks.
Google deploys its AI accelerator TPUs for training AI models and responding to user queries, a process known as inferencing. It currently works with Broadcom, which leads the custom AI chip market, to design its chips.
The TPUs designed by Google are used by many organisations like Anthropic to develop and run its AI software and chatbot Claude. It is also emerging as a threat to NVIDIA’s dominance in the space.
As businesses increasingly seek alternatives to the expensive chips in the market, the reports of Google’s latest move suggests an attempt at diversification to gain leverage over its current suppliers.
Russ Mould, the Investment Director at AJ Bell, spoke to Reuters saying: “It should be no surprise that rivals [of NVIDIA] will want to grab a piece of the market and the apparent growth on offer by developing their own product. It also makes sense for customers to diversify their sources of supply, if they can, so they can spread technological and supply chain risk.”
Challenging the status quo
A new revenue stream beyond NVIDIA has opened up as tech companies move towards building custom AI chips. By multi-sourcing these partnerships, companies aim to secure their supply chains as the spend on AI continues to grow.
Google is leading this shift with its recent plans of including Marvell alongside its existing relationship with Broadcom.
In March 2026, NVIDIA invested US$2bn in Marvell to ensure custom chips designed by the firm work easily with its central processors and networking gear. The aim is to make the integration of bespoke silicon into existing data centre architectures more seamless.
Marvell also expects its revenue to approach US$15bn in 2028. If the positive investor sentiment following the reports of its Google partnership holds through, the company is set to add more than US$6bn to its market value of US$122.15bn.
However, to take on the Silicon Valley giant NVIDIA, it will need to look past the latest deal talks and prove its custom AI chips can outperform the industry’s most powerful processors.
