Marvell Technology (NASDAQ: MRVL) moved sharply higher on Monday, with the stock rising about 6% to roughly $146 after a report said Google is exploring a potential partnership with the chip company on new artificial intelligence hardware. For investors, the jump was not only about a possible deal headline. It was also about what the reported talks say about the next stage of the AI infrastructure market, where the biggest cloud companies are trying to build faster, cheaper, and more specialized systems for running advanced models at scale.
The reported discussions center on two possible products. One is described as a memory-focused processor that could work alongside Google’s existing tensor processing units, or TPUs. The other is said to be a new TPU designed for inference, the part of the AI process that happens after a model is trained and begins serving real-world requests. That matters because inference is increasingly becoming one of the most expensive and strategically important layers of the AI business. Training giant models still captures most of the attention, but once those models are deployed across search, assistants, enterprise tools, and cloud platforms, the cost of running them continuously can climb fast.
That is where a company like Marvell starts to stand out. Unlike Nvidia, which dominates the market for high-performance GPUs, Marvell has built a strong position in custom silicon, networking, and infrastructure technologies used inside modern data centers. The company’s appeal is not based on selling one standard chip to everyone. Its value lies in helping major customers build hardware tuned to their own workloads, power needs, and system architecture. In a market where hyperscalers want more control over performance and cost, that model is becoming more valuable.
Why the reported Google talks carry weight
Google has invested in custom AI chips for years through its TPU program. Those chips are already central to the company’s internal AI systems and also support external workloads through Google Cloud. But the reported Marvell discussions suggest Google could be widening the circle of partners involved in its next phase of AI hardware development. That would be notable because Google has historically worked closely with Broadcom on chip design. If it is now examining additional options, it may reflect how quickly demand is growing and how important supply flexibility has become.
The broader industry context supports that reading. Large technology companies are under pressure to reduce dependence on outside vendors while also keeping pace in the AI race. Nvidia’s chips remain the benchmark for many advanced AI workloads, but they are expensive, in high demand, and not always the best answer for every use case. That has pushed cloud players to think beyond one-size-fits-all hardware. Meta has already expanded its own custom chip relationship with Broadcom, while other large platforms continue to pour money into internal silicon efforts. The bigger message is clear: AI infrastructure is no longer just about buying more GPUs. It is about building purpose-specific systems across compute, memory, interconnects, and software.
Marvell fits neatly into that trend. The company and Broadcom have increasingly been seen as important behind-the-scenes enablers of the custom AI chip boom. Their role is especially relevant as companies look for specialized processors for data centers that can deliver better efficiency in targeted tasks. Marvell’s growing relevance also helps explain why its stock is often treated as an AI infrastructure name rather than only a traditional semiconductor stock.
There is another reason investors reacted quickly. The AI spending cycle remains enormous. Major technology companies are expected to spend at least $630 billion this year on AI-related infrastructure, including data centers, networking, and specialized processors. In that environment, even early reports of a new design relationship can move stock prices because the market knows that successful custom silicon programs can turn into multi-year revenue streams.
What this could mean for Marvell’s growth
At around $146, Marvell’s rally implied a substantial increase in market value. If those gains hold, the company could add more than $7 billion to a market capitalization that was roughly $122.15 billion before the move. Those are big numbers, but the valuation discussion does not stop there. Marvell has also been trading at a richer forward earnings multiple than Broadcom, around 33.35 times compared with 27.84 times. That premium suggests investors already believe Marvell can capture a meaningful share of long-term AI demand.
The company’s own targets add to that optimism. Marvell has said it expects revenue to approach $15 billion in fiscal 2028, a figure that reflects confidence in expanding AI, cloud, and data center demand. A deeper connection with Google would strengthen that narrative, especially if any collaboration moved from design work into broader deployment across cloud infrastructure.
The latest report also lands against another important backdrop: Nvidia’s own strategic interest in Marvell. Last month, Nvidia invested $2 billion in Marvell as part of an effort to make it easier for customers to combine Marvell’s custom AI chips with Nvidia’s networking products and central processors. That relationship underlines how the AI hardware market is evolving. Even companies that compete in one area may work together in another when customers are asking for more integrated solutions.
The software side matters too. AI lab Anthropic, for example, uses multiple kinds of chips, including Google-designed TPUs, to build and run its Claude models. That kind of mixed environment is becoming more common as AI companies try to balance cost, access, and performance. It also reinforces the importance of having a broader supplier ecosystem rather than relying too heavily on one chip architecture.
Investors should still keep one point in mind: the discussions were reported, not finalized, and neither Google nor Marvell immediately commented publicly. That means Monday’s move reflects expectations as much as confirmed business. Even so, the stock reaction shows how strongly the market views Marvell as a leverage play on the custom silicon shift.
For readers tracking the semiconductor sector, the bigger takeaway is that AI’s next chapter may be less about a single dominant chip and more about a network of specialized components working together. Inference, memory optimization, custom design, and cloud integration are becoming central themes. Marvell’s rise to about $146 on a 6% gain suggests investors believe it has a credible place in that future.
More on the company’s technology portfolio and data infrastructure business is available on Marvell’s official website, while broader market coverage of the reported talks was published by Reuters.
You may also like: US gold price today slips $48.20 as COMEX gold falls $68 in sharp sell-off
Make Swikblog your go-to source on Google for reliable updates, smart insights, and daily trends.
