NVIDIA is moving ahead and beyond AI chips towards networking, photonics, and custom silicon. It seems that challenges have now shifted from computing power to connectivity. The Chip maker is partnering with some key suppliers. This includes Marvell Technology, Lumentum, and Coherent Corp with an investment of around $2 billion.Jensen Huang, CEO of Nvidia, called Marvell “The Next Trillion-Dollar Company” at Computex in Taiwan. He tried to renew investor interest in the AI networking and custom-chip maker. Marvell shares rose by more than 12% following the remarks. It extended a rally that has seen the stock more than double in 2026 as investors bet on growing demand for AI infrastructure.Data shows that Marvell shares are up by almost 24% in the pre-market trading session.Marvell sees AI demand accelerateThe partnership between Nvidia and Marvell Technology centers on advanced networking, custom silicon, and silicon photonics technologies. Marvell is also participating in Nvidia’s NVLink Fusion ecosystem, which enables cloud providers and other partners to connect custom accelerators and processors to Nvidia’s AI infrastructure. As AI clusters become larger and more complex, networking and interconnect technologies are becoming increasingly critical alongside computing power.Marvell’s recent financial results suggest why investors are bullish on it. Data shows that the company’s forecast for second-quarter revenue is around $2.7 billion. It is above all the Wall Street estimates.A report suggests that its custom-chip business could generate more than $10 billion in annual revenue by fiscal 2029. Marvell also raised its fiscal 2028 revenue outlook to about $16.5 billion and expects its data-center business to grow roughly 50% this year, driven by AI-related demand from hyperscale cloud providers.Despite the bullish outlook, Marvell remains well below the trillion-dollar threshold. The company currently has a market capitalization of roughly $190 billion, meaning it would need to increase its value by more than five times to reach $1 trillion. Nevertheless, investors have rewarded the company’s expanding role in AI infrastructure, with shares gaining more than 100% year to date as demand for custom AI chips, optical interconnects, and data-center networking solutions continues to accelerate.Marvell CEO Matt Murphy noted that AI infrastructure is entering a new phase of limitations. He said that computing was the initial limitation, followed by memory. Now, connectivity is the main limiting factor. This change is due to the large-scale use of AI cluster solutions, which require high-speed connections between different systems.Marvell’s AI infrastructure bet pays offMarvell has shifted its focus to areas experiencing rising demand. Data centers used to account for less than 10% of the company’s revenues. At present, their share has reached around 75% based on Marvell’s last financial report. Demand from top cloud companies has been increasing as they build new AI infrastructure.For the AI networking market, companies are defining their strategies based on how they address the data movement aspect of AI infrastructure. Each company is targeting different areas within the stack. For example, Marvell Technology targets custom silicon, optical interconnects, and networking semiconductors for hyperscalers. The company sees itself as a rapidly growing entrant in the market for next-gen AI connectivity with growing interest in silicon photonics and rack-scale interconnects.Broadcom is currently the dominant player in the data center networking market. The company has established itself as an absolute leader in the market for Ethernet switches and provides embedded infrastructure solutions to major cloud players. This leadership position has been achieved due to its scale and presence in most global data centers.On the other hand, AMD is not a major networking company but one that provides compute power. In the case of AI infrastructure, AMD can be considered an indirect participant in networking, which relies on the power of its CPUs and GPUs.If you're reading this, you’re already ahead. Stay there with our newsletter.