Nvidia's Next $1 Trillion Opportunity May Have Nothing to Do With GPUs

Shortly after the launch of ChatGPT in late November 2022, big tech hyperscalers realized that chipsets known as graphics processing units (GPUs) could be used to develop next-generation applications in artificial intelligence (AI). At the time, Nvidia (NASDAQ: NVDA) had a first-mover advantage in the GPU landscape. As a result, the company's revenue skyrocketed to record levels seemingly overnight -- as did its valuation.

After rapid and sustained share price appreciation, Nvidia quickly entered the trillion-dollar club. With a market capitalization of $4.7 trillion, Nvidia now sits at the top of this exclusive roster.

While GPUs have ushered in Nvidia's status as the world's most valuable company, I think its next trillion-dollar opportunity lies elsewhere. Luckily, Nvidia CEO Jensen Huang has given us some clues. Let's explore where Nvidia has been investing lately, and assess what these moves could mean for the company's trajectory as the AI infrastructure era takes shape.

Image source: Nvidia.

Understanding the critical role of AI networking

AI networking refers to the specialized interconnects that link GPU clusters inside data centers. Without extremely high bandwidth, ultra-low latency, and lossless performance, communication among model training and inference deployments creates congestion that leaves GPUs underutilized.

This results in diminished returns on multi-billion-dollar hardware investments. In this sense, networking can be seen as being as important as raw compute. In other words, the fastest GPUs ultimately deliver little value if data cannot flow efficiently between clusters. As models grow larger and applications become more complex, the underlying network stitching AI development together becomes the limiting factor for overall system performance.

Does Nvidia even offer AI networking solutions?

While its GPU business takes the spotlight, Nvidia has quietly assembled deep expertise across multiple networking technologies tailored for AI.

The company's InfiniBand platform offers ultra-low latency, in-network computing, and high-bandwidth connectivity optimized for high-performance computing (HPC) and AI workloads. Nvidia complements this with its Spectrum family of Ethernet switches -- particularly the AI-optimized Spectrum-X platform, which features adaptive routing, congestion control, and predictable behavior within standard Ethernet environments.

Rounding out the offering are the BlueField data processing units (DPUs), which offload networking, storage, and security tasks from CPUs and GPUs. Taken together, Nvidia's networking suite forms a comprehensive, high-performance foundation capable of powering hyperscale AI deployments.

Nvidia is partnering with several AI networking leaders

To accelerate its position and secure additional supply in the networking domain, Nvidia has made several targeted strategic investments.

In March, Nvidia invested $2 billion each in Coherent and Lumentum. The rationale behind these partnerships is to advance Nvidia's position in optical interconnects and silicon photonics, two essential layers for transmitting data at high speeds over long distances with lower power consumption. Nvidia invested another $2 billion in Marvell Technology to deepen its reach in designing custom AI accelerators.

These moves are far from random. Rather, each investment quietly strengthens Nvidia's ability to build end-to-end optical and electrical networking components layered atop its GPU ecosystem.

By combining its GPUs' architectures with an expanding networking portfolio, Nvidia is positioning itself as the key provider of complete AI factories -- integrated systems that capture compute, high-speed interconnects, DPUs, software, and advanced photonics. The goal is to enable customers to deploy turnkey infrastructure capable of training and running the largest AI models at unprecedented scale and efficiency.

From a valuation standpoint, Nvidia currently trades at a forward price-to-earnings (P/E) ratio of roughly 22. As the chart illustrates, this multiple sits well below the elevated levels Nvidia reached during the peak enthusiasm of the AI revolution's initial GPU phase.

NVDA PE Ratio (Forward) data by YCharts.

I don't think the market has fully incorporated Nvidia's expansion into networking. Since all signs point to an acceleration in AI infrastructure spending over the next several years, Nvidia's true earnings power could prove considerably larger than current expectations. Given these dynamics, I see Nvidia as a no-brainer stock to buy and hold, as hyperscalers bolster their capex budgets, allocating more to networking gear to meet their capacity needs.

Adam Spatacco has positions in Nvidia. The Motley Fool has positions in and recommends Coherent, Lumentum, Marvell Technology, and Nvidia. The Motley Fool has a disclosure policy.