What to buy after NVIDIA? Jensen Huang has already given the answer.

Yesterday, I talked about NVIDIA's GTC keynote speech. Today, let's look at NVIDIA's ecosystem because recently I’ve noticed that companies invested in or partnered with NVIDIA have all seen great gains, such as CoreWeave, Nebius, Marvell, and so on. For NVIDIA to become a $5 trillion company, there must be a reason, so studying its ecosystem I think is very important—just like if you want to understand Ethereum, you need to study its ecosystem, that’s the same principle.

In fact, based on a series of investments and collaborations from 2025 to 2026, Jensen Huang has already been building a massive AI infrastructure ecosystem. NVIDIA itself has started frequently using a term: AI Factory. Its strategy has upgraded from selling chips to controlling the entire AI industry chain!

NVIDIA Ecosystem Architecture

The entire ecosystem can be divided into 7 layers: AI application layer, large model layer, AI cloud platform layer, AI data center layer, network interconnection layer, optical communication layer, and GPU/CPU underlying layer (arranged from top to bottom). NVIDIA is also deploying allies at each layer.

Layer 1: GPU Computing Layer

This is NVIDIA’s core layer, with key products including:

Blackwell

Rubin

Grace CPU

Vera CPU

NVLink

Spectrum-X

BlueField DPU

Layer 2: Chip Manufacturing Layer

Because NVIDIA designs chips but does not manufacture them itself, its foundry partner is TSMC.

Layer 3: Custom AI Chip Layer (ASIC)

Since many clients have customized needs that NVIDIA’s products sometimes cannot meet, here comes Marvell Technology, a new member. In 2026, NVIDIA invested $2 billion in Marvell.

Marvell is responsible for custom ASICs, helping clients with tailored requirements, rather than standard AI chips. Of course, it also provides networking, optical interconnects, and switching chips for AI data centers.

Layer 4: Network Interconnection

Responsible for communication between GPUs, such as a complete communication module, like optical switches.

Main partners at this layer include:

Marvell Technology: as mentioned above.

Broadcom: responsible for switches in AI data centers, such as Tomahawk switches, Jericho switches. They also design AI chips.

Layer 5: Optical Communication Layer

Future AI factories might be clusters of 1 million GPUs. The biggest challenge for AI factories is connectivity—how to link these 1 million GPUs. NVIDIA has previously stated that in the future, this will be done via fiber optics, not copper wires. So, a key partner here is Coherent, which mainly handles optical communication, optical networks, and lasers. NVIDIA invested $2 billion in Coherent in 2026.

Another partner is Lumentum Holdings, whose core technologies include high-speed optical transceivers, optical path switching, integrated optical modules, and OCS optical switches—claimed to be the fiber optic neural system of AI factories.

Layer 6: AI Cloud Platform

NVIDIA’s concept of an AI factory is cloud-based, so this will be the biggest cake layer in the future. Customers ultimately buy computing power, not just GPUs.

Currently, there are two major partners:

CoreWeave: the current AI cloud leasing platform, the largest GPU rental platform, akin to AWS in the AI era.

Nebius: the European version of CoreWeave, also responsible for AI data center construction.

Layer 7: Large Model Layer

This is where real compute power is consumed, and it’s also NVIDIA’s clients, such as OpenAI, Meta, Google, etc.

Layer 8: Robotics Ecosystem

Jensen Huang also emphasizes Physical AI, which should be a future focus, such as Tesla, Yutu Technology, and others.

NVIDIA is copying Microsoft’s strategy from the past. Looking at Jensen Huang, he has already controlled GPUs, networking, optical communication, AI cloud, and AI factories. His goal is no longer just to be a chip company but to be the “standard setter” for AI infrastructure in the era.

From an investment perspective, every important company in each of these layers is worth paying attention to and researching!