Investing in the AI Supply Chain Beyond the Chip

The gold rush analogy is overused in technology investing, but it captures something important about the current AI moment. In the California gold rush of 1849, the most reliable fortunes were made not by the miners themselves but by the suppliers of picks, shovels, and denim trousers. The AI equivalent of this insight — that the infrastructure enabling AI may be a more durable investment than the AI applications themselves — has been widely adopted. What is less widely appreciated is how broad and deep the AI supply chain actually is.

Nvidia's dominance in AI training chips is well-documented, and its stock performance has reflected it. But the GPU is only one component of an AI compute cluster. The networking fabric that connects thousands of GPUs — InfiniBand switches, high-speed interconnects, custom silicon for network processing — is a multi-billion dollar market dominated by a small number of specialist companies. The power infrastructure that feeds a hyperscale AI campus — transformers, switchgear, uninterruptible power supplies, cooling systems — represents a capital expenditure that often exceeds the cost of the compute hardware itself. And the software layer that orchestrates AI workloads across distributed infrastructure — orchestration platforms, MLOps tools, inference optimisation software — is a rapidly growing market with high switching costs and attractive unit economics.

For private investors, the most interesting parts of the AI supply chain are those that are not yet publicly traded or are trading at valuations that do not fully reflect their strategic importance. Specialist cooling technology companies, power electronics manufacturers, and networking silicon designers are all areas where private capital can access the AI infrastructure theme at earlier stages and more attractive entry points than the public markets currently offer.

At Lensbridge, our investment in PacketFabric — a leading Network-as-a-Service operator providing the programmable networking infrastructure that AI workloads require — is an expression of this supply chain thesis. We look for assets that are essential to the AI infrastructure stack, that have durable competitive positions, and that are not yet priced to reflect their strategic value. The supply chain beyond the chip is full of such opportunities.