Enterprise AI Shift Drives Demand for Data Center 'Plumbing' Stocks
As firms move AI workloads from the cloud to in-house clusters, infrastructure vendors are seeing a surge in demand for connectivity and cooling.
The artificial intelligence gold rush is expanding beyond hyperscale cloud providers into the broader enterprise sector. Companies, including law firms, are increasingly shifting from cloud-only experiments to building in-house data centers to maintain greater control over proprietary data and AI workloads.
This transition is fueling a massive upgrade in data center physical infrastructure. To prevent bottlenecks between GPUs and memory, firms are investing in high-speed connectivity and specialized server hardware. This shift benefits a specific group of "pick and shovel" vendors—most notably Astera Labs, Delta Electronics, and Quanta Computer—who provide the essential power, cooling, and connectivity required for dense GPU racks.
The Connectivity Bottleneck
Astera Labs (ALAB) has emerged as a primary beneficiary of this infrastructure buildout. The company reported explosive financial growth, with revenue climbing from $396.3 million in 2024 to $852.5 million in 2025. This growth pushed the company into profitability, with a reported 2025 net income of $219.1 million.
A key driver is the Scorpio P-Series smart fabric switches, which contributed more than 10% of total revenue in 2025. According to CEO Jitendra Mohan, the combined opportunity for PCI Express and UALink—specifically regarding the Scorpio X—represents a $10 billion total addressable market. Other industry reports suggest the broader merchant scale-up switching market could reach $20 billion by 2030.
Power and Thermal Management
As GPU density increases, the heat generated by these systems requires advanced thermal management. Delta Electronics is addressing this need by providing chip-level cooling solutions. Specifically, the company provides a micro-channel cold plate designed for NVIDIA Vera Rubin NVL72 systems, ensuring that high-performance AI clusters can operate without overheating.
Complementing the connectivity and cooling is the hardware assembly. Quanta Computer is currently producing NVIDIA GB300 AI servers. The company has reported an "unimaginable" number of orders for these platforms, signaling a massive appetite for ready-to-deploy AI server racks among enterprise clients.
Market Implications
This shift is significant because it diversifies the AI hardware market. Previously, growth was concentrated among a few giant cloud providers like Microsoft and Google. Now, the market is expanding to thousands of individual enterprises. This creates a more resilient growth path for infrastructure vendors, as their success is tied to the physical requirements of AI—power and connectivity—rather than the success of any single AI model or software application.
What to Watch
Investors and industry analysts are now monitoring the pace of these private cluster deployments. While the order volume for GB300 servers is confirmed to be massive, the exact timing of shipment acceleration remains a key metric for the coming year. Additionally, the ability of Astera Labs to scale its Scorpio line will determine if it can capture the projected multi-billion dollar switching market as enterprises move toward more complex, multi-GPU architectures.