The AI data center sector is currently navigating a complex supply-and-demand dynamic. Global demand for intelligent compute capacity is surging; IDC projects China's intelligent computing power alone will grow 87.9% year-over-year in 2026 to 2,576.5 EFLOPS. At the same time, the physical constraints of power and cooling are tightening. Average rack densities have climbed from 16 kW in 2025 to 27 kW in 2026, while AI-specific racks now hit 246 kW, pushing liquid cooling from an option to a structural necessity. Data center electricity consumption reached roughly 500 TWh in 2025 and could approach 950 TWh by 2030, straining grid interconnection queues that already span three to seven years.
Against this backdrop, the market's skepticism toward CoreWeave's debt load is understandable. The neocloud model relies on massive upfront capital expenditure financed by debt, secured against long-term GPU rental contracts. Any disruption to utilization rates or financing costs can amplify balance-sheet risk. Galaxy Digital's Helios business, by contrast, sits inside a diversified financial services firm with multiple revenue streams and a different capital structure.
Whether the market closes the valuation gap depends on two variables. First, CoreWeave's debt trajectory must stabilize enough to stop the sector-wide overhang. Second, Galaxy must demonstrate that Helios can convert its infrastructure into contracted, recurring revenue at margins that justify the build-out. The next few quarters of utilization data and lease-up velocity will be the proof points.