AI Data Center Leases: The New Foundation Of Compute Power
AI data center leases are multi-year agreements through which infrastructure providers reserve and operate dedicated GPU capacity for technology customers, turning short-term cloud rentals into predictable long-horizon commitments that shape how artificial intelligence workloads will scale, be financed, and be physically deployed over the next decade. The latest wave of AI infrastructure deals shows a clear pattern: GPU capacity that once floated in a spot-market frenzy is being pinned down in long contracts with specific megawatt allocations. Riot Platforms has signed leases covering 241MW of critical IT capacity with two major technology customers. Duos Technologies has executed five-year hosting service orders with Axe Compute for high-density AI facilities. Boost Run and HIVE are following similar paths. The story now is not about who can find GPUs, but who can guarantee them for years.
Riot And Duos: From Experiments To Long-Haul Commitments
The most striking sign that AI compute is entering a long-haul phase comes from Riot Platforms. The company has agreed to a 191MW build-to-suit lease at its Rockdale campus with a leading frontier AI lab, tied to an initial 20-year term that runs through June 2048. That is not a tactical experiment; it is a bet that AI workloads will demand industrial-scale GPU capacity for decades. Initial 96MW delivery is targeted for December 2027, with the full 191MW scheduled by June 2028, underscoring how far ahead these projects now plan. Duos Technologies is taking a similar stance. Its two project entities have executed five-year hosting agreements with Axe Compute covering 55MW of total AI facility capacity across multiple sites. Duos’ CEO openly frames these executed agreements as an advancement of an AI infrastructure strategy and proof that the company can turn development opportunities into long-term commercial relationships.
Boost Run And HIVE: Scaling GPU Contracts Into Recurring Revenue
While Riot and Duos focus on campus-scale facilities, Boost Run and HIVE show how mid-sized players are racing to secure GPU capacity contracts that convert volatile demand into recurring revenue. Boost Run has locked in a long-term lease with 10X Infrastructure Partners that provides immediate access to 20MW of GPU infrastructure capacity and plans to expand to 111MW across three sites through 2027. The initial 20MW is expected to be delivered in the fourth quarter of 2026 at an existing, fully energized site, letting the company deploy GPUs faster to meet demand for high-performance computing. HIVE, through its BUZZ HPC unit, has landed its largest AI deal yet: a five-year GPU cloud services agreement that will see 2,016 Nvidia Blackwell Ultra GPUs deployed at its Bell AI Fabric facility. The contract is expected to add about RM70 million in annual revenue to BUZZ HPC’s business, lifting its annualized revenue run rate to roughly RM180 million.
Why Infrastructure Deals Are Becoming Strategic Weapons
These AI infrastructure deals are more than routine hosting contracts; they are strategic weapons in the fight for GPU capacity. Axe Compute is described as a neocloud platform delivering dedicated enterprise GPU compute at global scale, and it is choosing partners that can meet the density, cooling, and reliability demands of next-generation GPU systems. Boost Run’s CEO highlights that its agreement reflects strong alignment between contracted demand and deployment capacity, underpinned by proven colocation capabilities. The common thread is control. Rather than scramble for short-term GPU rentals, hyperscale-minded customers want guaranteed megawatts, tailored facilities, and predictable economics. The result is a lock-in effect: once a lab or enterprise commits to a multi-year AI data center lease, switching providers is no longer trivial. That hardens market share and turns infrastructure into a moat.
The Next Phase: AI Compute As Long-Term Utility
The timelines attached to these contracts reveal the deeper shift. Duos and Axe Compute target initial project readiness beginning in late 2026 and continuing into early 2027, subject to construction, commissioning, performance testing, and formal acceptance. Boost Run expects its first 20MW at an energized site in the fourth quarter of 2026. Riot’s flagship AI lease does not reach full deployment until mid-2028. In other words, AI compute is being planned like a utility: slow to build, designed to last, and backed by long-term agreements. HIVE’s stock already shows how markets reward these moves, with shares rallying in pre-market trade and retail sentiment trending bullish as its contract pushes BUZZ HPC toward its annual recurring revenue goal. The conclusion is clear: AI is leaving the improvisational phase. GPU capacity contracts and AI data center leases are becoming the backbone of how serious players intend to scale.
