Artificial intelligence (AI) and the chips that power them are advancing at a rapid pace, but a major constraint is holding back their broader deployment – power. Better chips mean better models and greater compute capabilities, but a shortage in sustainable, accessible power supply is starting to emerge as a roadblock in AI advancement. According to a report from the International Energy Agency, global electricity consumption by data centres is projected to more than double by?2030, which is slightly more than Japan’s total electricity consumption today.
The organisations capable of scaling compute sustainably won’t just be those that can procure chips – they will be the ones that can deploy them where energy is most abundant, stable and clean. If that sounds familiar, it’s because the Bitcoin mining sector experienced this lesson a decade ago. Industrial miners mastered the art of pairing computation with low-cost renewable energy, deploying modular infrastructure at scale, and shifting hardware workloads based on energy unit economics.
Cango (NYSE: CANG) – one of the largest listed Bitcoin mining companies- is now applying those lessons to AI through what it describes as a “power-to-inference” transformation. Drawing on its global, energy-anchored footprint, the company is redeploying its infrastructure into a decentralised AI compute grid. The aim is to create a capillary network for the global AI economy—one that channels computing power along the same pathways that deliver electricity.
Constraints to cloud energy risk AI advancement
Evidence of power becoming a binding constraint is already visible across major data-centre markets. In Europe, the Middle East and Africa, a Savills report found that power supply shortages slowed new data-centre rollouts in 2025, with just 850 megawatts of capacity coming online—an 11 per cent decline from the previous year despite strong demand from cloud and AI workloads. In the United States, McKinsey has warned that grid bottlenecks in established hubs such as Northern Virginia, Santa Clara and Phoenix are lengthening connection lead times, forcing utilities to ration electricity in phases and delaying full deployments. Similar pressures are emerging across Europe’s largest data-centre clusters—Frankfurt, London, Amsterdam, Paris and Dublin, where constrained grids have prompted local authorities to limit new construction, underscoring how power availability, rather than demand, is increasingly dictating where AI infrastructure can scale.
Cango is approaching the problem from a different starting point. Rather than expanding capacity in a small number of congested data-centre hubs, the company is working on the assumption that compute should follow energy, not geography. Across parts of North America, the Middle East, South America and East Africa, significant volumes of low-cost renewable power remain underutilised. By locating infrastructure closer to these energy sources, Cango is seeking to align AI compute more directly with available, sustainable electricity, offering an alternative model to the increasingly power-constrained centres of traditional cloud development.
Modular GPU buildouts for an AI-driven era
Cango’s approach draws on lessons learned from a decade of Bitcoin mining, where speed, modularity, and standardisation were essential to staying competitive. While traditional hyperscale data centres can take years to build, Cango uses standardised, modular containers and uniform server configurations, allowing new deployments to go live in a fraction of the time. Each high-density unit hosts optimised GPU clusters capable of handling AI inference, rendering, HPC, and balanced mining workloads—mirroring the operational efficiencies miners developed to rapidly scale compute where power is abundant.
Scaling global energy capacity with strategic ownership
A major milestone in Cango’s shift toward owned-and-operated energy compute infrastructure came in August 2025, when the company acquired a fully operational 50 MW facility in Georgia. The project forms part of a global expansion aimed at jurisdictions combining renewable power, geopolitical stability, and regulatory clarity.
These sites provide Cango with a scalable foundation for long term growth, a risk-balanced portfolio of competitive power sources, and infrastructure ready for denser?GPU deployments. Together, they illustrate the company’s evolution into the “first right of access to power” network, delivering capacity to the market faster than traditional data centre players.
Why mining’s operational lessons matter now
The AI boom is uncovering challenges: power scarcity, delays, equipment lead times, and land constraints. Bitcoin miners solved many of these challenges years ago out of necessity—and their model translates exactly to today’s compute landscape:
Build where power is cheapest
Deploy modularly
Standardize configurations
Cango is now applying this framework on a global scale.
“As AI, HPC, and Bitcoin mining continue converging, the winners will be those with the most energy-efficient and deployment-ready infrastructure,” CEO Paul Yu says. “The future of compute will be modular, distributed, and energy-anchored, and Cango is building that foundation.”
Cango’s journey marks the convergence of two worlds — energy and intelligence. By empowering the global mining ecosystem into a synchronised, capillary?network?of?AI?compute, the company is establishing the bridge between the energy-rich world of mining and the compute-hungry world of artificial intelligence.
Original Article: https://ambcrypto.com/how-bitcoin-minings-hard-won-lessons-are-solving-ais-power-crisis/
