AI Infrastructure Bottleneck: An Opportunity for Bitcoin Miners, According to Grayscale

AI Facing the Bottleneck of Its Energy Infrastructures

The development of AI is currently facing a sort of existential crisis regarding the risks associated with its potential loss of control. However, another more down-to-earth reality could also quickly become problematic: << the physical, and potentially social and political, limits to the rapid construction of data centers >>.

This observation has been made by analysts at Grayscale in a lengthy document titled: << Investing in the Face of the Computing Power Bottleneck >>. Indeed, the constraints on the infrastructures necessary for AI computation should soon lead to << a persistent imbalance between the supply and demand for computing power, which will primarily benefit holders of computing capacities with effective access to energy >>.

The results produced by AI are digital, but the infrastructure that makes them possible is fundamentally physical. Although this infrastructure is developing rapidly, it still struggles to keep pace with demand. Bottlenecks exist at every stage of the energy-fed computing capacity chain, with perhaps the most difficult to resolve being obtaining the necessary permits to access the electrical grid. Grayscale

As a result, despite a rapid increase in capacities provided by American data centers, the resources still available continue to fall to historically low levels due to a demand that immediately absorbs every new capacity brought to market, to the point of making computing power a true << tradable commodity >>.

Available capacities of data centers decline despite the increase in supply

A Situation Potentially Very Favorable to Bitcoin Miners

According to Grayscale analysts, this bottleneck could sustainably keep computing power prices at high levels. This reality is already being taken into account by market players such as CME Group and Silicon Data, who plan to launch two futures contracts in this specific market in October of this year.

However, one problem persists: the difficulty for individual investors to gain direct exposure to this AI infrastructure market, as industry giants (hyperscalers) control about 70% of the available computing capacity using debt or privately financed infrastructures, against global investment needs that could exceed $5 trillion by 2030.

📰 Bitcoin: Miners Pivot to AI, Hashrate Drops

Consequently, Grayscale analysts identify two categories of publicly traded companies accessible to investors: operators of data centers specializing in AI/HPC - including Bitcoin miners who have integrated artificial intelligence into their development strategy - and neoclouds, these providers specializing in GPU rental and AI computation.

Many of these companies are former or current Bitcoin or cryptocurrency miners. Investing in Bitcoin miners thus allows for an indirect bet on the value that their connections to the electrical grid could take if they were used to power infrastructures dedicated to AI.

The conclusion of this analysis appears quite simple: << rather than trying to predict which model or application will ultimately prevail, we believe that the most sustainable opportunity lies with holders of computing capacities with secure access to electricity >>.

Source: Grayscale