This has led to a phenomenon some call the "Great Techno-Industrial Adjustment," a moment that marks the end of the era of unlimited digital growth and a reckoning with an immovable "physical ceiling": the capacity to generate and distribute electrons.
A crisis that is not only redrawing the strategies of giants like Microsoft, but is also forcing the Bitcoin mining industry to reinvent itself as the new backbone of data center infrastructure, all within a volatile geopolitical context marked by the conflict in Iran and shortages of essential electronic components.
Anchorage and Google launch Agentic Banking, banking for AI agents
The end of energy abundance
But why is all this happening? Well, bear in mind that the mass adoption of large language models (LLMs) and generative AI services has drastically altered the structure of data center demand. While a traditional Google search consumes approximately 0.3 Wh, an interaction with advanced models such as ChatGPT requires about 2.9 Wh, representing a 1000% increase in energy intensity per query.
This disparity has led to projections in which AI could account for more than 50% of all data center energy consumption by the end of 2026. At the systemic level, the electricity consumption of these facilities is heading toward 1000 TWh by 2030, a figure that rivals the national consumption of powers such as Germany and France.
An increase of more than 200% in consumption
But this demand is not only quantitative, it is qualitative. Unlike traditional digital workloads, AI requires extreme power density.
Let's put this in context. In 2020, the average density of a data center rack was approximately 8.4 kW; but by 2026, clusters of latest-generation GPUs, such as the NVIDIA GB200 NVL72 system, demand more than 132 kW per rack. This technological leap has rendered conventional infrastructures obsolete, creating fierce competition for "electrons" between the software industry and cryptoasset mining.
This situation makes the energy crisis manifest as a "physical governor" of growth. In nerve centers such as northern Virginia, wholesale electricity prices have seen increases of up to 267% over the past 4 years, limiting the expansion of other industries and putting pressure on local inflation. This reality has forced technology developers to seek solutions outside traditional interconnection models, turning to on-site generation and large-scale energy agreements.
Cloudflare and Stripe team up so AI can set up and run businesses for you
Microsoft and the clean energy crossroads
Perhaps the greatest example of this situation is precisely Microsoft, a company that has historically led environmental responsibility rankings in the tech sector and now faces a fundamental contradiction between its AI ambitions and its sustainability goals.
In fact, its flagship 2030 goal, known as «100/100/0», sought to match every hour of electricity consumption with zero-emission energy purchases on the same grid. However, the reality of 2026 shows that this goal is logistically unattainable given the speed at which its AI infrastructure is being deployed.
Since late 2022, Microsoft's carbon emissions have risen 23%, while competitors such as Google and Meta have reported increases of 51% and 64% respectively. As a result, the company's leadership has initiated internal discussions to delay or formally abandon the hourly matching goal. All because the high cost of clean energy and the intermittency of sources such as solar and wind do not allow powering 132 kW racks that require 99.999% availability.
Against this backdrop, natural gas has established itself as the "fuel of choice" for the sector, due to its speed of deployment compared with the long approval cycles of renewable projects. And Microsoft is even investing to restart the Three Mile Island nuclear reactor (which will be operational again in 2027), which indicates the level of pressure they are under on this issue.
The war in Iran and the supply crisis
As if the energy problem were not enough, the tech sector is also battling the consequences of the war in Iran. The conflict has caused a systemic disruption to maritime traffic in the Strait of Hormuz, blocking the transit of 84% of crude oil destined for Asian importers.
This situation has sent operating costs soaring in key manufacturing hubs such as Taiwan, Japan and South Korea, where electricity prices have reached historic highs, forcing energy rationing for the electronics industry.
Mythos, Anthropic's AI that has raised alarms over its destructive speed
The crisis in printed circuit boards (PCBs) and critical resins
A devastating side effect of the conflict has been the attack on the Jubail petrochemical complex in Saudi Arabia, operated by SABIC. This plant is responsible for producing 70% of the world's supply of high-purity polyphenylene ether (PPE) resin, an indispensable material for manufacturing the laminates used in printed circuit boards (PCBs). PCBs are the fundamental substrate of all electronic devices, and their scarcity has caused a 40% price increase in a single month.
So it should not surprise you if, on your next visit to the tech store to buy a smartphone or any electronic device, you see that prices have risen considerably. This supply crisis acts as a cost multiplier for AI and cryptoasset infrastructure. Server manufacturers such as Dell and Lenovo have in fact already begun passing these extra costs on to end customers, raising the capital expenditure (CapEx) needed to sustain the expansion of computing capacity.
Cryptoasset miners as AI infrastructure partners
In the midst of this cost storm, Bitcoin miners have found a new strategic purpose. Historically criticized for their energy consumption, they are now seen as holders of invaluable assets: approved power contracts, operational electrical substations and land with access to dark fiber. As a result, the industry is undergoing a deep reorganization, splitting between "pure miners" and "hybrid infrastructure providers."
Early monetization and capital optimization
Bitcoin mining has established itself as a financing tool for AI infrastructure. Since building AI data centers can take more than 18 months, miners use already-contracted power to generate revenue from day one through modular ASIC equipment. Once the data center is ready to host AI GPUs, mining does not disappear; instead, it becomes an "adaptable load" that consumes surplus energy or scales back during periods of peak prices.
Therefore, and to survive a world of high energy costs and geopolitical volatility, mining farm and data center operators have implemented cutting-edge technical optimization strategies. Efficiency is no longer just a sustainability metric, but a condition for economic survival.
Anchorage and Google launch Agentic Banking, banking for AI agents
An example of this is the so-called Virtual Power Plants (VPPs), which have gone from pilot projects to a strategic imperative for the power grid. VPPs aggregate distributed resources such as industrial batteries and mining loads to provide flexibility to the grid.
Basically, miners actively participate in demand response programs, where they are paid to automatically reduce their load during periods of system stress. This automatic adjustment capability is what allows many mining facilities to obtain faster interconnection approvals, since utilities see them as an asset that improves grid stability rather than a threat.
In any case, the current energy crisis has shown that the digital economy is not immaterial, but is deeply anchored in the physical reality of power generation and global logistics. Competition for energy between AI and digital assets has forced an accelerated maturation of the tech sector, eliminating inefficient business models and consolidating a new hybrid infrastructure architecture.
Microsoft's dilemma underscores that even the most powerful corporations must yield to the limitations of the grid. The return to nuclear power and the acceptance of natural gas as a necessary evil mark a climate realism that prioritizes technological sovereignty.
On the other hand, the transformation of miners into AI data center operators represents one of the most successful industrial adaptations of the decade, turning a previously stigmatized activity into the pillar of digital resilience.
Data, energy and blockchain: the pillars of the new economic system
Ultimately, success in this new paradigm will depend on the ability to integrate fast energy storage systems, advanced cooling and stochastic optimization models that can navigate geopolitical and industrial volatility. Energy has ceased to be an operating cost and become the sovereign resource par excellence, and whoever controls its flow and efficiency will control the future of global computing, amid a crisis that looks set to stay with us for a good couple of years.