The artificial intelligence revolution has arrived accompanied by an insatiable hunger for energy. Data centers dedicated to AI workloads and high-density GPUs already consume about 2% of global electricity, a number that the International Energy Agency projects will double by 2030. In the United States, the share of these processing centers in national consumption is expected to jump from 3-4% to 11-12% in the same period.
The problem goes beyond the numbers. If this explosive demand is met by fossil fuels, the climate consequences will be severe. Goldman Sachs estimates that data centers powered by traditional sources could add 220 million tons of CO₂ per year to the atmosphere. Added to this are the diesel generators used as backup, a polluting solution that giants like Microsoft have promised to eliminate by 2030.
Cooling represents another critical challenge. Continuously operating AI servers generate intense heat, requiring cooling systems that consume up to 40% of the data center's total energy. Many projects still rely on millions of liters of water, creating conflicts in water-scarce regions. The case of TikTok in Caucaia, Ceará, illustrates this dilemma well: the company plans a mega data center in a city that has faced water emergencies in 16 of the last 21 years.
Why green hydrogen is emerging as a solution.
Unlike expensive, short-lived batteries or highly polluting diesel generators, green hydrogen allows for the storage of renewable energy for extended periods and its conversion into electricity on demand, emitting only water vapor. This characteristic makes it ideal for both emergency backup and continuous generation, ensuring 24/7 operation with clean and reliable energy.
The technology has already moved beyond proof-of-concept testing to real-scale validation:
- In 2025, Microsoft and Caterpillar demonstrated a 3-megawatt fuel cell system powering a data center for 48 uninterrupted hours.
- In Dublin, Microsoft installed a 250 kW module powered by 100% green hydrogen to test the supply in parts of the campus.
- Amazon, IBM, and Meta are also actively investing in H₂ projects as part of their decarbonization strategies.
Hydrogen acts as an energy vector: it stores electricity obtained from renewable sources via water electrolysis. In the case of green hydrogen, this process is powered by wind or solar energy, resulting in carbon-free hydrogen. Subsequently, fuel cells combine the H₂ with oxygen to generate electricity, completing a 100% clean cycle.
The major strategic advantage is overcoming the intermittency of solar and wind power. During the day, surplus renewable energy can be converted into hydrogen; at night, this hydrogen generates electricity for the servers. Thus, data centers operate uninterruptedly with clean energy, without resorting to the fossil fuel grid as a "hidden supplement," a common practice even among companies that claim to be 100% renewable.
Pecém 2026: the first integrated ecosystem at scale.
The Port of Pecém, in Ceará, offers the most concrete example of this convergence. In November 2025, the federal government approved an investment package in the Export Processing Zone focused on the synergy between data centers and green hydrogen: seven co-located hyperscale data centers (including ByteDance/TikTok) and a co-located green hydrogen/ammonia plant.
The numbers are impressive. R$ 571 billion is earmarked for data center installations and R$ 12 billion for the Casa dos Ventos green ammonia plant. The project is expected to generate 95,000 jobs in construction and R$ 80 billion annually in digital service exports when fully operational.
The strategic advantage lies in the commitment to additional energy production. The companies have committed to using exclusively energy from new renewable sources, a capacity that did not exist before the project. All electricity supply will come from additional wind and solar farms, avoiding overloading the current grid.
The choice of Pecém was not accidental. The region has a privileged geographical location, close to several submarine internet cables (ensuring low latency), in addition to the ZPE (Special Economic Zone) with tax incentives and less bureaucracy for exporting services.
Thermal synergy: transforming challenges into opportunities
Co-location paves the way for integrated solutions that increase overall efficiency. Pioneering projects in the Middle East have demonstrated how to use hydrogen-powered turbines to generate electricity while capturing waste heat and redirecting it to absorption cooling systems.
This cogeneration approach achieves overall efficiency exceeding 80%, compared to approximately 50% for pure electrical conversion. The heat that would otherwise be discarded becomes a useful resource, powering chillers that supply chilled water for data center cooling, virtually eliminating the use of water in evaporative cooling towers.
Large operators like Meta have already announced that all their data centers will be liquid-cooled by 2030, starting with AI workloads. Combined with direct liquid cooling on the servers, thermal efficiency allows large data centers to operate in extreme conditions without exceeding water or energy consumption limits.
The business model that completes the equation.
Co-location creates a win-win model by balancing different economic profiles. Data centers are willing to pay a premium for reliable and clean energy, given their highly profitable model. Hydrogen production has tighter margins and requires cheap electricity, but benefits from firm long-term demand.
For H₂ project developers, having a local anchor client dramatically improves financial viability. Instead of relying solely on nascent global markets with uncertain margins, they guarantee stable consumption that sustains the venture.
Operational flexibility amplifies the benefits. AI training data centers can adjust loads for times of higher renewable energy supply, while electrolyzers consume energy when there is surplus. This orchestration maximizes asset utilization and stabilizes demand for renewables.
Mechanisms such as the ReData special regime in Brazil were created to attract these investments, reducing taxes on equipment and requiring the use of additional renewable energy. There is an alignment of incentives between the public sector, energy companies, and technology giants.
Conclusion
The marriage between AI and green hydrogen unites economic convenience and climate conviction. The convenience comes from complementarity: data centers need demonstrable sustainability and uninterrupted power; H₂ projects need scale and firm buyers. The conviction comes from the imperative to fuel the artificial intelligence revolution with clean energy; otherwise, every advance in AI will come at the expense of the planet.
In 2026, this marriage began to take shape. From the deserts of the Middle East to the beaches of Ceará, a future is emerging in which intelligent algorithms will be powered by green molecules.
