Large technology companies resort to hydrogen AI data centers to feed

Yuval Bachar knows the data centers. Has worked on them to Goal, Microsoft and CiscoBut now, its start is looking to help Silicon Valley execute data centers with lower carbon dioxide emissions.

ECL, the Bachar startup, builds data centers with hydrogen.

Hydrogen is a new energy source for data centers that is more ecological and, more importantly, for technology companies that need to quickly expand their infrastructure, data centers that are executed in hydrogen can be put into service in the Half the time needed to build centers data that connects to the grid, Bachar said.

There is one of these data centers with hydrogen, with a capacity of 1 megavatio miserable, next to the ECL headquarters in Mountain View, California. Twice a month, a diesel truck is taken into hydrogen in a tank in southern California or northern Nevada. Hydrogen is mainly derived from natural gas, which is the main source of energy for electricity in the US.

Bachar and others that develop technologies that can feed data centers with minimal broadcasts discuss their work in a new CNBC documentary, which can be seen before.

Since Operai launched chatgpt in 2022, Amazon, GoogleMicrosoft and other companies have been running to open data centers that can handle generative artificial intelligence. These buildings are usually full of energy Nvidia Graphic processing units. GPUs are the standard for training and execution of large language models that produce impressive pieces of text with some words of human entry. Executives in all industries have seen what Chatgpt can do, and now they want to infuse generative in their internal products and operations, sometimes with the hope of increasing productivity.

If your data center does not have enough power for GPUs today, executives will look elsewhere. Bachar knows it. It is a large part of its launch.

He likes to say that public services in some places, such as California and Virginia, cannot help him at this time if he wants a lot of energy for a data center. Sam Altman de Openai has invested hundreds of millions in new nuclear companies, but they will not be ready to deliver energy for years, Bachar said.

After establishing ECL in 2021, Bachar has registered two clients who pay, with several other organizations that have made orders for future delivery.

“They are the Microsoft, Facebook, Amazonas and Google of the World … that require that all this technology be placed somewhere, and at this time, somewhere is not anywhere,” said Bachar, explaining that the centers of traditional data in the United States cannot be easily reused to work with AI.

ECL has plans to operate its sites efficiently, but from now on, it is small, with 10 employees and 18 contractors. That is much smaller than Altman’s nuclear fusion investment, Helion, and the fission startup he supported, Oklo. Together, the two use almost 600 people, representatives said.

Microsoft has pledged to work with Helion, and the software company also signed an energy purchase agreement in September to restart a nuclear reactor in Pennsylvania Tres Miles Island That closed in 2019.

Inherently, inherently impose questions about safety and waste management, but their carbon without carbon makes them attractive. Amazon, Google and Oracle have everything explored Small modular reactors with less capacity than Three Mile Island.

He Large technology companies are carefully observing their emissions in the AI ​​era.

By 2030, Google wants to have zero net emissions, while Microsoft’s goal is to be a negative carbon for that year. Amazon has pledged to reach net-zero carbon by 2040.

“We are working with the main technology companies, as well as with several industrial players, to help them integrate our plug and game solution for the generation of energy at the site in data centers,” Bret Kugelmass, founder and CEO of Last Energy , a Washington, a Washington. Startup working on small modular reactors.

Bachar is fascinated with nuclear energy, but said that obtaining more of those online facilities will take time.

“We have a problem that we have to solve right now,” he said.

In addition to its nuclear investments, Altman de Openai has opted for the solar exowatt startup. It has partners who develop data centers that consume more than half of the energy available in their states in some places, said co -founder and CEO Hannan Happy.

Geothermal energy has also generated a new interest in the modern era of AI, with Google collaborating with Startup Fervo Energy In Nevada. Tim Latimer, the CEO of the startup, said Fervo has found a way to generate electricity gigawatts in one place drilling horizontal holes underground, instead of the traditional vertical form.

Gigawatts are a serious amount, but holes drilling for geothermal plants can be expensive, said Adrian Cockcroft, a former Amazon sustainability executive.

ECL intends to build a large -scale data center of 1 Gigawatt in Texas over the next four years, with the help of hydrogen pipes. It will probably have been moving to zero carbon green for so long using electrolyzers that convert water into hydrogen and oxygen, Bachar said.

But generating green hydrogen through electrolysis is not cheap, said Kittu Koluri, managing director of Neotribe Ventures.

The price of green hydrogen must be determined, especially now that Donald Trump He is the president of the United States again, Bachar said.

Even so, each Gigawatt matters.

In 2028, the demand of the US Data Center December report of the National Laboratory of Lawrence Berkeley. Data centers can represent 6.7% to 12% of the total energy consumption of the USA in 2028, compared to 4.4% in 2023, according to the report.

“The concern we have is that we can grow quickly enough to address the unprecedented demand of AI data centers,” Bachar said.