During the dot-com era, 2,888 related startups went public. So far, zero generative AI startups have sold their shares to the public.
That could change soon if AI data center provider CoreWeave’s IPO succeeds. And Crusoe Energy Systems – a Denver, Colo.-based supplier of renewable energy powered AI data centers – could follow suit.
Generative AI has a huge appetite for energy. More specifically, U.S. data centers where AI chatbots are trained and operated consumed a whopping 46 gigawatts of energy in the third quarter of 2024 – a figure expected to grow to 59 GW by 2029, according to CarbonCredits.com.
With digital services, cloud computing, and AI applications driving growth in demand, investors have recognized the opportunity to profit from financing the building of data centers.
Indeed, since 2018, the global investment in “new data center capacity is estimated to be around $443 billion on average through 2028,” according to data from PitchBook and Moody’s.
A tiny proportion – about 4.5% – of that capital is going to build AI-optimized data centers powered by renewable energy sources such as wind and solar. Google, Microsoft, and Meta are planning to invest “up to $20 billion in renewable energy projects to directly power their new data centers,” according to Canary Media.
One recipient of that capital is Crusoe which could be siphoning some revenue from AWS and other hyperscalers. Meanwhile AWS is investing in renewable projects. The company is “continuing to invest in new sources of solar and wind energy,” as well as nuclear, according to a March 4 email from an Amazon spokesperson.
Crusoe – which in January 2025 announced a $600 million Series D funding round, at a $2.8 billion valuation – envisions an IPO at some point. “We will go public but we don’t have a specific plan,” Crusoe CEO and co-founder Chase Lochmiller told me in a February 21 interview.
In the meantime, Crusoe’s very rapid growth compares favorably to cloud industry leader AWS’ 19% revenue growth in 2024.
Amazon is proud of its actions on renewable energy. In 2019, the company pledged to be net-zero carbon by 2040. Amazon has invested in more than 600 solar and wind projects and matches 100% of the energy used in its operations with the energy produced by its renewable investments, according to a company release.
Powering Generative AI Data Centers
Generative AI will consumer vast amounts of electricity and capital in the years ahead.
To train and operate large language models, OpenAI, Meta, and Google require huge amounts of electricity to power AI chips – notably graphics processing units from Nvidia housed in servers.
Indeed, training LLMs such as GPT-4 requires thousands of GPUs consuming as much as 1 gigawatt-hour per model. To satisfy that demand, AI-driven data centers will consume an additional 200 terawatt-hours of electricity annually from 2023 to 2030, notes Goldman Sachs Research featured by CarbonCredits.com.
Tech giants Amazon, Google and Microsoft operate 445 data centers in the U.S. and have 249 in the pipeline, according to a Synergy Research Group report featured by the Wall Street Journal. The $50o billion Stargate project plans to build at least 20 such data centers – spending “hundreds of billions of dollars annually,” noted the Journal.
Before the AI boom, companies used data centers primarily to store information and run applications. To connect to people faster, data centers were located near population centers and technology hubs, noted the Journal.
For that reason, most data centers are located near Washington. D.C. “Northern Virginia is the data center capital of the world,” Lochmiller told me. “That’s because people needed to build internet applications with a few milliseconds of latency. That is not so important for training and inferencing of LLMs. For that, concerns about energy availability are more important.”
That is why more data centers are being built in rural Texas. “It doesn’t make sense to be in New York City where power is better used to operate big office buildings or high-density housing,” Lochmiller told the Journal. “But in a place like West Texas where you don’t have a lot of people but you have land and power, that’s a great fit,” he added.
Crusoe Energy Systems Growth And Strategy
Crusoe – whose latest funding is backed by Peter Theil’s Founders Fund and others, according to a company release – is “developing Stargate’s Abilene data center and building a gas power plant that will provide backup energy to it,” the Journal reported.
Last October, Crusoe announced this data center’s construction which is scheduled to be completed soon. With more than 1.2 GW planned, Crusoe said the project will “inject $1 billion into the local economy over the next 20 years and create more than 800 local jobs,” according to the company release.
Crusoe’s data centers help meet the demand for data center capacity and energy. “In 2025, nobody has enough data center capacity or energy,” Lochmiller told me. “The density has increased from between two and four kilowatts per data center rack up to 100 KW/rack with Nvidia’s Blackwell,” he added
Crusoe’s Abilene facility meets such demand efficiently. The Abilene plants “has wind and solar at a low cost – creating a negative price point,” he said. “The community needs us. People care less about low cost and more about reliability, security, and sovereignty. We look for naturally windy and sunny regions such as Nevada and West Texas,” he added.
How Crusoe Got Started
Lochmiller earned degrees in physics and computer science from MIT and Stanford respectively and spent a decade using computers to manage money. “Before Crusoe, I spent 10 years as a quantitative portfolio manager,” he explained.
“We were deep learning early adopters. We were actually predicting when the current price was different than what it should be. We were large users of computing infrastructure and deep learning. I saw how much computing was required to do intelligent workloads at scale,” he added.
Lochmiller recognized the broad applicability of the computing architecture required to do deep learning. “It was a meta science,” he said.
He left quantitative finance and went into digital assets, joining Coinbase as an early employee and becoming a partner at Polychain Capital. “Adding to individual intelligence was going to be broadly deployed across every sector,” he said. “There was so much energy required to run computing tasks at scale.”
Before joining Crusoe, he wanted to do something on his bucket list. “In 2018, I left to climb Mount Everest,” he said. “After that I went to Crusoe which was developing an AI cloud platform that could be used for digital currency mining. We got our first publicly-paying customers in early 2022.”
How And Why Crusoe Has Grown
ChatGPT propelled demand for Crusoe. “When ChatGPT launched in November 2022 it transformed the arc of history.” he said.
“There was a boom in demand for GPU computing. Crusoe’s mission was to be a competitive AI cloud provider using a data center architecture that would transform power density and thermal complexity.”
Crusoe’s customer base has grown by over seven-fold in the past 12 months and the company employs more than 550 people, according to a company release.
Why such fast growth? Crusoe’s two services — AI-focused data centers and AI cloud services – give customers more of what they want. “We win the competition for data centers by providing purpose-built 100 MW data center architecture and much faster execution,” Lochmiller told me.
“We complete construction in less than a year– a data center we started in June 2024 will be finished in February 2025. This compares to two or more years for Microsoft and Oracle,” he said
Crusoe’s AI services offer more bang for the buck. “We win in cloud services because of our narrow AI focus, our better performance, price, support and solution implementation in text, video, and image generation,” he added.
He sees a bright future for generative AI. “As performance gets better, models are smarter, there is more utility, costs go down and people use them even more,” he said.
Companies are using generative AI to develop code more efficiently which reduces the number of engineer-hours to complete the task. Other uses include boosting the productivity of call center agents and sending out press releases around the world.
“More energy usage is not bad,” he concluded. “It’s actually good and correlates with better quality of life. The negative externality is climate. Crusoe’s mission is to provide more clean energy and make it economical.”
In 2023, traditional cloud providers such as Azure and Google Cloud had more than six month waitlists for NVIDIA H100 GPUs, with spot market prices reaching $8/GPU-hour, researcher Sacra. Crusoe’s ability to deploy GPUs quickly using “stranded energy” likely siphoned off customers frustrated by delays.
Will Public Company Investors Tap Crusoe’s Growth?
AWS is huge yet is growing more slowly than Crusoe. The cloud computing giant achieved $108 billion in 2024 revenue, according to Amazon.
Yet Crusoe may be growing much faster. Crusoe’s revenue grew 82% from $150 million in 2023 to $276 million in 2024 – with Crusoe Cloud, the company’s AI-focused offering, rising from 14% to 45% of revenue, Sacra estimates.
How much would Crusoe be worth in an IPO? I don’t know – however, assuming Crusoe maintains its current growth rate, 2025 revenue could total between $500 million and $550 million.
Crusoe rival CoreWeave filed for an IPO – valuing its $1.92 billion in 2024 revenue at $35 billion, according to Fortune. Applying that 18.2x revenue multiple, Crusoe’s IPO could be valued at $9.1 billion.
If CoreWeave goes public and its stock price soars, Crusoe could follow.
Read the full article here
