"At the end of AI is electricity" narrative heats up, "Electricity upstart" ERock is pushing for a US stock IPO to raise $600 million.

ERock announced its initial public offering (IPO) terms and regulations in the US stock market on Monday. The company focuses on manufacturing and deploying natural gas power generation systems for enterprises or large AI data centers. The timing of ERock's decision to go public is considered the best in terms of the company's valuation prospects, as the global AI super wave is facing an "electricity supply shock." The delivery cycle for large gas turbines in the United States is already booked until 2029, and electricity resources in the construction process of large AI data centers focusing on AI training and inference have transformed from "supporting resources" to "one of the biggest bottlenecks in the AI computing power industry." The natural gas power company based in Houston, Texas plans to raise approximately $600 million by issuing 27.9 million shares of stock (28% of which will be sold in a non-standard IPO structure as a synthetic secondary market). The company plans to raise funds in a price range of $20 to $23 per share. Based on the midpoint of the proposed price range, ERock's fully diluted market value will reach approximately $5.9 billion. ERock is a vertically integrated distributed power supply and deployment company that designs, deploys, operates, and maintains modular natural gas generator systems for multiple data centers, utilities, and large commercial and industrial clients in nine states in the US. The company deploys its natural gas power systems in three configuration modes: bridge power, providing main power supply system before integrating with the grid system; backup power to support continuous operation of large power-consuming scenarios such as data centers during grid interruptions; and dispatchable power, providing flexible power capacity on-demand for peak load management and grid stability. ERock generates cash flow through sales and installation services of natural gas power systems before operation, as well as ongoing operation and maintenance contracts post-deployment. Founded in 2018, ERock achieved comprehensive revenue of $191 million in the 12 months ending March 31, 2026. As mentioned above, ERock focuses on designing, deploying, operating, and maintaining modular natural gas generator systems for data centers, utilities, and large industrial and commercial enterprises. Therefore, the company's core revenue streams include sales and installation services of power systems before operation, as well as long-term operation and maintenance contracts after deployment. The application scenarios mainly include "bridge power" before grid connection, backup power during grid interruptions, and dispatchable power for peak load management and grid stability. The company plans to list on the New York Stock Exchange under the ticker symbol "EROC." Morgan Stanley, J.P. Morgan, Barclays, and Bank of America Securities are the joint bookrunners for this transaction. The company is expected to price its IPO in the week of June 8, 2026. The end of AI is electricity! The AI infrastructure frenzy has spawned the "new nobility of electricity." As the demand for AI computing power infrastructure dominated by NVIDIA AI GPUs and Google TPU power clusters becomes stronger, electricity resources have evolved from a backend cost to a bottleneck in the forefront; whoever can deliver stable power systems faster may become a critical variable in the pace of data center construction. Therefore, ERock's decision to go public at this time is extremely precise, aiming to leverage the market narrative of "the end of AI is electricity" to achieve financing and valuation pricing in the most optimistic window. The global AI super wave is facing an unprecedented "electricity supply shock." In the United States, the delivery cycle for large gas turbines has already been booked until 2029, and manufacturers such as GE Vernova, Siemens Energy, Mitsubishi Electric have warned that they will be unable to meet the surging global electricity system demand for at least the next 3 years. The global construction and expansion process of AI data centers led by Google, Microsoft, and Meta, the parent company of Facebook, is in full swing, and this process increasingly highlights the importance of electricity resource supply. This is why the investment theme of "the end of AI is electricity" is becoming increasingly popular. More importantly, if the "self-power supply" path is eventually institutionalized in the entire United States and other regions such as Europe, it will undoubtedly transfer a significant portion of AI capital expenditure systematically to the power equipment and grid technology stack. The current policy direction advocated by the US government, based on "self-power supply," requires large-scale cloud computing companies and AI tech companies that increase loads to bear all the costs of energy and infrastructure needed for their data centers, without transferring them to ordinary residents. This means that the Trump administration requires large tech companies to "build their own power plants for data centers," which is no longer an abstract policy slogan, but a push for these tech giants dedicated to massively building or expanding AI data centers towards a "Bring Your Own Power" mode. From the perspective of power system engineering, ERock's value lies not in being a traditional large power plant developer, but in providing modular, distributed, and quickly deployable natural gas power solutions that are closer to the data center delivery cycle. When there is a temporary bottleneck in the delivery of large combined cycle gas power plants, transformers, gas turbines, switchgear, and grid connections often for AI data center owners who urgently need a transition solution of "having power first and waiting for the grid." At the same time, AI inference workloads have high power density and variability, requiring more efficient power supply for backup power, microgrids, and dispatchable capacity than traditional commercial loads. Goldman Sachs, the Wall Street financial giant, recently raised its forecast for the massive power demand driven by global data centers up to 2030, significantly expanding by 220% from 2023 (Goldman's previous forecast was +175%), equivalent to adding the power resource load of another "top ten electricity-consuming country" globally. According to Goldman's strategy analyst team, the end of large AI models is electricity - the agency emphasizes that the AI, known as the "power-consuming beast," will bring an unprecedented global electricity "super-demand cycle" and a "super-bull market" for power stocks.