As artificial intelligence drives unprecedented growth in cloud computing, a new sodium-ion energy storage architecture is emerging to address the extreme power volatility of AI-native data centers, combining resilience, cost efficiency and supply chain security.

Sodium-ion energy storage for artificial intelligence
Powering neoclouds with sustainable energy solutions optimized for the volatile loads of artificial intelligence is the goal behind the partnership between Peak Energy and Energy Vault. The two companies are working on the launch of a sodium-ion energy storage architecture designed specifically for AI-native cloud platforms and data centers built around artificial intelligence. On paper, the solution promises a set of capabilities that are increasingly critical for a market expanding at unprecedented speed.
Managing power peaks in AI workloads
Whether an AI model is being trained or handling millions of simultaneous inference requests, the power demand of graphics processing units represents a critical challenge. In both scenarios, energy consumption can jump from a few kilowatts to hundreds of megawatts in a matter of milliseconds. Unlike traditional IT workloads, AI computing loads exhibit highly volatile consumption profiles, requiring energy storage systems capable of absorbing instantaneous demand spikes without compromising grid stability.
In this context, conventional battery systems are poorly suited to managing the extreme power dynamics generated by AI workloads. Most lithium-ion systems are designed to deliver energy steadily over several hours. This does not make them obsolete, but when lithium-ion batteries are forced to deliver extremely high power over very short periods, the cells experience significant thermal stress. Avoiding thermal runaway under these conditions requires extensive cooling infrastructure, increasing both complexity and energy consumption.
Integrating sodium-ion batteries with Vault OS software
The response to this challenge comes from the new architecture co-developed by the two companies. The system combines Peak Energy’s sodium-ion battery technology with Energy Vault’s system design and its Vault OS™ software controls. According to the companies, this integration delivers a range of advantages for AI-focused data centers.
The result is a single platform capable of simplifying electrical design, reducing reliance on uninterruptible power supply systems, lowering cooling requirements and enabling more effective management of highly volatile compute loads.
“The rapid growth of artificial intelligence is exposing the limits of conventional energy infrastructure,” said Marco Terruzzin, chief revenue officer at Energy Vault. “This solution enables faster deployment, lower costs and improved safety.”
Technical advantages and costs of sodium-ion BESS
Choosing sodium-ion battery energy storage systems offers several technical advantages. Rechargeable sodium-ion batteries operate across a wider temperature range than lithium-ion systems, reducing the need for large-scale cooling installations and lowering overall energy consumption. They can also be fully discharged to 0% without damaging the internal chemistry, providing greater operational flexibility.
There are, however, trade-offs. Sodium-ion batteries have lower energy density and currently require higher upfront investment. Market data show that, on average, sodium-ion batteries still cost more than lithium-ion batteries for the same storage capacity. Analysts at Wood Mackenzie estimated average 2025 costs of 52 US dollars per kWh for LFP batteries, compared with 59 US dollars per kWh for sodium-ion systems. “We believe sodium-ion battery costs are declining slightly faster than lithium-ion costs, but we do not expect them to reach parity with LFP before around 2035,” Wood Mackenzie wrote in November last year.
In the United States, however, there are additional reasons to invest in this technology. Sodium-ion chemistry offers the prospect of building new supply chains that are less dependent on China, which currently produces around 80% of the world’s lithium-ion batteries, 64% of refined lithium and nearly 90% of cathode materials.
In addition, the federal budget law approved in July 2025, known as the “one big beautiful bill”, introduced new restrictions on eligibility for tax credits. Projects are excluded from incentives if they are linked to any of four countries: China, Russia, Iran and North Korea.
For initiatives like Energy Vault’s, this matters. Because Peak Energy’s sodium-ion BESS are manufactured in the United States, customers can avoid these restrictions and qualify for domestic content incentives, reducing upfront investment by 10-30%.
Targeting neoclouds and global expansion
The supply agreement between the two partners covers the delivery of 1.5 GWh of sodium-ion energy storage systems. Operational control of the batteries will be managed by Vault OS™, Energy Vault’s proprietary energy management system, enhanced with AI to operate across multiple functions, from demand forecasting and energy arbitrage to cell health management.
“Reducing energy costs is essential to winning the AI race, and our proprietary sodium-ion energy storage systems offer the fastest and most cost-effective way to connect data centers to the grid,” said Landon Mossburg, chief executive officer of Peak Energy. “This partnership confirms that Peak’s technology is central to meeting AI’s energy needs, now and in the future.”
The rise of neoclouds
Who are the target customers for this new offering? Neoclouds. Initially developed as stopgap solutions to address GPU shortages, this new generation of cloud service providers is now expanding rapidly on a global scale.
According to recent data, more than 100 neocloud operators are active worldwide, around ten of them in the United States alone, each managing dozens of data centers. Including emerging startups and smaller regional players, the total number easily exceeds 200.
Energy Vault has also secured exclusive channel partner rights for Peak Energy’s sodium-ion technology in Australia and Japan.











