Data center solution
The demand for data center power generation is unprecedented
Power generation shortage for AI data center demand is solved by an "additively manufactured gas turbine" for short lead times
Novra Power is initially focused on the market of power solutions for data centers. Given that the demand for electricity powering data centers is far outstripping the current capacity of the electricity grid in the US and globally, data center developers are planning off-grid ("behind-the-meter") dedicated power plants for their data centers; in the past, these were usually for redundancy purposes when the grid fails. New installations will also cater for prime power in the absence of grid connection.
The global data center power demand will be 160GW by 2030 at a growth rate of 20% (Source: BCG report - fig. 1 below). In the US, this means 10% of the entire electricity demand will be consumed by data centers, up from around 2% today. The data center market segments into hyperscaler, co-location and enterprise data centers, with projected sizes per project between 20MW and 1GW. Additionally, redundancy requirements range from N+1 to N+N, which translates into a potential doubling of capacity. The majority of the growth and the most Gigawatts will be in larger installations of up to 1GW each (fig. 2). The power installations, being initially off-grid, will have the characteristics of large-scale power projects, as well as giving rise to bridging power opportunities with potential integration into the utility infrastructure at a later point in time.
An interesting opportunity also arises from the AI industry's shift towards inference vs initially more training weighted (fig. 3). This will drive significant volatility into the load profile for very large loads which is unprecedented (for power grids) and supports the use of dedicated power plants and technology (like a large amount of modular gas turbines) that can handle the load profiles, whilst at the same time operate efficiently. Current grid connection queues (the time until a grid connection is available for new builds) are 5-7 years in the US. This leads to many data centers being projected to be powered (initially) by off-grid, gas-fired power plants. (Nuclear is also in the discussion, but has long lead times to become available).
Novra Power offers bridging power
​In response to the grid bottleneck and supply chain challenge for providing timely power to the fast AI data center growth, Novra Power is developing a bridging power solution and expects to be able to offer first deployments by 2027, comprising a large fleet of rental assets based on a highly-proven gas turbine model. The fleet will cater for specific requirements of AI data center operations. Whilst the focus is on time-to-power, there will be a rental model for contract lengths of 5 years and upwards on the basis of capacity charges. At the end of the term, a sale to the utility is one possible option, once the grid connection becomes available. In addition, the assets can be kept for redundancy operations or decommissioned as they are transportable. The core turbine is also suitable for additive manufacturing of critical long lead items like rotor forgings. This leads to an ability to scale the supply chain to continue to grow the fleet quickly to support the AI data center industry into the next decade. Other advantages will be the integration with absorption chilling in a very efficient way due to high exhaust gas temperatures. This supports hyperscaler direct-to-chip cooling integration. Fast load response to cater for inference load profile volatility is another feature.
Additive manufacturing cuts power generator lead times
Novra Power has provisionally patented a gas turbine that is designed to be manufactured with additive manufacturing technologies (3D-printing). This will be applied to existing turbines first and later to new designs and components. This caters for optimal component sizes, material choices and geometries, meaning long lead time items like forgings and castings are not needed. The total part count is significantly reduced, eliminating today’s bottlenecks in the supply chain. Conventional gas turbines, such as those that are deployed for large gas-fired, off-grid data center projects, have today lead times of up to 5 years due to the AI data center driven backlog.
​
With "order to delivery" of Novra Power’s product taking weeks rather than months, lead times are cut by 10x. Additionally, the supply chain can be scaled for GW scale production by multiplying printing machines without increasing lead times. This avoids today’s backlog in the supply chain for power generation equipment that leads to multi-year lead times. Novra Power’s gas turbine has superior performance specs to traditional turbines in line with large industrial gas turbines. This also relates to low NOx emissions being deployable anywhere. It has high power density to enable staking into an industry leading footprint. This is complemented by start-up times for cold start and superior load acceptance profiles and part load performance.
​
By designing a product that enables the industrialization of additive manufacturing for a gas turbine power generator, infinite scaling of power generation manufacturing is enabled in order to solve today’s and future AI-driven energy problems.
The supply chain opportunity
Novra Power revolutionizes the supply chain for large gas turbine power generators that are the prime choice to power the unprecedented growth of AI data centers off-grid.
Novra Power solves the power infrastructure bottlenecks of the AI age.
Novra Power’s proprietary and patented gas turbine design, specifically tailored to be manufactured with additive manufacturing technologies (3D-printing), enables time-to-power to reduce 10x to deliver in weeks vs months/years. It also enables scalability of the supply chain to produce GW of power without increasing lead times or backlogging the supply chain. The design eliminates forging of high stress components as well as castings for large structural components. Both processes are typically responsible for lead times of over one year (even without a supply chain backlog) for large industrial gas turbines or recip engines, which are the "go to" prime movers for data center power outside of the grid.
The time is now for re-developing the supply chain in the era of the power super cycle with lasting benefits for the future of energy. The AI and power super cycle generates grid connection queues of 5-7 years and, at the same time, dedicated power solutions for data centers using eg gas turbines are also suffering from lead times of the same order.