Download Whitepaper

The Problem

The exponential growth of data centers, fueled by the computational demands of cloud services, big data, and, most recently, artificial intelligence, has created an unprecedented energy challenge for electrical grids worldwide. Goldman Sachs Research estimates that the electricity demand from data centers worldwide will increase 50% by 2027 and up to 165% by 2030 from 2023 levels. This rapid increase is straining power grids globally with a massive volume of new capacity being built and connected, notably in Northern Virginia and other US markets. This is resulting in data centers and other large loads being forced to wait, often years, until utilities and grid operators can prepare and ensure sufficient generation and delivery capacity.

To meet this challenge, there is an opportunity to mitigate demand by throttling applications that are not time-critical running in data centers. Existing energy management solutions, which typically operate at a coarse facility-wide level, are no longer sufficiently adequate to meet this challenge. They often fail to differentiate between critical, latency-sensitive application workloads and flexible, non-essential tasks, leading to inefficient curtailment and potential reduction in quality of service or disruption. These limitations underscore the critical need for a new approach that must empower individual applications, VMs, and containers to dynamically adjust their electricity consumption in direct response to real-time and predicted grid conditions.

The Solution

To meet this need, BluWave-ai provides the Data Center Autopilot, a software-as-a-service product using patent-pending technology. It provides application owners, colocation and hyperscale data center operators with a fast-to-deploy solution to manage their immense and volatile power demands in concert with the dynamics of the local grid load. The Data Center Autopilot realizes this by moving tasks that can be scheduled to operate at times where grid congestion and energy costs are lower, smoothing out “spiky” AI workloads while maintaining quality of service and uptime.  

This provides the following benefits for:

Applications Owners: Application owners can subscribe to the service and opt in to energy use reduction during grid constrained times and receive a direct revenue stream from BluWave-ai reducing their overall cost of operations. BluWave-ai aggregates multiple applications within a data center or across data centers in an electricity system operator service area

Data Center Operators: The platform provides joint optimization for all connected assets (on-site batteries, data center management system and applications) to perform peak demand reduction and energy arbitrage. This delivers a demonstrable ROI with net reduction in total electricity expenditure and measurable improvements in power usage efficiency using existing hardware assets.

ESG and Carbon Emissions Reduction: By preferentially scheduling workloads to run during periods of high renewable energy availability, the solution actively helps operators achieve aggressive carbon usage effectiveness targets and meet net-zero sustainability mandates which are quantified by the platform for ESG reporting purposes  

A central element of the Data Center Autopilot is BluWave-ai’s "Grid Sentiment" service, BluScore, which provides a composite, actionable signal about the electrical grid's state. This goes beyond simple energy price data, encompassing critical factors such as grid health, carbon emissions, renewable energy content, and real-time market pricing, as well as solar, wind, and load forecasts. The Data Center Autopilot uses this signal acting as a management layer sitting above traditional hypervisors (e.g., Hyper-V, Proxmox). Its purpose is not to perform the primary functions of VM orchestration, such as provisioning or migration, but to focus exclusively on translating the BluScore signal into actionable energy-saving commands. By acting as a central controller, the Data Center Autopilot ensures that power management is coordinated and effective across multiple virtualized environments and entire server racks, providing a unified, top-down approach to energy management

The Data Center Autopilot encompasses two control models: top-down and bottom-up. With the top-down model, it acts as the central command and control hub. When it receives a signal indicating poor grid conditions, it initiates a power reduction command that is disseminated to the data center's existing management systems or traditional hypervisors. This model is ideal for single-tenant, purpose-built data centers or colocation facilities that require centralized control. In the bottom-up model, the control and response are decentralized and autonomous. Individual applications or VMs directly subscribe to the Data Center Autopilot. When these subscribing applications receive a BluScore signal, they autonomously adjust their resource consumption. The bottom-up model is perfectly suited for multi-tenant, hyperscale environments where individual tenants and applications can be managed and incentivized independently. This approach is particularly effective for flexible, latency insensitive workloads, such as large language model training, batch processing, or other background tasks.

The Data Center Autopilot offers a much-needed source of predictable and dispatchable load reduction that can be aggregated to provide a significant, verifiable response to grid signals. This enhances grid stability, reduces strain during peak demand, and improves the integration of intermittent renewable energy sources, ultimately contributing to a more resilient and sustainable energy future. By enabling targeted power reduction at the application and VM level, it reduces data center operation costs, provides new revenue streams through demand response payments, and enhances sustainability by facilitating the use of cleaner energy sources.