Date -2026-04-22 Location - WILMINGTON, Del.

EnterpriseDB (EDB), the leading sovereign AI and data company, today announced a suite of validated performance efficiencies within EDB Postgres AI (EDB PG AI), designed to drastically reduce data center power consumption, lower token usage, and deliver an unprecedented “intelligence per watt” standard for the enterprise.

By 2027, half of enterprises will be using AI agents to redefine how humans and machines collaborate. By 2030, more than 1 billion agents will be actively deployed and will execute roughly 217 billion actions a day. These agents will consume trillions of tokens daily, and electricity demand from data centers worldwide will more than double by 2030 to around 945 terawatt-hours (TWh), with AI being the most significant driver of that increase.

“The AI energy conversation has been about what happens with the models and GPUs. Almost nobody is talking about what happens at the data layer that every agent, every model, every inference call depends on,” said Quais Taraki, CTO at EDB. “You can’t control consumption at the model layer. Agents consume what they consume. But you can control efficiency at the data layer, and for most enterprises, that’s the only lever they actually have.”

Sovereign architecture unlocks efficiency at the core and the data layer

EDB PG AI addresses the agentic energy challenge on two complementary fronts: first, by shrinking the core infrastructure footprint required to run enterprise applications; and second, by making the data-layer operations that power agentic AI—especially search, retrieval, and vector indexing—far more efficient. Together, those gains improve not just how much infrastructure enterprises need but how effectively that infrastructure is used per unit of energy.

At the infrastructure level, EDB PG AI helps enterprises reduce the servers and cores required to run applications, lowering both data center energy use and emissions. An analysis of three BFSI customers operating more than 120 data centers, independently validated by Incendium Consulting, showed up to 94% reduction in compute cores in one case, resulting in up to 87% expected emissions reduction—approximately 153,000 metric tons of avoided CO2e, equivalent to removing 33,000 cars from the road.

At the workload level, EDB is targeting one of the most underappreciated drivers of AI energy cost: the intensive data-layer operations as agents create databases, adjust queries, and move data across enterprise environments 24/7/365. Building and maintaining vector indexes is among the most resource-intensive activities in modern databases—and one that scales directly with the number of agents in production.

New benchmarks show EDB PG AI delivers:

• 5x–12x faster vector index builds with comparable or superior throughput at 1 billion vectors on a 128 GB server versus 1,000+ GB for traditional vector engines—a step-change reduction in the compute and memory required for AI-scale data retrieval*
• Up to 57% reduction in AI token consumption, with 90% quality preserved and a 72% scenario win rate, demonstrated in a pilot with a leading global telecom provider—directly reducing the energy cost of every agentic interaction

These results build on EDB PG AI’s broader efficiency gains, including 50x–100x faster analytical workload completion on live operational data, and up to 58% lower cost with the lowest concurrency degradation among leading cloud analytics platforms. These are capabilities that reduce the energy overhead of data storage, retrieval, and analysis across the enterprise stack.

The intelligence per watt framework

Building on these demonstrated efficiency gains, EDB PG AI delivers an “intelligence per watt” standard for global enterprises to measure, improve, and operationalize AI efficiency at scale, as autonomous systems create more databases, pipelines, and queries over time.

The platform is built around three principles that compound as agentic workloads scale:

• Measure: Quantify the energy and infrastructure cost per unit of AI intelligence produced, extending the Incendium-validated methodology to agentic, RAG, and multi-agent workloads.
• Optimize: Reduce compute, storage, and network demand per AI operation through database consolidation, storage tiering, query acceleration, vector indexing, and token reduction.
• Govern: Maintain visibility and control over data layer operations as autonomous agents create databases, indexes, pipelines, and queries at machine speed.

“Enterprises succeeding with AI at scale are 275% more likely to prioritize energy-efficient data infrastructure than the rest of the market. They’re also seeing 5x the ROI. That’s the connection most of this industry is missing. This idea of ‘intelligence per watt’ isn’t just an environmental metric—it’s a performance indicator. The companies getting the most from AI are the ones demanding the most from their data layer,” said Kevin Dallas, CEO of EDB. 

Organizations can quantify their own intelligence per watt with the EDB PG AI Efficiency Calculator at www.enterprisedb.com/calculator/efficiency, or visit enterprisedb.com to learn more.

*Based on completed and independently validated EDB benchmarks. Published report forthcoming.

About EDB

EDB Postgres® AI (EDB PG AI) is the first open, enterprise-grade sovereign data and AI platform—secure, compliant, and scalable, on-premises and across clouds. Built on Postgres, the world’s leading database, EDB PG AI unifies transactional, analytical, and AI workloads, enabling organizations to operationalize their data and LLMs while maintaining control over sovereign environments. EDB PG AI is supported by a global partner network and delivers up to 99.999% availability as well as hybrid management and a built-in AI factory. As one of the most active contributors to the PostgreSQL project, EDB is deeply invested in the vitality of the global community. To learn more, visit www.enterprisedb.com. 

Media Contact:

Steph McGuirk 
stephanie@interdependence.com
(845) 269-8868