An answer to the datacenter energy crisis

Partner Content This
year, the global build-out of datacenters has become impossible to
ignore, with the debate spilling into national media, local
newspapers, and community council meetings alike. From Arkansas to
Southern California, Nevada, Pennsylvania, West Virginia, and most
recently Box Elder, Utah, communities are weighing the economic
promise of datacenter expansion against mounting concerns over
energy, infrastructure, and residential impact. The same dynamic is
playing out in the UK, where OpenAI’s “Stargate UK” project
has been partly shelved
amid energy consumption concerns and regulatory pressure.

A typical new hyperscale datacenter can face grid-connection
bottlenecks of up to seven
years in certain markets, well before the necessary
transmission, substations, generation capacity, and transformers are
in place. McKinsey, meanwhile, estimates that global datacenter
spending could reach $7
trillion by 2030 – a figure comparable to the size of a
top-12 global economy.

AI intelligence at scale now dominates enterprise strategy and
global politics because the promise of the technology is matched only
by the infrastructure required to deliver it. Energy consumption is
unavoidable in this new world, and the bet enterprise leaders are
making is that the value AI creates will outstrip the cost of the
power feeding it. That trade-off has produced a new equation for
executives: intelligence
per watt.

Is your agentic ambition constrained by energy?

AI-driven datacenters already account for
roughly 1.5 percent of global electricity consumption, and the IEA
expects that demand to more
than double by 2030, approaching three percent of global
electricity use. That’s more than many major industrial sectors,
including agriculture.

The pressure will compound over the next three years, with IDC
projecting onebillion
agents running 217 billion daily actions by 2029. From
Seattle to Barnsley in the UK, the race to build more datacenters
close to energy sources is now a daily occurrence.

If the right datacenter, grid, and power infrastructure for the
first billion agents takes up to seven years to build, supporting
two, three, or even eight billion agents implies timelines the
industry has yet to cost. The mismatch between enterprise intent and
energy capacity is widening.

For enterprise leaders, this is a defining moment of decision.
With 95
percent of global enterprises intending to become their
own AI and data platforms in less than 780 days, AI, data, and energy
can no longer be treated as separate priorities; they are now
interconnected parts of a single platform strategy. The harder
question is how executives can pursue those AI ambitions while
managing energy efficiently at agentic scale.

BFSI might be showing us the way forward

Banking, financial services, and insurance
(BFSI) enterprises have traditionally invested more heavily in
technology than any other major sector. McKinsey estimates banking IT
spending typically runs at between
six and 12 percent of revenue, compared with 3.75 percent
to five percent for the next-highest sector.

The pressure to deliver new technology value, particularly through
AI and agentic systems, is creating an operating language shared by
CIOs, CTOs, and business leaders alike. AI and data are increasingly
framed as the new competitive moat, yet the energy costs associated
with maintaining that moat introduce a fresh dynamic into technology
decision-making.

The 13
percent of global enterprises winning with AI and agentic
systems are more likely to build their data strategies around
control, efficiency, and sustainability. The common pattern is
repatriation: pulling AI and data out of single-hyperscaler silos and
into their own control planes, where they can govern and manage
information across clouds, on-premises environments, and systems they
own.

The pattern recurs among agentic AI leaders across EMEA, North
America, Singapore, and Japan. The principle is straightforward:
bring AI to the data, because the two must work together across the
front lines and back offices of the business rather than operating as
separate concerns.

That logic explains why BFSI leaders such as Wells Fargo,
Mastercard, HSBC, JPMorgan Chase, Bank of America, Citigroup, Goldman
Sachs, BNP Paribas, ING, Crédit Agricole, UBS, and NatWest have made
public carbon-neutrality commitments alongside ambitious plans to
become their own sovereign AI and data platforms.

AI and data sovereignty in Postgres wins on OpEx, environment, and
ROI

Agents operate at the data layer, which
means energy must be managed at the same layer, since this is where
much of the work happens. The alternative is the equivalent of
turning on the heat while leaving every window open in the middle of
winter. Only by controlling the data layer, agents, and broader data
estate can enterprises build the foundation for managing energy
consumption.

Energy efficiency has to begin where enterprise operations already
run, which is why PostgreSQL^®, the world’s most widely used database
among developers, is well suited to the challenge. EDB
Postgres AI is built specifically to address the
energy-intensive nature of modern datacenters by improving database
and AI efficiency at the point where workloads execute.

By shrinking core usage requirements and tightening data-intensive
agentic operations such as search, retrieval, and vector indexing,
EDB Postgres AI can cut datacenter energy consumption by up to 81
percent and reduce emissions by as
much as 87 percent.

The ambition to become an AI and data platform carries one
foundational requirement: AI and data sovereignty. Organizations that
adopt this model not only achieve 5x
ROI and deploy 2x more AI and agentic AI systems; they
also gain more control, greater efficiency, and a smarter way to
design and operate datacenters for the agentic era.

The formula for success is sovereignty in Postgres — the most
practical path to achieving more intelligence per watt.

Contributed by EDB.