On September 16th 2026, the U.S. House passed legislation addressing the electricity infrastructure costs created by large data centers by an extraordinary 417–3 vote.
The basic principle is straightforward: state utility regulators should ensure that data centers bear the costs of the new generation and transmission infrastructure required to serve them rather than shifting those costs onto households and small businesses. AP News
That vote matters.
But I think the energy industry should take the conversation one step further.
The question shouldn’t only be:
“Who pays for the next megawatt?”
It should also be:
“Do we actually need to build that megawatt?”
This week, Google, NVIDIA, Emerald AI and other energy and technology companies launched the AI Energy Management Alliance, focused on making data centers flexible grid participants capable of reducing electricity consumption during constrained periods. Axios
These two developments belong together.
Imagine a proposed 500-MW data center.
Traditional planning might assume:
500 MW load → 500 MW grid requirement → infrastructure upgrades → customer/ratepayer cost allocation
A flexibility-based model could look different:
500 MW peak demand
− 75 MW flexible compute
− 50 MW BESS
− 25 MW demand response
= 350 MW firm grid requirement during constrained periods
Those numbers are illustrative, but the architecture isn’t.
And today we’re getting another reminder of why it matters.
Reuters reports growing community opposition to AI data-center expansion in Silicon Valley over electricity consumption, water use, pollution and how infrastructure costs are allocated. California policymakers are considering additional requirements around energy costs and transparency. Reuters
This is where I believe flexible interconnection becomes much more than an engineering concept.
It becomes an affordability mechanism.
Instead of forcing utilities to choose between:
Build everything
or
Delay interconnection
we can create a third option:
Connect sooner—but operate within enforceable grid constraints.
That requires an architecture capable of coordinating:
Operating envelopes + DERMS + BESS + flexible compute + demand response + onsite generation + real-time grid conditions.
The key word is enforceable.
A data center saying it could reduce demand doesn’t justify avoiding a transmission or substation upgrade.
A data center contractually committed to a 350-MW operating envelope, with telemetry and automated controls capable of maintaining that limit, presents a fundamentally different planning proposition.
This is where energy policy, software and infrastructure economics are beginning to converge.
The House is asking:
Who should pay for AI’s electricity infrastructure?
The technology industry is asking:
How can AI become more flexible?
Utilities should connect those questions.
Because the cheapest way to protect ratepayers may not simply be making data centers pay for every new substation and transmission line.
It may be giving them a financial incentive not to require those assets in the first place.
That is how flexibility becomes infrastructure.
Relevant reading: AP — House passes data-center energy-cost legislation, September 16, 2026 | Reuters — Silicon Valley’s AI data-center boom meets local resistance, September 17, 2026 | Axios — Tech companies launch flexible-power coalition for data centers
#Energy #EnergyAffordability #ArtificialIntelligence #DataCenters #GridModernization #FlexibleInterconnection #DERMS #OperatingEnvelopes #DemandResponse #BatteryStorage #GridFlexibility #Utilities #EnergyPolicy #ElectricGrid #EnergyInfrastructure
Photo courtesy Ricardo Gomez Angel