Utilities are being asked to grow like technology companies. Most of their trading desks still run on systems built for a much smaller, slower world.
AI Data Center Demand Is Outpacing Interconnection Queues
AI data center demand is forcing multi-gigawatt load growth into timelines that used to take decades. It’s happening from Texas to the Netherlands, wherever hyperscalers are chasing power and grid capacity to build the next generation of data centers. The numbers are extraordinary.
The Electric Reliability Council of Texas (ERCOT), the grid operator for most of the state, was tracking approximately 474 GW of large-load interconnection requests as of June 20261, with data centers accounting for more than 90% of that total, up from roughly 63 GW just 18 months earlier2. PJM, the largest grid operator in the United States, received applications representing roughly 220 GW of proposed projects for its next interconnection cycle alone3, and projects there now spend an average of more than seven years from application to operation. The capacity waiting in interconnection queues exceeds the entire installed generating capacity of the United States power grid4.
To meet it, generators and retailers operating in deregulated and liberalized markets are signing more power purchase agreements, structuring non-standard deals like hourly-matched clean energy contracts, and negotiating directly with hyperscalers who bring credit and settlement requirements most trading desks have never handled at this scale.
Deregulated utilities are feeling this first, and they’re also the ones best placed to act on it. Unlike rate-regulated utilities, where any significant technology investment can mean months of regulatory filings and approval cycles, deregulated generators and retailers can usually get sign-off from their own board once the case is clear. That’s a real advantage right now. It also means there’s less excuse to wait.
Regulatory and Ratepayer Pushback: Texas SB6 and Beyond
The scale of interconnection requests isn’t just a planning challenge, it has become a political one. Residential ratepayers and state legislators are asking hard questions about who pays when a data center developer needs a new substation, a transmission upgrade, or a grid expansion that benefits primarily one customer and lasts fifty years.
Texas has become the most visible laboratory for working this out. In June 2025, the Texas legislature enacted Senate Bill 6, a package of reforms that formalized ERCOT’s large-load interconnection process, required data centers to post $50,000 per MW in financial security before entering the queue5, and directed the Public Utility Commission of Texas to make data centers responsible for the infrastructure costs they create. The core philosophy, as one observer put it, is simple: data centers should pay their own way, and stand down during grid emergencies so that residential ratepayers don’t absorb the cost6. Governor Greg Abbott escalated further in June 2026, directing ERCOT to audit all data center projects in the interconnection queue before approving any new connections7, and ordering that no grid upgrade costs for data center service be passed on to residential customers.
Texas is not alone. In PJM and the other regional markets, pressure has been building on the Federal Energy Regulatory Commission (FERC), which regulates interstate electricity transmission in the United States, to establish a consistent federal framework for large load interconnection8. Without one, utilities are managing these requests inconsistently, often with the same outdated, slow processes that have failed generation interconnection for years.
The regulatory environment around data center load is moving fast and in one clear direction: more scrutiny, more cost responsibility on developers, and more pressure on utilities to demonstrate that their systems and processes can handle what’s coming without passing the bill to consumers.
For utilities navigating this environment, the operational question and the political question are the same: can your trading infrastructure handle the scale and complexity of what’s in your pipeline, and can you demonstrate to regulators and ratepayers that you’re managing it efficiently?
Where Legacy ETRM Systems Break Under New Deal Complexity
When deal volume and product complexity increase exponentially, the cracks show up fast: a trader building a deal in a spreadsheet because no system can configure it, a settlement team reconciling by hand because two tools won’t talk to each other, a risk report that takes days to assemble because the data lives in five different places. None of this shows up as an outage. It shows up as a slow accumulation of manual work that quietly becomes the process, until it’s the bottleneck standing between the business and the growth its board is expecting.
The data center load surge adds a specific new wrinkle. A hyperscaler requesting several hundred megawatts, or in some cases, a gigawatt-scale campus, brings deal structures, credit requirements, and settlement complexity that many legacy energy trading and risk management ‘ETRM’ systems were never designed to handle. Hourly-matched clean energy contracts, co-location arrangements, curtailment obligations under new regulatory frameworks like SB6, and bilateral offtake agreements with non-utility counterparties all require purpose-built configuration. When the system can’t handle the deal type, the workaround becomes the process.
Why AI Tools Fail on Fragmented Trading Data
Here is where the story takes an unexpected turn. The technology driving the power demand is also being positioned as the solution to managing it. Utilities across the industry are turning to AI. AI-powered reporting, natural language queries, intelligent scheduling tools, automated settlement; all to help their lean trading and operations teams cope with a workload that is growing faster than headcount ever will.
The pitch is compelling, and in the right environment, it works. But there is a problem that is quietly derailing AI initiatives across the industry before they ever produce a meaningful result: the data underneath most trading operations is a mess.
AI does not fix bad data. It amplifies it. A natural language query tool built on top of fragmented, inconsistently structured, manually maintained trading data does not give you faster answers. It gives you faster wrong answers, with enough confidence in the output that nobody questions them until something breaks. The same applies to AI-powered risk reports, position summaries, and settlement reconciliation tools. The intelligence is only as good as the foundation it sits on.
There is almost too neat. The AI boom depends on data centers drawing unprecedented amounts of power. Utilities are being asked to manage that demand with more sophistication than ever before. They are turning to AI to help. And they are discovering that years of working around inadequate trading systems has left them with exactly the kind of fragmented, poorly organized, multi-source data that makes AI tools unreliable at best and dangerous at worst.
The demand that is straining the grid is the same force exposing the data problems that were always there. It just took this much pressure to surface them.
The path forward is not to abandon AI, it is to do the work in the right order. Clean, well-structured, integrated trading data is the foundation. It is not the exciting part of the conversation, but it is the part that determines whether the AI investment pays off or quietly becomes another layer of complexity on top of a system that was already struggling.
What an AI Enablement Assessment Covers
The mistake is assuming the fix is point solutions – add a feature, update a calculation, buy a best of breed product – without first understanding the larger impact of what’s in place today. Some utilities need a new platform. Others need to consolidate several legacy tools into one. Others have a perfectly capable system that’s just badly configured or barely integrated. You can’t know which one you are without looking at the whole picture first: every spreadsheet, every legacy platform, every manual handoff involved in getting a deal from execution to settlement.
That’s what a AI Enablement assessment is for. It’s a vendor-neutral look at everything currently supporting your trading operation, spreadsheets and all, to find out where the real gaps are before they become the reason growth stalls.
In practice, that means looking at:
Your full trading technology landscape.
Every spreadsheet, legacy platform, and manual handoff involved in getting a deal from execution to settlement, not just the named ETRM/CTRM system if one exists.
Capacity and scalability.
Whether the current setup can handle the deal volume and transaction throughput your growth plans actually require, including the non-standard deal structures that hyperscale demand is driving.
Product and deal structure coverage.
Whether the platform can configure the deal types data center demand is creating, like hourly-matched contracts, complex tolling structures, and curtailment-obligated offtake agreements, without manual workarounds.
Data architecture and AI readiness.
Whether the underlying data is structured, consistent, and integrated well enough to support the AI and automation tools your organization is planning to deploy, because if it isn’t, that is the first problem to solve.
Integration and onboarding friction.
How cleanly new counterparties, market data feeds, and downstream systems like credit and settlement can be onboarded as the business scales.
Total cost of ownership and upgrade path.
The real cost of staying, upgrading, consolidating, or replacing, mapped against your timeline and budget cycle.
A vendor-neutral recommendation.
A clear answer on whether the fix is configuration, consolidation, or a new platform, without a predetermined push toward any specific vendor or product.
Sometimes the answer is a configuration fix. Sometimes it’s consolidation. Sometimes it’s a new platform. The point is knowing which one, on your timeline, before the patchwork breaks under the weight of what’s coming.
The interconnection queue numbers make the direction of travel clear. The regulatory response in Texas and elsewhere makes the stakes clear. If your trading operation is being asked to scale for AI-driven demand and you’re not sure what’s actually holding it together right now, that’s worth finding out before the next deal lands on your desk that your current system can’t configure.
Get Started with capSpire
capSpire consultants have spent decades inside energy trading operations, assessing what trading technology can actually support and then selecting, implementing, and optimizing the ETRM/CTRM systems that close the gap. We are vendor neutral, which means the recommendation you get is the one your operation needs. Contact us to find out where your technology stands today.
John Harris
Client Account Manager
1ERCOT large-load interconnection requests (474 GW as of June 2026, 90%+ data centers):
https://www.powermag.com/abbott-orders-full-audit-of-texas-data-center-interconnection-queue-threatens-to-deny-grid-access/
2ERCOT large-load requests up from 63 GW in December 2024 (226 GW as of November 2025):
https://etalytics.com/resources/blog/when-data-centers-request-gigawatts
3PJM 220 GW proposed projects, seven-year average from application to operation:
https://www.datacenterknowledge.com/energy-power-supply/why-ai-data-center-projects-face-years-of-delays-after-approval
4U.S. interconnection queue capacity exceeding total installed generating capacity:
https://www.hanwhadatacenters.com/blog/data-center-grid-limitations-the-power-bottleneck
5Texas Senate Bill 6 — large-load interconnection reforms, $50,000/MW financial security requirement:
https://www.gtlaw.com/en/insights/2026/3/texas-senate-bill-6-update-what-data-centers-large-load-customers-should-know-about-proposed-interconnection-standards
6Texas Senate Bill 6 — ratepayer protection philosophy and implementation:
https://www.latitudemedia.com/news/the-multimillion-dollar-debate-over-powering-data-centers-in-texas/
7Governor Abbott June 2026 directives — audit of data center queue, ratepayer cost protections:
https://www.troutman.com/insights/texas-hits-pause-on-data-center-grid-connections-amid-growing-oversight-push/
8FERC and federal framework for large-load interconnection:
https://ifp.org/interconnection-for-ai/
9BofA data center load forecast — 125 GW of U.S. electric load addition, 4.1% CAGR through 2030:
https://www.utilitydive.com/news/ai-data-center-growth-utilities-generation-plans/825541/
10Texas and Virginia as regulatory models for other states:
https://www.belfercenter.org/research-analysis/ai-data-centers-us-electric-grid