The Grid Is Not Ready for Your Proposed Data Center

AI needs more data centers, but investors need projects that can actually obtain power.

By Paul Anthony Claxton, General Partner, Digerati Investments | Managing Partner, Reciprocity ROI, LLC - Wednesday September 23rd, 2026 - 8 minute read

Subscribe on Substack | Follow on LinkedIn​

Let’s scenario play for a moment. You play data center, I play investor.

Someone shows me a parcel near a transmission line, a rendering of a data center, and a slide claiming hundreds of megawatts of capacity. The presentation moves quickly from the size of the AI market to the amount of money the sponsor wants. The utility commitment, the cost of upgrades, and the date at which the proposed capacity can actually be delivered tend to move more slowly and sometimes they do not appear at all.

There is a real shortage of data center capacity in the places where buyers want it, so data center power availability to meet demand is a real issue, but there is no shortage of proposed projects claiming they will meet that demand. Those are very different things, even when the renderings and megawatt figures look equally impressive in PowerPoint. A parcel near a transmission line is not available capacity. Neither is a financial model showing revenue from a facility that has yet to secure power or be built. Until the developer can document a viable path to delivering the facility, the capacity remains hypothetical. And even a believable project does not become available capacity until it can actually serve customers.

We need more data centers, obviously. We also need to stop pretending that every development announcement represents a financeable project, or that any facility with an AI customer is automatically a responsible addition to the grid. The investment question is whether the sponsor can secure power on workable terms, build without imposing unexamined costs on others, and convert that capacity into durable customer revenue.

AI demand cannot approve your interconnection

The International Energy Agency reported that data center electricity consumption rose 17% in 2025 and projects it to roughly double between 2025 and 2030. A Lawrence Berkeley National Laboratory report estimates that data centers could account for 11.8% of U.S. electricity use by 2030 in its reference case.

Those figures support an infrastructure opportunity, but they do not establish that a specific parcel can receive power, that its developer can afford the interconnection, or that anyone has agreed to buy the capacity. Quoting global AI demand while leaving project-level evidence unresolved is a fantastically efficient way to avoid underwriting the actual project.

The IEA estimates that grid constraints could delay around 20% of global data center capacity planned for construction by 2030. That is a risk estimate for planned capacity, not a finding that 20% of developers are dishonest or that no new data centers can be built. It does mean the grid cannot accommodate every proposal on the sponsor's preferred schedule just because the sponsor has found willing investors.

So this is the awakening that the market needs wherein demand cannot distract reality. Everyone wants to build a data center, but right now the grid cannot handle it. Without a real, tangible and documented path to reliable power, you do not have a data center project.

Some regions can support new capacity sooner than others, and grid upgrades, new generation, storage, or flexible operations can change the answer. That is exactly why a sponsor must demonstrate the answer for its location and timeline. The claim “there is a power line nearby” tells me almost nothing about available capacity, transmission constraints, upgrade costs, or the right to interconnect.

All That Money, No Data Center Power Availability.

Powered land requires more than money folks. When a sponsor advertises a site as powered, I want to know what that means in documents.

1.     Is there existing service at the required load?

2.     A binding utility agreement?

3.     A completed facilities study?

4.     An approved interconnection with outstanding upgrades?

5.     Or just an application and a plausible-sounding date?

Those are all different stages of development and deserve different prices and financing structures.

Show me when the power can be delivered, what upgrades are required, and who will pay for them. If the only evidence is an application or a letter of intent, the investor is financing the pursuit of power, but the land is not powered yet. But getting power to the land is only part of the issue, because a large facility can strain the local grid and require new equipment or power generation, and generating power on site does not make them disappear.. Federal regulators are examining these questions.

Before funding a project, I want clear answers:

1.     How much power will it use, and when?

2.     Can the grid supply it, or are upgrades needed?

3.     Who pays for those upgrades?

4.     Can the facility use less power when the grid is under strain without failing its customers?

5.     If it plans to generate or store power on site or “behind the meter”, are the equipment, fuel, permits, costs, and emissions accounted for?

If the project depends on a power plant that has not been built, investors need to assess that project too. Building where power is available and reducing demand when possible can help, as the IEA notes, but neither one replaces proof that the data center can reliably power what it promises to deliver.

Powerful Destruction

A data center can bring investment to a region while leaving nearby residents with higher electricity bills, greater pressure on water supplies, and pollution from the power built to serve it. That is not an acceptable trade simply because the facility supports AI.

Federal regulators are all over this and are already examining how to keep the costs of grid upgrades for large users from falling on other customers.

Before I invest, I want to know what the project does to the people living around it.

1.     Will households pay more for power?

2.     Will the facility compete for water the community needs?

3.     Will onsite generation or backup equipment worsen local air quality?

4.     Who pays for new grid infrastructure, and what happens to those costs if the data center is delayed or never built?

An ESG slide answers none of that.

Your Data Center Customer Has Customers to Lose

A company may need computing capacity and still hesitate to commit to your facility. If the project draws opposition over electricity costs, water use, or its effect on the surrounding community, that company may have to explain its decision to its own customers. Whether those customers would actually leave depends on the company and the project, but the risk deserves examination before projected demand becomes projected revenue. Data center developers have acknowledged that community opposition can disrupt tenants and make facilities harder to lease.

A site tour, request for a proposal, or letter of intent therefore tells me only that someone is interested. I want to know whether the buyer has assessed the project’s community and reputational risks, what it has committed to purchase, and whether it can delay or cancel if the facility runs into trouble.

I also separate proposed, reserved, contracted, delivered, and billable capacity. For example, a project could propose 100 MW, reserve 70 MW, contract 50 MW, deliver 30 MW, and bill for 20 MW. Those are five different claims, not 100 MW of current revenue. In its filing for the quarter ended June 30, 2026, Core Scientific reported 590 MW of leased customer power capacity and 395 MW it was actively billing. Even with a signed customer, delivering capacity and beginning to bill take time.

Customers can be assets or your greatest liabilities. This means that a major buyer’s commitment can help finance a data center, but if that buyer accounts for most of the planned revenue, the project depends heavily on its decisions. What happens if it delays deployment, cannot pay, or decides that association with the facility could cost it customers of its own? Interest alone offers little protection against those risks. I want an enforceable agreement, capacity the operator can actually deliver and bill for, and a realistic assessment of whether the buyer will pay.

Assets and Liabilities: Turning Depreciation Into Appreciation

A data center should not arrive in a community, demand enormous amounts of power, and leave residents wondering whether their bills will rise. If the project helps fund new local generation, residents should have a way to share in the benefit through lower electricity costs. A utility partnership or community energy program might make that possible, depending on local rules. It takes a real supply plan and a measurable discount, not a promise on an ESG slide.

There may also be an opportunity to use homes as small compute sites. A company that owns GPUs could place suitable equipment with willing residents and run paying customers’ workloads on it. Residents would be paid to host the equipment and could benefit from its heat. Alternatively, residents who want to use the GPUs themselves could lease them and pay for that use. Either arrangement needs a contract, and each home’s electrical connection must be able to handle the added load.

An idle GPU loses value while producing no income. Once it is placed with a resident under a sound agreement, it can begin earning lease payments or revenue from compute customers. As utilization and reliable payments are proven, the combined value of the equipment and its contracts may grow even while the GPU itself continues to depreciate. That stronger cash flow can support financing. It only works if revenue exceeds the costs of power, installation, resident compensation, maintenance, insurance, and replacement.

Furthermore, with an ESG component, a nonprofit partner could help ensure that promised savings reach households, and an eligible clean-energy project might qualify for tax benefits. I invest in advancement: a project should earn money while leaving its neighbors better off.

Finance the stage the project has reached

Early-stage projects can obtain capital, but that does not mean they have real value. Funding studies, securing long-term site rights, engineering an interconnection, obtaining permits, or negotiating customer agreements can create real value. The sponsor should say which milestones the money will buy, when they can be verified, and how much capital is at risk if they fail.

All capital is not the same. If you have a parcel, a rendering, and an application for power, you are asking for inception capital to find out whether the project is viable. Price the risk accordingly. Development capital should follow as the project secures land rights, power, permits, construction commitments, and customers; release it in stages against evidence. Calling an unproved proposal “shovel-ready” and demanding the terms of a financeable facility asks investors to pay you for work you have not done.

Here is the test I would apply before considering a serious commitment:

1.     What land rights exist today?

2.     What electricity has actually been committed, at what price and date?

3.     Who funds the required upgrades, and how will the facility affect other grid users?

4.     What are the expected energy, emissions, and water impacts, and who will verify them?

5.     What remains to be built and approved?

6.     Who has signed a binding contract to pay, and when does billing start?

7.     If a milestone fails, who takes the loss?

Those questions do not need to demand perfection. They demand a description of the project that survives contact with the documents. A developer who answers them clearly may have a financeable development opportunity even if major risks remain. A developer who advertises a powered, shovel-ready, ESG-aligned data center while promising to secure the land, electricity, permits, sustainability plan, and customers after funding has a collection of claims and potential legal trouble if they take the funding. There may be a business in there someday; but not today, and not tomorrow.

Dreams cannot finance reality, and if you want to finance a dream, be my guest, just don’t call me when you finally wake up.

Continue the conversation

For candid analysis of venture capital, artificial intelligence, founder readiness and capital markets, subscribe to VC & AI w/Paul Anthony Claxton on LinkedIn or VC and AI with Paul Claxton on Substack. You can also follow Paul Anthony Claxton on LinkedIn for shorter commentary and updates.

The ideas, investment judgments, and conclusions in this article are entirely mine. I used AI to assist with drafting, editing, and visual development. I reviewed and revised the final work and take responsibility for what I publish.

This article is provided for general informational and educational purposes only. It does not constitute investment, legal, tax or financial advice; an offer to sell or a solicitation of an offer to buy any security or interest in any fund; or a recommendation or endorsement of any company, security or investment strategy.