Data Centers

For much of the past two years, corporate America has treated advanced artificial intelligence as an unlimited competitive race. Companies spent freely on the most powerful models available, often with limited regard for cost but that is now changing. It’s no longer the Wild West when it comes to A.I.

Companies are no longer defaulting to the highest priced models for every task. Instead, they are mixing systems by prioritizing the expensive pay per use models for complex planning while sending more routine work to cheaper alternatives, including “open weight” public models. 

Essentially, A.I. has quickly gone from being categorized as a necessity to now being treated more like a utility expense that needs to be managed. 

You may be wondering how this relates to data centers. Data centers are the physical infrastructure that makes large scale A.I. possible. Their rapid expansion has been driven by expectations of continuously increasing demand for high performance computing and the enormous capital investments that have followed.

A single data center can require enormous investments in water, wastewater, electric infrastructure, and transmission capacity. Those improvements are often built with the expectation that the project will operate for decades and generate enough economic activity to justify the cost. A large share of current A.I. related capacity is tied, either directly or indirectly, to a relatively small number of high spending customers.

If those customers reduce their growth rates and shift toward more efficient models, revenue that was projected to cover debt service may fall short. It’s reasonable to ask whether every projected data center will still be necessary which is why communities should be cautious about making long term infrastructure commitments based solely on optimistic projections.

It’s also important to understand how data centers are financed. It is not uncommon for these projects to be structured through Special Purpose Vehicles (SPVs), which are separate legal entities created specifically for one development, created specifically to own the facility, hold the debt, and sign customer contracts.

While SPVs are not uncommon and they are legitimate financing tools, their intended use it to limit the financial exposure of the parent company. If a project is delayed, downsized, abandoned, or fails, the SPV may have few assets available to cover outstanding obligations. This is why it is critical that we have strong agreements in place, so that the public is not left with infrastructure that was built for a project that never fully materialized.

I acknowledge that some of these decisions may not be entirely within our local control. Arkansas has adopted laws that encourage data center development and limits the ability of local governments to prohibit these projects. As a result, cities like Fort Smith need to place our focus on ensuring we have the right safeguards in place. This is why policy matters, with government the policy typically moves slower than the technology, now is the time to be proactive and not reactive.

For Fort Smith specifically, the concern extends beyond our city limits because we are a regional water provider. If we expand treatment capacity, transmission lines, or other water infrastructure to serve large industrial users, those investments impact the entire system. If anticipated demand never materializes, the costs of that infrastructure don’t just disappear and must still be paid, placing additional financial pressure on existing ratepayers and the communities that depend on our water system.

Before investing taxpayer dollars or expanding public utility infrastructure to serve a data center, local governments should have policies in place that ensure the private developer and not the public bears the financial risk if market conditions change or projected demand fails to materialize.

Developers should be responsible for the infrastructure built specifically to serve their projects through enforceable agreements, financial guarantees, performance bonds, letters of credit, or other mechanisms that protect taxpayers and utility ratepayers from financial liability.

Long term infrastructure decisions, such as pipe sizing, treatment plant expansions, and system redundancy, are based on projected demand. If those projections don’t materialize, the community can be left with infrastructure that is underutilized while taxpayers and utility ratepayers continue paying for the investments that were made.

These are not short-term decisions. Water infrastructure is built to last for decades. A transmission line or treatment plant expansion approved today will likely still be serving the region 50 years from now. That’s why we should be cautious about locking in permanent infrastructure commitments based on uncertain market projections or rapidly evolving technology.

Arkansas has made it clear that data centers are part of the state’s economic development strategy. Whether we ultimately host one may not be entirely within our control. What is within our control is ensuring that any project relying on Fort Smith’s regional water system does so on terms that protect existing customers, safeguard taxpayers, and leave the financial risk of speculative development with the developer and not the public.