Back to Blog

The $3 Billion Lesson From Spartanburg: How the Next Phase of AI Infrastructure Gets Built

photo by Jenny Gregg

TL;DR FAQ: What Does Spartanburg’s Data Center Decision Mean for AI Infrastructure and Hiring?

▼ Q: Did Spartanburg reject data center development?

A: No. Spartanburg did not reject data center development itself. The community pushed back on the specific terms of a proposed $3B project, including power usage, tax structure, and infrastructure cost allocation. The decision reflects deal structure concerns, not opposition to AI infrastructure.

▼ Q: Why are data center projects facing more scrutiny from communities?

A: Data center projects now operate at a scale where energy use, grid impact, and long-term tax agreements directly affect local residents. Communities are asking for transparent terms, fair infrastructure cost allocation, and clear economic benefits before approving large-scale developments.

▼ Q: Are data centers still a strong economic opportunity for communities?

A: Yes. Well-structured data center projects can generate significant tax revenue, fund public infrastructure, and support long-term economic growth. Regions like Loudoun County, VA demonstrate how properly aligned projects can deliver sustained financial benefits.

▼ Q: Is all AI infrastructure being built as massive hyperscale campuses?

A: No. While hyperscale campuses remain important, a growing share of AI infrastructure demand is shifting toward regional data centers and colocation facilities. Enterprises are increasingly deploying AI workloads in distributed environments closer to their operations and data.

▼ Q: Why are companies exploring private or self-hosted AI (LLMs)?

A: As AI usage scales, many organizations are evaluating private infrastructure due to cost, data control, and compliance requirements. Self-hosted or colocated LLMs allow companies to keep sensitive data in-house, reduce long-term inference costs, and maintain greater operational control.

▼ Q: How does AI infrastructure growth impact STEM hiring?

A: Data center expansion is driving demand for specialized technical roles, including electrical and mechanical technicians, network engineers, controls specialists, and critical facilities operators. These roles are high-skill, difficult to fill, and increasingly important as infrastructure expands into more regions.

▼ Q: What is the biggest takeaway for data center developers and operators?

A: The success of future projects will depend on deal structure as much as capital investment. Developers that align early with communities, fund infrastructure responsibly, and communicate clear long-term benefits are far more likely to get projects approved and built.

The United States is entering the largest digital infrastructure expansion in its history.

Artificial intelligence, cloud computing, and enterprise data workloads are driving unprecedented demand for data centers, power systems, and fiber networks. Hundreds of billions of dollars in private capital are moving into the physical infrastructure that will support the next generation of the digital economy.

For communities across the country, this represents one of the most significant economic development opportunities in decades.

A single data center campus can generate tax revenue comparable to major commercial districts while requiring far less public infrastructure. The regions that have successfully partnered with the industry are funding schools, expanding transportation systems, and investing in workforce development in ways that would have been difficult otherwise.

That’s why it’s important to be clear about what happened earlier this year in Spartanburg, South Carolina.

In February 2026, a developer withdrew a proposed $3 billion AI data center project, citing a lack of community alignment.

But the outcome wasn’t a rejection of data center development.

It was a reaction to the specific structure of that proposal.

And the lesson from Spartanburg isn’t that communities are turning against digital infrastructure. It’s that the structure of these projects matters enormously for long-term success.

Why the Spartanburg Deal Fell Apart

On paper, the project appeared to be a generational economic development opportunity.

The proposed campus represented billions in private investment and would have positioned the Greenville–Spartanburg region as a major player in the AI infrastructure economy.

But the details raised legitimate concerns.

The facility was expected to draw around 2 gigawatts of power, making it one of the largest proposed data center loads in the country. At the same time, the proposal included a long-term tax reduction structure and raised questions about how grid infrastructure upgrades would be funded.

When residents evaluated the proposal, the concern was straightforward: the community could potentially bear infrastructure costs while the developer received long-term tax advantages.

Those kinds of concerns don’t mean a community is anti-technology.

They mean the terms of the deal didn’t align.

That distinction matters. Spartanburg is part of one of the most dynamic advanced manufacturing regions in the United States, and there is every reason to believe the region could still attract significant digital infrastructure investment under the right structure.

The lesson isn’t that projects shouldn’t be built.

The lesson is that deal structure determines whether projects succeed.

Communities Are Competing for Data Centers

Despite occasional headlines about opposition, communities across the United States are actively competing to attract data center investment.

And the places that succeed often follow a similar model.

They:

  • align early with utilities on energy infrastructure
  • communicate transparently with residents
  • structure tax agreements tied to performance
  • ensure that infrastructure upgrades are clearly funded

When those pieces come together, the economic impact can be enormous.

Loudoun County, Virginia, home to the largest concentration of data centers in the world, generates hundreds of millions of dollars annually in tax revenue from the industry. That revenue supports schools, transportation infrastructure, and public services while helping keep local tax rates comparatively low.

Other regions have seen similar results at smaller scales, with data center investment supporting new hospitals, improved broadband connectivity, and expanded workforce training programs.

These successes demonstrate something important: when structured well, data center development can be one of the most powerful economic engines available to a community.

The Demand for AI Infrastructure Is Expanding

The growth of data centers is being driven by something very simple: computing demand.

Every AI model, cloud application, enterprise workload, and digital service ultimately requires physical infrastructure somewhere in the world.

And the demand curve is steepening.

What’s changing now is how that infrastructure is being deployed.

Large hyperscale campuses will continue to dominate the headlines, but a second layer of demand is emerging from enterprises across healthcare, finance, manufacturing, research, and defense.

Many organizations are beginning to evaluate whether some AI workloads should run on privately controlled infrastructure rather than exclusively in public cloud environments.

For the past several years, the default approach for most companies was simple: run AI workloads through public APIs offered by major cloud providers.

But as usage scales, the economics can change.

Organizations that run high-volume AI inference workloads can reach a point where the cost of paying per token, per query, or per API call begins to exceed the cost of operating dedicated infrastructure.

At the same time, data governance requirements are pushing some industries to keep sensitive information closer to home.

That combination is driving growing interest in privately hosted large language models.

Instead of sending proprietary data to a public cloud model, organizations can deploy open-weight models on dedicated hardware in colocation facilities or private data centers.

In this model, the model comes to the data, rather than the data traveling to the model.

For healthcare systems, financial institutions, manufacturers, and research organizations, that approach can simplify compliance with regulatory frameworks while maintaining control over sensitive intellectual property.

Performance has improved rapidly as well. New generations of open-weight models are increasingly capable for enterprise workloads, particularly when used in retrieval-augmented systems that supply relevant context during inference.

The result is a three-layer AI infrastructure model that many organizations are now exploring:

  • edge and on-device inference for lightweight workloads
  • privately hosted models in colocation facilities for sensitive or high-volume workloads
  • hyperscale cloud APIs for frontier models and large training runs

This architecture doesn’t replace hyperscale infrastructure.

But it does expand the role of regional data centers and colocation facilities that support enterprise AI deployments.

The Workforce Behind the Infrastructure

Another reason communities compete for data center development is the workforce impact.

Permanent headcounts inside a facility may be modest, but the roles that support these operations are highly skilled and increasingly difficult to fill.

Facilities rely on professionals such as:

• electrical and mechanical technicians
• critical facilities engineers
• HVAC and cooling specialists
• network engineers
• automation and controls technicians
• commissioning managers
• operations leaders

During construction, a large project can support thousands of workers across electrical, mechanical, and engineering trades.

Over time, these facilities anchor ecosystems of contractors, suppliers, and service providers that support regional employment.

As AI infrastructure expands into more markets, demand for these specialized technical roles will grow alongside it.

For companies operating in the digital infrastructure ecosystem, workforce strategy is becoming just as important as site selection.

Engineering Is Improving the Industry’s Footprint

Many of the concerns communities raise about data centers involve practical issues around energy use, water consumption, and noise.

Fortunately, the technology behind these facilities continues to evolve rapidly.

Modern cooling systems significantly reduce noise and energy consumption compared with older designs. Closed-loop cooling systems help address water usage concerns in water-constrained regions. And phased deployment strategies allow infrastructure to expand gradually rather than placing large demands on utilities immediately.

These innovations are helping modern facilities become more efficient and easier to integrate into growing communities.

The Real Lesson From Spartanburg

The outcome in Spartanburg doesn’t suggest communities are rejecting digital infrastructure.

If anything, it reinforces an important lesson for an industry that will continue building massive amounts of capacity over the next decade.

Data center development works best when it is structured as a long-term partnership between developers, utilities, employers, and communities.

The demand for AI infrastructure is real and accelerating.

The regions that benefit most from this expansion will be the ones that structure projects transparently, align infrastructure planning early, and ensure the long-term economic benefits are clear.

When those pieces come together, data centers become more than buildings filled with servers.

They become part of the foundation that supports the next generation of the digital economy.

And that’s why the next decade will see many more of these facilities built across the United States.

Sources:

Recruiting redefined; built for high-tech,
high-growth teams

Hire Talent
Job Board