Written by Alex Rosenthal, Partner and Head of Data Center Group
This year’s Datacloud USA conference in Austin made clear that the data center industry has reached maturity and is now entering an exciting and challenging new phase. After the sector’s initial market splash, participants in this new ecosystem now face the hard part: extracting power and value from this promising resource.
The conference was a full-blown display of American ingenuity. The room included manufacturers of nearly every component that goes into a data center, developers, engineers, energy providers, investors, software companies, neocloud executives, and entrepreneurs attacking problems that barely existed a few years ago. Everyone was looking for opportunities, but they were also trading information, forming partnerships, and trying to solve a common set of problems at enormous speed.
The clearest takeaway for me and my partner Allison Gabala, who leads our firm’s Environmental Law practice, was that the market has reached a more demanding stage. Capital and customer demand remain abundant, but execution has become the harder problem. A site can be controlled, zoned, and ready to build, yet still fail if power does not arrive when promised. A project can have a credible power plan but stall under a moratorium, a zoning change, community opposition, or water constraints. A lease can promise flexibility to a customer while the development, utility, and financing documents make that flexibility impossible to deliver.
For developers, operators, investors, and their counsel, these issues cannot be handled in separate lanes. They must be handled together in real time to assess the risks that exist in the intersection of the parts.
1. Power availability is driving the entire schedule
Getting the land, approvals, and agreements in place is only the first challenge. The next hurdle is whether the power will be there on Day One— in the required quantity and on workable economic terms.
The legal work begins long before lease drafting and reaches far beyond confirming that a utility letter exists. Before that point, the project team must determine:
FERC’s June action, directing all six regional grid operators under its jurisdiction to justify or reform their tariffs governing the integration of data centers and other large loads, is a good measure of how quickly large-load regulation is evolving.
Industry insiders are examining options such as behind-the-meter and on-site generation to provide bridge, phased or long-term power, but these ideas also raise questions about generation ownership, fuel supply, interconnection, emissions and other environmental permits, noise, self-use rules, tenant consumption, retail power sales, utility regulation, and decommissioning. The answers vary by jurisdiction.
Connectivity raises similar issues. Fiber routes, redundant entry points, easements, rights-of-way and carrier access need to be evaluated alongside power. A powered site without resilient connectivity is still not a viable data center.
Additionally, utility arrangements, site-control documents, development schedules, lease commencement provisions, and financing assumptions must be analyzed I depth to ensure a project will be successful, on time, and on budget.
2. Diligence has to account for a moving regulatory target
Industry participants must also face the possibility that a project can be derailed by a moratorium or a change in local or state rules. Confirming that a use is permitted under today’s zoning map is only the first question.
Diligence must examine:
Local elections and changing incentive regimes can matter as much as the existing ordinance.
Site-control and acquisition documents should reflect that uncertainty through realistic diligence periods, extension rights, milestone conditions, outside dates, and exit rights. Otherwise, a developer can end up owning an entitled site that cannot be energized or operated as modeled.
3. Community support has become part of project delivery
Data centers have historically done a poor job with public relations. The industry’s relative anonymity once had practical and security advantages. At today’s scale, that approach no longer works.
The cloud is not in the sky. Phones, applications, remote work, data storage, streaming, artificial intelligence, and much of daily commerce depend on physical data centers. The industry has a strong story to tell about the infrastructure it provides and the economic activity it supports.
The industry’s narrative needs to address actual community concerns. Residents want to understand water consumption, utility costs, noise, air quality, appearance, construction impacts, tax benefits, and the number and quality of permanent jobs. Writing every objection off as NIMBYism will not build trust. Credibility comes from listening early, responding with project-specific information and, where appropriate, changing the design.
Community strategy should inform the site plan from the beginning. Buffers, sound attenuation, generator placement, construction hours, landscaping, reclaimed or non-potable water, and closed-loop, liquid or hybrid cooling can all affect both public acceptance and the approvals process. The first public hearing is too late to begin that work.
The same flexibility will matter as the market moves from its early emphasis on model training toward much broader inference demand. Training frontier models typically concentrate enormous compute in massive clusters. Inference occurs whenever a model is used, and agentic systems can turn a single user request into many model calls. As those systems move into production, power and cooling plans will increasingly need to follow actual usage and allow capacity to grow in smaller increments. The market is likely to require a mix of giant training campuses, centralized inference clusters, and regional or modular capacity closer to users and data where latency, privacy, or data gravity matter.
Robotics adds another layer. A robot may need to make immediate decisions on-device or at the edge, while the simulation, model training, fleet learning, and data processing behind it will continue to rely on substantial centralized compute. That points toward a more interconnected architecture spanning large campuses, regional inference facilities and edge systems. Location, fiber diversity, latency, phased power delivery, and the ability to expand or shift capacity will matter even more, and those requirements should shape site selection, interconnection strategy, and contract documents from the beginning.
4. Optionality has to be engineered across the entire deal
Neoclouds, the AI-focused cloud providers offering GPU capacity on demand, need to scale quickly while workloads, customer commitments and hardware cycles continue to change. That creates demand for power and space ramp rights, phased delivery, expansion and contraction rights, partial termination options, renewals, rights of first offer or refusal, and access to the last available capacity at a campus.
These rights are all highly interdependent. An expansion right has little value if the utility allocation is unavailable or the adjacent land has been committed. A contraction right may conflict with the owner’s financing or a long-term take-or-pay power obligation. A promised delivery phase may depend on permits, equipment, and transmission upgrades controlled by third parties.
The lease, colocation or services agreement, development documents, utility arrangements and financing documents have to be mapped against one another. For neoclouds in particular, the mismatch between long-lived infrastructure obligations and faster-moving customer and chip cycles makes that coordination essential. Otherwise, flexibility exists only on paper.
5. Neocloud consolidation will reach down to the site documents
In a strong indication that this market is beginning to mature, there was substantial discussion in Austin about imminent consolidation among neoclouds in light of increasing capital intensity, customer concentration and the value of scale.
The effects of consolidation will not stop at the corporate level. Assignments and change-of-control provisions, guaranties and other credit support, capacity reservations, exclusivity rights, equipment liens, landlord waivers, financing consents, licenses, permits and customer portability may all determine how easily a platform can be combined or sold.
A roll-up that makes perfect strategic sense can become operationally painful if dozens of site-level agreements are not properly assignable. Parties negotiating these agreements today should assume that ownership, financing, and operating structures may change during the life of the project.
6. Mega-campuses and modular development will grow together
The industry is moving from facilities in the 50-to-100-megawatt range toward multi-building campuses that exceed a gigawatt. At the same time, modular and containerized data centers offer a useful counterweight. They allow power and compute to be added in increments, potentially making deployment smaller, faster, and more flexible.
The legal consequences are meaningful. The parties still need to address whether the equipment is a fixture or personal property; and address security interests and lien priority; building and fire code treatment; zoning and permitting status; tax treatment; access and transportation; incremental interconnection; phasing; and removal and restoration. Modular equipment does not automatically create a modular approvals process. Permits and project documents must also accommodate phased deployment.
7. The next bottleneck may be people
Concerns remain about the staffing required to manage full pipelines, rising demand and supply constraints. The need is not limited to construction labor. The industry needs more electricians, mechanical trades, designers, cooling and controls specialists, fiber professionals, commissioning teams, operators, and experienced personnel at utilities and reviewing agencies.
Workforce limitations eventually show up in the contracts. Construction milestones, labor and supply-chain force majeure provisions, liquidated damages, contractor capacity, commissioning requirements, contingency planning, and service-level commitments should be based on the labor market that exists. A schedule that ignores workforce capacity merely converts an operating problem into a future dispute.
8. The legal work is increasingly happening at the seams
A data center project may once have looked like a real estate transaction with several specialized components added to it. Today, the asset is a coordinated system of land, power, connectivity, water and cooling, permits, community support, construction and contracts. Lacking discipline in any one of those areas can strand value in all the others.
At AY Strauss, our data center group brings together real estate, zoning and land use, environmental, power and interconnection, connectivity, regulatory and transactional experience at the outset. The most significant legal work in this sector often happens between the traditional specialties, where a promise made in one document creates risk somewhere else.
The energy in Austin was unmistakable, and so was the scale of the opportunity. The companies that lead the next phase will be the ones that can turn ambition into delivered megawatts, durable approvals, and operating capacity on schedule. I’m energized to continue working with our clients to help them achieve their goals in this exciting space.
For questions, contact Alex Rosenthal at 973-376-3840 or arosenthal@aystrauss.com or Allison Gabala at 973-952-6866 or agabala@aystrauss.com.