Data Center Intelligence - Weekly Roundup (Nov 17-23)

November 26, 2025

The data center industry is rapidly evolving, driven by major tech advancements and strategic business decisions. Companies aren’t just adapting to changes; they’re actively reshaping the market with new hardware partnerships, expansive campuses, and a stronger commitment to clean energy. With the rise of AI and the increasing challenges posed by power demands, the gap will only widen between those who can scale effectively and those who can’t.

For the week of November 17–23, 2025, here are the key developments shaping the future of data centers, plus some insights on how analysts might forecast these trends over the next five years.

1. Industry Trends — Acquisitions, Customer Commitments, Capital Moves

OpenAI and Foxconn Team Up for AI Hardware Manufacturing

OpenAI and Foxconn have announced a partnership to co-design and manufacture AI data center hardware in the U.S., including racks, power systems, cabling, and networking equipment, utilizing Foxconn’s facilities in Wisconsin, Ohio, and Texas. OpenAI will get early access and a purchase option on the hardware, while Foxconn gains insights into future high-performance computing needs, without committing to a rigid volume agreement just yet. Reuters+1

Why it matters:

This partnership reflects a shift in the AI supply chain: instead of merely buying servers, AI platforms are influencing how hardware components are developed. This can enhance the collaboration between model developers and hardware makers, potentially speeding up deployment times for new AI clusters.

How an analyst should forecast it (5-year view):

• View this as the beginning of vertical integration in AI infrastructure. In models, consider:
  • A larger share of AI capital expenditures will flow through a select group of strategic manufacturing partners (like Foxconn and a few others).
  • Profit margins on standard, commodity hardware will shrink, whereas customized AI racks and integrated power/cooling setups will command higher prices.
  • Lead-time risks (transformers, switchgear, busway, racks) will gradually lessen after 2027 as these aligned supply relationships develop further.

AI Infrastructure Spending Fuels Significant Capex and Market Volatility

By late November, analysts predict that hyperscalers will invest about $441 billion in infrastructure for 2025, marking an increase of around 184% from 2023, with AI driving much of this growth. Some companies have seen their stocks climb (like Google with its TPU/Gemini 3 boost), while others, like Oracle, face scrutiny over potential overbuilding risks, funding strategies (debt vs. cash), and long-term returns on AI investments. Investors.com+1

Why it matters:

AI infrastructure spending has transitioned into a significant-capex line item; it’s now the central narrative for hyperscalers. However, public market sentiment is starting to raise questions about how rapidly this spending will lead to sustainable cash flow.

How an analyst should forecast it (5-year view):

• Assume high but stabilizing growth in AI capex: anticipate double-digit annual growth through around 2028, but expect more scrutiny concerning utilization and ROI.
• Develop scenarios where capex intensity peaks in the middle of the decade, then normalizes as infrastructure costs are spread out and software revenues catch up.
• Treat the funding mix (debt vs. cash) as a critical factor in equity valuation and credit standings, particularly for companies that are heavily leveraged or are late entrants in the market.

AI Data Center Billionaires Highlight Emerging Value Pools

A Bloomberg report recently identified 16 new billionaires linked to AI data center ventures, including operators, chip-related suppliers, and specialized infrastructure firms. It underscores the growing wealth concentrated among those at the intersection of GPUs, power, and purpose-built campuses. Bloomberg

Why it matters:

This trend isn’t just about hyperscalers; there’s a rising ecosystem of dedicated infrastructure companies riding the wave of AI data center growth, encompassing operators and component suppliers.

How an analyst should forecast it (5-year view):

• In your coverage models, differentiate AI-focused infrastructure companies (like campus developers, advanced cooling, interconnects, and chip-adjacent hardware) from traditional REITs and generalists.
• Expect consolidation: stronger players are likely to acquire niche specialists as industry complexity increases (thermal management, power electronics, advanced busway, etc.).
• Monitor revenue mixes: the greater the share from AI/HPC customers, the more cyclical yet higher-growth the profile will be.

2. Future Expansion — Land, Power, and Build Adjustments

CleanArc Breaks Ground on a 900 MW Campus in Virginia

CleanArc Data Centers has begun construction on a 900 MW hyperscale campus in Virginia. It’s set up to be a sustainability-driven site with grid-scale capacity for AI and cloud workloads. The project will be rolled out in phases, but its headline capacity puts it among the biggest data center complexes in the U.S. Morningstar

Why it matters:

Virginia continues to be the benchmark for scale, even as other areas develop. A 900 MW design signals that large campuses are becoming standard for AI and cloud operations, not just exceptions.

How an analyst should forecast it (5-year view):

• Assume Virginia remains a Tier-1 market but will face increasing grid and policy limitations, so the ramp-up schedule is crucial, not just the total capacity.
• In megawatt and revenue projections, develop phased adoption curves (like 100–150 MW increments) instead of assuming the entire 900 MW will come online rapidly.
• Use this as a model: other regions that can secure similar grid capacity could see mega-campus announcements, particularly in the Midwest and Southwest.

West Texas and New Mexico: Scaling AI Power and Land

New Era Energy Digital has expanded its planned AI/HPC campus in Ector County, Texas, to cover 438 acres, aiming for more than 1 GW of data center capacity. At the same time, it has secured a 3,500-acre option in Lea County, New Mexico, related to over 2 GW of gas generation and a proposed 5+ GW nuclear facility, specifically targeting AI and HPC workloads. Midland Reporter-Telegram

Why it matters:

This reflects the emerging “energy state” model: develop data centers in locations where you can bundle substantial land tracts, gas, and future nuclear setups into a single AI platform, and then transmit data over long-distance fiber rather than chasing after users.

How an analyst should forecast it (5-year view):

• Treat these as long-term but high-potential campuses, significant capital expenditure, but phased commissioning expected after 2026.
• Create scenarios where these sites evolve into primary AI training centers, enjoying lower power costs compared to coastal Tier-1 markets.
• Factor in higher initial development risks (permitting, nuclear timelines, infrastructure build-out), but also potential for better long-term margins if power is secured under favorable conditions.

Not Every Project Survives: Great Falls, Montana Campus Canceled

The proposed $1–1.5 billion hyperscale campus in Great Falls, Montana; a 2 million sq ft development over 569 acres, was officially shelved this week by Ardent TAC Data Centers. Local issues, economic factors, or shifting priorities seem to have led to its cancellation. The Electric -

Why it matters:

We are now seeing a divergence between announced projects and actual developments. Community resistance, grid limitations, or economic factors might lead to some sites being abandoned, particularly in edge or frontier markets.

How an analyst should forecast it (5-year view):

• Include a reduction factor for announced pipeline megawatts, not every press release translates into a functioning campus.
• Monitor markets with repeated cancellations or stalled sites as higher-risk areas for future new projects.
• Recognize operators with proven execution (land + permits + power + community backing) rather than simply large announced plans.

3. Green Energy — Power, Renewables, and Environmental Infrastructure

Google’s 1.5 TWh Ohio Solar PPA Sets a Precedent

TotalEnergies and Google recently established a 15-year PPA for 1.5 TWh of certified renewable energy sourced from the Montpelier solar farm in Ohio, which will support Google’s data center operations on the PJM grid. This week’s industry discussions framed this as a model for how hyperscalers can secure long-term clean energy in constrained grids. TotalEnergies.com+1

Why it matters:

PPAs of this scale show that data centers are now becoming anchor customers for renewable energy, not just purchasers. The structure and duration of these agreements can significantly influence grid planning and future capacity expansions.

How an analyst should forecast it (5-year view):

• Consider that large PPAs will continue to be the primary decarbonization strategy for hyperscalers, but gradually transition from straightforward “bulk MWh” agreements to more 24/7-aligned products.
• In your models, connect renewable PPAs with both ESG narrative value and potential power-cost stabilization against fluctuating spot markets.
• Keep an eye on regions like PJM where these agreements might affect nodal pricing and congestion, impacting both data center economics and local utility operations.

Debate Grows Around Local Solar and Community Benefits

In Virginia, there’s a growing discussion on whether AI data centers should invest more significantly in local solar initiatives and SRECs, tying their growth to community energy benefits, reduced peak demand, and lowered tax burdens. Virginia Mercury

Why it matters:

The conversation is transitioning from “Are you 100% renewable on paper?” to “How are you supporting the local grid and community where you operate?” That’s a different standard — and one that regulators may increasingly enforce.

How an analyst should forecast it (5-year view):

• Anticipate more location-specific requirements: community solar, local storage, or grid-service obligations will be integrated into project approvals.
• Model additional capital or operational expenditures for local energy and community initiatives as part of the total cost of business in sensitive markets.
• Distinguish between markets with light vs. heavy local requirements — this will impact their attractiveness for new AI campuses.

4. Government Policies — Regulations & Public-Sector Influence

States Adjust Data Center Incentives and Regulations

An NCSL policy snapshot published on November 17 reveals a trend in 2025 state legislation reassessing data center incentives, such as Louisiana’s recent allowance for data centers to be categorized as “industrial” for certain local partnerships while still imposing significant minimum capital expenditure and capping tax incentives at 20–30 years. NCSL

Meanwhile, a report by Good Jobs First shows that most states providing data center subsidies don’t detail which companies benefit, despite at least 36 states now offering targeted incentives. Stateline+1

Why it matters:

The political atmosphere is shifting from “whatever it takes to attract data centers” to “incentives with conditions and closer scrutiny.” Look for more stipulations about wages, renewable energy use, and community impact.

How an analyst should forecast it (5-year view):

• Treat incentives as conditional, not guarantees, expect stricter eligibility requirements and compliance reports over time.
• Develop scenarios where effective tax rates increase or certain abatements expire sooner than anticipated.
• Monitor transparency and disclosure trends: investors will likely favor markets and operators that provide clearer disclosures and lower reputational risks.

Alabama’s ‘Project Marvel’ Highlights Community Concerns

In Bessemer, Alabama, city officials have rezoned 100 acres of forest for a colossal hyperscale project nicknamed “Project Marvel,” which one of the largest proposed campuses, forecasted to require 300–500 diesel generators for backup power. Environmental groups are already raising alarms about air pollution, noise, and impacts on local health. Southern Environmental Law Center+1

Why it matters:

This case illustrates the emergence of a new opposition narrative: it’s not only about power and water anymore; it’s about numerous diesel units and their effect on local air quality. Expect tighter regulations on backup power and emissions for future projects.

How an analyst should forecast it (5-year view):

• Anticipate higher compliance and mitigation costs for backup power systems (e.g., diesel controls, fuel switching, or prioritizing battery use).
• Extend entitlement timelines and increase the likelihood of project cancellations in regions with strong environmental activism.
• Keep tabs on the shift from diesel-heavy to battery-enhanced or gas-hybrid backup designs, as these will impact both costs and competitive positioning.

Wisconsin and Other States Implement “Guardrails” for Green Incentives

Wisconsin is proposing conditions for its data center tax exemption, requiring qualifying sites to use 70% renewable energy and meet wage standards, as part of a broader trend to ensure that ratepayers aren’t shouldering the cost of AI expansion without receiving benefits. WPR+2News From The States+2

Why it matters:

States still want jobs and investments, but they’re increasingly clear about who bears the cost of new grid capacity and how green the energy load must be.

How an analyst should forecast it (5-year view):

• Factor in minimum renewable energy standards and wage expectations as standard conditions for approvals in many U.S. states by 2030.
• Expect that some power-abundant but policy-stringent markets will favor hyperscalers and top-tier operators who can meet these requirements, pushing out those with limited capital.
• Assess the durability of incentives: policies focused on renewables and labor tend to be politically sustainable and may prove more stable long-term.

As we navigate this constantly shifting landscape, it’s clear that the future of data centers is being shaped by a blend of technology, thoughtful development, and closer alignment with energy and policy realities. The industry isn’t just reacting to change; it’s actively advancing with new hardware collaborations, expansive campuses in energy-rich regions, and increasingly sophisticated renewable strategies.

For analysts, the next five years will be more than just tracking megawatts, it will involve understanding where those megawatts are located, how they’re powered, and what policy and community considerations projects must meet. Analysts who can piece together the connections between AI demand, energy grid challenges, and changing incentives will have a much clearer picture of who comes out on top in the AI data center race.

“Content is based on public information and personal analysis. Not financial or investment advice.”