Data Center Intelligence — Weekly Roundup (Mar. 16–22)
Weekly Data Center Dispatch
Last week showed that the AI infrastructure market is no longer expanding one component at a time.
Data center developers are securing campuses. Technology companies are investing in connectivity. Utilities and energy companies are assembling dedicated power. Cooling providers are becoming acquisition targets. Governments are using infrastructure policy to build domestic AI capacity.
The pieces are beginning to move together.
That creates opportunity, but it also raises the execution standard. A data center cannot scale because one part of the plan works. The land, power, equipment, cooling, financing, customers, and approvals must all arrive in sequence.
Below are the developments that mattered, in plain English, with what they mean for operators, customers, investors, and anyone doing FP&A.
1) Industry momentum
New campuses, cooling acquisitions, supplier investment, and the rising value of infrastructure control
Bell announces a 300-megawatt Saskatchewan campus
Bell Canada and the Government of Saskatchewan announced a 300-megawatt data center campus outside Regina, with CoreWeave and Cerebras identified as initial customers.
What this means: Large AI campuses are expanding beyond the most established North American markets. Saskatchewan offers land, energy resources, cooler temperatures, and government support.
Analogy: When the central warehouse reaches capacity, companies begin building regional hubs closer to available land, labor, and transportation.
Ecolab agrees to acquire CoolIT Systems for $4.75 billion
Ecolab announced plans to buy liquid-cooling provider CoolIT Systems from KKR for approximately $4.75 billion in cash.
CoolIT supplies cooling systems used by hyperscale, colocation, and chip customers, including Nvidia and AMD.
What this means: Cooling is becoming strategic infrastructure rather than a supporting mechanical system. Higher-density AI equipment is increasing the value of companies that can remove heat efficiently and manage water responsibly.
CoolIT’s projected sales highlight the size of the cooling opportunity
Ecolab expects CoolIT to generate approximately $550 million in revenue over the next 12 months.
What this means: AI infrastructure spending is moving through the full supply chain. The opportunity includes not only processors and buildings, but also pumps, coolant distribution, controls, monitoring, and long-term maintenance.
Analogy: As factories become more productive, the systems that prevent them from overheating become more valuable.
MaxLinear launches new AI connectivity technology
MaxLinear introduced its Annapurna 224G retimer, designed to extend high-speed copper connectivity inside AI data centers.
What this means: Larger AI clusters require faster and more reliable communication between processors, memory, and networking equipment. Connectivity can become a bottleneck even when sufficient computing hardware is available.
Germany moves to double its domestic data center capacity
Germany outlined measures intended to at least double national data center capacity and increase domestic AI-processing capability fourfold by 2030.
What this means: Data centers are becoming part of national industrial and security policy. Governments want computing capacity located within their borders and operating under domestic rules.
2) Future expansion
Gigawatt projects, new power capacity, grid flexibility, and the infrastructure required to support growth
SoftBank and AEP announce a massive Ohio AI campus
SoftBank, American Electric Power, and the U.S. Department of Energy announced plans for an AI data center development at the former Portsmouth uranium-enrichment site in Ohio.
The proposed project includes approximately 10 gigawatts of computing capacity supported by as much as 9.2 gigawatts of natural gas generation.
What this means: The scale of AI development is moving beyond traditional data center campuses. Projects of this size resemble regional industrial and energy programs.
Analogy: This is not adding a production line. It is planning an entirely new manufacturing district with its own utility system.
The Ohio project includes major transmission investment
The proposal includes approximately $4.2 billion of transmission upgrades in southern Ohio.
What this means: Even a campus with dedicated generation still depends on broader grid infrastructure for reliability, balancing, backup, and regional integration.
The data center and the power system cannot be modeled as separate investments.
Solaris adds 900 megawatts of generation capacity
Solaris Energy Infrastructure announced plans to add approximately 900 megawatts of incremental power-generation capacity between 2026 and 2029.
The expansion would bring its pro forma generation capacity to roughly 3.1 gigawatts.
What this means: Specialized power providers are scaling to serve customers that cannot wait for conventional utility timelines.
This creates another path to capacity, but also introduces fuel, equipment, financing, and operating risk.
Google expands demand-response agreements with five utilities
Google signed agreements with five U.S. utilities that allow it to reduce data center electricity use during periods of peak demand.
The arrangements cover utilities in markets including Arkansas and Minnesota.
What this means: Data center flexibility can unlock capacity without waiting for every new power plant or transmission line to be completed.
Analogy: A factory that can temporarily shift production away from the busiest hours is easier and cheaper for the grid to support.
Germany proposes dedicated land and faster development support
Germany’s data center strategy includes identifying land and creating measures to encourage additional investment.
What this means: Governments that want sovereign AI capacity must address more than funding. They need sites, grid connections, permits, fiber, and a predictable development process.
Announcing a national target is easier than creating the delivery pipeline behind it.
3) Green energy and environmental builds
Liquid cooling, flexible demand, dedicated generation, and the growing need to manage environmental trade-offs
Ecolab moves directly into liquid cooling
The CoolIT acquisition combines Ecolab’s water, chemistry, and monitoring capabilities with data center cooling hardware and engineering.
What this means: Cooling strategy increasingly connects energy efficiency, water use, equipment performance, and operating reliability.
The strongest solutions will optimize the full thermal system rather than one individual component.
Google agrees to curtail demand during grid peaks
Google’s utility agreements allow participating data centers to reduce their electrical load when local systems are under stress.
What this means: Flexibility is becoming an infrastructure resource.
A data center that can shift workloads, use batteries, or temporarily reduce noncritical operations may receive better access to constrained power markets.
SoftBank’s Ohio campus relies heavily on natural gas
The proposed Ohio development would use a large dedicated gas-generation system to support its AI load.
What this means: Private generation can improve speed and control, but it also creates long-term fuel exposure, emissions, maintenance, and permitting obligations.
Analogy: Owning a private fleet gives a company control over delivery, but it also makes the company responsible for fuel, repairs, drivers, and emissions.
The Ohio plan raises questions about consumer cost protection
Supporters argue that the project structure could make the data center carry a larger share of its transmission and generation costs rather than shifting those costs to households.
What this means: The environmental debate and the ratepayer debate are becoming connected. A project can reduce pressure on consumer bills while still creating concerns about emissions and fuel dependence.
Both questions need to be evaluated separately.
Liquid cooling becomes a water-management issue
Liquid systems can improve thermal efficiency at high rack densities, but they also require careful fluid selection, monitoring, maintenance, and water strategy.
What this means: “Liquid cooled” does not automatically mean environmentally superior.
The result depends on system design, water source, heat rejection, operating conditions, and whether waste heat can be reused.
4) Government policy that affects data centers
National AI strategy, public-private development, ratepayer protection, and government involvement in site selection
Germany adopts a national capacity-growth strategy
Germany’s government moved toward policies designed to double domestic data center capacity and expand AI-processing capability by 2030.
What this means: European governments increasingly view computing capacity as strategic infrastructure comparable to energy, transportation, and communications.
The U.S. government supports development on federal land
The Ohio campus would be located at a former federal uranium-enrichment site managed by the Department of Energy.
What this means: Public land and former industrial properties may become important tools for accelerating AI development.
These sites can offer acreage and infrastructure, but they may also carry environmental-remediation and community-history considerations.
The Ohio project forms part of a U.S.–Japan investment framework
The proposed development is connected to a broader U.S.–Japan trade and investment agreement involving critical technology, energy, and industrial infrastructure.
What this means: Data center development is becoming part of international economic policy.
Capital, energy, technology supply chains, and national security are increasingly being negotiated together.
AEP’s role creates a public-utility accountability question
AEP’s involvement in the Ohio project means regulators and customers will closely examine how transmission costs, risks, and long-term benefits are allocated.
What this means: Large projects must demonstrate not only that power can be delivered, but that existing customers will not be left carrying the downside if demand changes.
Google’s utility agreements offer an alternative policy model
The demand-response agreements show that policymakers and utilities do not have to choose only between approving new generation and denying new loads.
Flexible contracts can help data centers participate in grid management.
What this means: Future regulation may reward customers that can reduce demand, provide storage, commit to minimum payments, or fund required infrastructure.
What FP&A should take from this week
If you only remember three things for forecasting:
Cooling is becoming a strategic capital category. Higher-density systems will increase investment in liquid cooling, water treatment, controls, maintenance, and thermal-management expertise.
Dedicated power does not remove grid or policy exposure. Private generation can accelerate delivery, but transmission, fuel, permits, emissions, and regulatory oversight remain part of the economics.
Flexibility has measurable value. A data center that can adjust consumption may secure power sooner, reduce utility costs, and improve its position with regulators and communities.
Closing thought
Last week demonstrated that AI infrastructure is becoming a vertically connected system.
The data center developer needs the site. The customer needs the compute. The utility needs confidence in the load. The equipment supplier needs production capacity. The cooling system must manage the density. The government wants economic growth without shifting the cost to the public.
The best analogy is a large production network.
It does not matter how advanced the final assembly plant is if the electricity is unavailable, the cooling fails, the components arrive late, or the transportation system cannot handle the volume.
The winners will be the organizations that manage these dependencies as one operating model.
Because the AI infrastructure race will not be decided by the largest announcement.
It will be decided by who can connect capital, power, technology, cooling, and public policy into capacity that actually works.
The content is based on public information and personal analysis. This is not financial or investment advice.
