March 27, 2026

Aligned Data Centers just grabbed a $2.58bn revolving credit facility — and it’s hard to read it as anything other than lenders leaning into the AI-era data centre land grab. The timing matters: this comes right ahead of Aligned’s pending ~$40bn sale to a BlackRock Global Infrastructure Partners-led group. In other words, the buyout story isn’t just a headline; the balance sheet is already being set up to build.

The Big Stories

Aligned Data Centers raises $2.58bn revolving credit for expansion to finish six later-stage US facilities and fund future development, backed by assets in Dallas, Phoenix, and Northern Virginia. PGIM is an anchor lender; the facility matures in three years with two one-year extension options. The signal here is straightforward: the cost of capital still matters, but the market is still willing to underwrite scale for operators sitting in the core AI corridors — especially with a marquee sponsor transaction expected to close later this year.

Ecolab to acquire liquid data-center cooling provider CoolIT for $4.75bn from KKR, reportedly delivering ~15x on KKR’s 2023 equity investment. Ecolab’s pitch is about pairing CoolIT’s engineered liquid cooling with Ecolab’s water, chemistry, and digital services to sell something closer to “Cooling-as-a-Service” for AI halls. This is the increasingly obvious industrialization of AI infrastructure: cooling is moving from “component” to “platform,” and strategic buyers are paying up to own the bottlenecks.

Arm launches Arm AGI CPU purpose-built for AI data centers, with Meta as lead partner and first customer. The chip is designed by Ampere, manufactured by TSMC on 3nm, and is specced at up to 136 Neoverse V3 cores with a 300W TDP; Arm is claiming more than 2x performance per rack versus x86. Even if you discount marketing math, the more important point is ecosystem intent: Arm isn’t just licensing cores into other people’s roadmaps — it’s trying to shape the CPU lane of “agentic AI” data centres alongside hyperscaler demand.

Panasonic: Hyperscalers Reserve 80% Battery Output for AI through FY2029, with plans to convert automotive cell lines from FY2027 and build a new module plant near its Mexico facility. The detail that should make operators sweat is the allocation dynamic: structural supply tightness that “favours hyperscalers over enterprise and colocation customers.” Rack-level backup and AI power spikes are turning what used to be a procurement line item into a competitive differentiator.

Policy and regulatory pushback is also getting less theoretical. Sanders proposes pause on data center builds pending AI rules — unlikely to pass, but politically potent as a framing device tying data centres to AI governance. Meanwhile, Groups sue PSC to halt Georgia Power data center expansion over a December approval that the challengers say effectively blesses unnecessary investment (they cite $50–60bn and 757MW) to serve data-centre-driven load growth. Add the utility’s own claim that 80% of future demand will come from data centres, and you get the real investor takeaway: the next constraint may be less “can you build?” and more “can you build without triggering a ratepayer and permitting backlash?”

Behind the Headlines

AI inference drives data center network and optical demands makes a point the market still routinely underprices: inference, not training, is where the volume shows up. Training clusters are lumpy capex events; inference is the always-on business that forces sustained investment in DCI, optical, and scalable connectivity. If over 40% of new builds will be AI-focused (per the cited Ciena survey), the second-order winners won’t just be GPU and power players — they’ll be the companies selling the fabric that keeps distributed inference performant across metros and regions.

Veritone shifts AI workloads to Oracle Cloud Infrastructure (OCI) is a useful case study in where “sovereign” and “federal-compliant” requirements are pulling AI platforms. The company is migrating key products (aiWARE, Data Refinery, Data Marketplace) under a multi-year deal and emphasizes years of re-architecting and containerization to enable portability. The quiet message: cloud choice is becoming as much about compliance posture and deployment control as it is about raw price/performance.

atNorth joins European Data Center Association to shape policy reads like a membership update, but the subtext is the real story: operators are trying to get ahead of the EU policy machine while they expand AI-ready capacity. atNorth points to eight Nordic data centres, new builds in Kouvola (Finland) and Ølgod (Denmark), a Stockholm campus, and a “mega site” in Sollefteå (Sweden), alongside a climate-neutral-by-2030 target. As grid, water, and permitting debates heat up, trade-group influence is becoming part of the build strategy — especially in markets selling themselves on sustainability as a competitive moat.