AI Infrastructure Newsletter

From this week’s Data Center Frontier Trends Summit, Flexential CEO Chris Downie warns that AI is forcing the data center industry to toss out its old playbook.

Despite the fever-pitch hype around generative AI, reality is proving more complicated: MIT NANDA’s recent report on the “Gen AI Divide” found only 5% of enterprises are seeing real ROI so far, despite pouring tens of billions of dollars into pilots that are rarely making it to production-level scale.

AI infrastructure is being built at breakneck speed, and the strategies are diverging. At the hyperscale end, Meta just signed a reported $10 billion deal with Google Cloud to secure the compute power for its 1.3 million GPU ambitions.

A new Google study puts up some fascinating numbers about the holistic environmental impacts of AI, and how to measure them: Not just water, not just power, and not just based on conditions during active queries.

AI is reshaping both physical infrastructure and political debates — and the two of them often interconnect. On the physical infrastructure side: In Texas, we see Vantage going big with plans for a $25 billion hyperscale campus to fuel AI growth.

Developments around Intel’s future and ownership are shifting quickly. SoftBank has announced a $2 billion investment in the company, becoming the chipmaker’s fifth-largest shareholder and deepening its global AI footprint as Intel works to regain ground in AI and foundry services.

CoreWeave is the latest company to report on quarterly earnings that showed strong growth in AI infrastructure revenues — along with ballooning capex that reflects the steep costs of scaling an infrastructure business.

The race to build AI-ready capacity is heating up, with Dell’Oro forecasting the physical data center infrastructure market to reach $63.1 billion by 2029, growing at a 15% CAGR on the back of liquid cooling, busway systems, and hyperscaler-scale deployments.

Cisco closed fiscal 2025 with AI infrastructure orders surpassing $2 billion—double its goal—and locked in two $1 billion webscale deals for 2026, underscoring hyperscalers’ accelerating network buildouts.

AI may be driving some workforce cuts, but it’s also generating various new jobs — and in both AI infrastructure and AI in general, finding enough of the right people is definitely an industry pain point.