AI is now said to be rewiring the economics of data centers. This, according to a recent update shared by CB Insights which noted that steady growth in large language models is increasingly pushing data center buildouts into the multi‑GW range — equivalent to “a major city’s worth of electricity.” Combined with grid interconnection queues (waitlists to plug into the grid) that stretch for years, this is “creating a structural power bottleneck for AI infrastructure.”
The CB Insights update further stated that the ability to deliver power, cooling, and interconnect at speed is “becoming a defining advantage for the small cohort of platforms with the capital and execution capabilities to operate at scale.”
CB Insights pointed out that across the data center value chain, new companies are emerging “to keep pace with AI buildouts.”
From power generation and AI chips to supporting infrastructure and cloud services, vendors are “competing for a share of a data center market that could reach nearly $7T.”
Using CB Insights’ latest Mosaic Scores, Hiring Insights, markets, and business relationships, they reportedly map 367 companies across 45 markets to show where “advantage is consolidating — and look at the strategic moves data center developers should prioritize now.”
They mapped the data center value chain across 7 layers:
- Power generation: Energy production technologies that power AI infrastructure, from next-generation nuclear reactors (fusion, SMRs, molten salt) to renewable sources like geothermal power and battery storage systems that balance grid demand.
- AI computing hardware: Specialized processors and chips designed for AI workloads, including neural processing units (NPUs), AI training and inference processors, GPUs, CPUs, and accelerator modules that power machine learning computation.
- Supporting hardware infrastructure: Physical systems that enable efficient data center operations, from advanced cooling solutions (immersion and direct liquid cooling) to energy optimization software, infrastructure management platforms, and virtual power plants.
- Networking & connectivity: Network infrastructure that enables secure, high-speed data movement across data centers, including security tools (NDR, firewalls), core networking services (DNS, CDN, WAN optimization), and interconnection fabric for multi-site deployments.
- Data storage & database infrastructure: Storage and database platforms optimized for AI data pipelines, from vector databases and real-time platforms to NoSQL and relational systems, plus streaming and messaging infrastructure that moves data between systems.
- AI cloud services: Cloud-native platforms and services that simplify AI deployment, including management and monitoring tools, migration services, cloud storage, GPU compute resources, and AI-specific security solutions.
- Data center facilities & operations: Physical data center infrastructure across scales and form factors, from edge and micro data centers to hyperscale facilities, plus modular solutions and emerging concepts like space-based data centers.
They selected companies for inclusion “based on Mosaic startup health and potential scores (550+).”
As clarified in the latest update, this recently updated market map is “not exhaustive of companies operating in the space.”
Key takeaways from CB Insights are as follows:
- AI-ready capacity is concentrating with mega-financed platforms that can translate capital into multi-GW delivery. CB Insights’ Mosaic and Hiring Insights flag Aligned and Vantage Data Centers as emerging leaders converting large financings into rapid deployment, with Aligned in the top 1% by Mosaic score and Vantage in the top 2% by Hiring Momentum score. Aligned is anchoring BlackRock’s $100B AI infrastructure buildout plan by using “skidded” manufacturing — pre-assembled, modular systems — to accelerate timelines, while
- Vantage recently raised $1.6B and is recruiting heavily to deliver the Stargate project with Oracle and OpenAI. As these players lock in resources via integrated power solutions — like Vantage’s push for modular nuclear and microgrids — expect pricing power and interconnect priority to consolidate with this cohort.
- Direct liquid cooling for chips is becoming the standard over conventional air‑cooled solutions. AI chips generate 3-5x the heat density of traditional servers, driving a rapid shift to direct liquid cooling — 2025 YTD funding has surged nearly 5x YoY to $546M. Submer, which has the highest Mosaic score in this segment, is seeing adoption accelerate — revenues jumped 76% year over year to $156M in 2024 and are projected to grow further in 2025.
- Virtual power plant (VPP) orchestration is shifting from a grid stability tool to a critical enabler for AI data center expansion. Our recent VPP brief highlighted momentum tilting toward capacity markets like PJM, driven by the urgent need to secure power for hyperscale infrastructure.