AI Infrastructure Platforms Race to Cut Deployment Time as Power Constraints Bite
DG Matrix and InfraPartners launch an end-to-end AI-ready power and factory platform to accelerate deployment. Meanwhile, Delos Data offers AI chip startups a fast track to rack-scale validation, addressing networking and power hurdles.
Two new infrastructure platforms are targeting the biggest bottleneck in AI growth: the time it takes to get new compute capacity online. On June 12, 2026, DG Matrix and InfraPartners unveiled what they call the first end-to-end AI-ready power and AI factory platform, combining software-defined power distribution with prefabricated data center architecture. Separately, Delos Data announced a service that lets AI chip startups validate their hardware at rack scale without building their own testbeds.
The DG Matrix and InfraPartners platform is designed to reduce time-to-compute by integrating power infrastructure directly with the AI factory build. The system uses software-configurable power distribution that supports both AC and DC feeds, including 800VDC readiness, and is built around prefabricated, upgradeable modules. The goal is to cut deployment timelines that currently stretch into years due to grid interconnection delays and traditional construction schedules.
Software-defined power meets modular data centers
The platform addresses three specific pain points for AI operators: stranded capital risk from inflexible infrastructure, the need to support evolving GPU and accelerator architectures, and the challenge of scaling globally with consistent designs. Key features include:
- Software-configurable power distribution that can switch between AC and DC feeds without hardware changes.
- Prefabricated data center modules that can be deployed in months rather than years.
- Native support for 800VDC distribution, which reduces power losses and supports higher rack densities.
- Energy optimization software that balances load across power sources and cooling systems.
- Integration with liquid cooling and energy storage systems to handle peak AI workloads.
The platform is aimed at hyperscale operators, colocation providers, and enterprises building private AI factories. DG Matrix provides the power electronics and software, while InfraPartners contributes the modular data center design and construction expertise.
Startups get a fast track to rack scale
Meanwhile, Delos Data is tackling a different part of the AI infrastructure problem: validation. According to The Register, Delos Data offers AI chip startups a service that lets them test their hardware at rack scale without having to build their own networking and power infrastructure. The company provides pre-configured racks with networking, cooling, and power distribution, allowing startups to focus on chip performance rather than infrastructure integration.
This approach addresses a common pain point for AI chip startups: building an Nvidia NVL or AMD Helios competitor requires not just a good chip, but also the networking and power systems to make it work at scale. Delos Data's service reduces the time and cost of getting from a single chip to a validated rack-scale system.
Both platforms reflect a broader shift in the AI infrastructure market. As power constraints become the defining challenge for AI growth, the industry is moving away from bespoke, one-off builds toward standardized, integrated platforms that can be deployed faster and scaled more easily. The next step for both DG Matrix and Delos Data will be landing anchor customers and proving that their platforms can deliver on the promise of faster time-to-compute.
Fact check
-
DG Matrix and InfraPartners unveiled the first end-to-end AI-ready power and AI factory platform on June 12, 2026.
reported · source
-
The platform uses software-configurable power distribution supporting AC and DC feeds, including 800VDC readiness.
reported · source
-
Delos Data offers AI chip startups a service to validate hardware at rack scale without building their own testbeds.
reported · source
-
Building an Nvidia NVL or AMD Helios competitor requires networking and power systems to work at scale.
reported · source
Source reporting (3)
Join the conversation
You need to be registered and logged in to comment on blog articles.
0 Comments
No comments yet
Be the first to share your thoughts on this article.