What happened
Capital is flowing into the physical infrastructure of AI at a scale that now rivals the funding rounds of the AI companies themselves. Three raises reported around November 9, 2026 illustrate the wave.
Mavria raised $130 million cumulatively at a reported $1 billion valuation, according to The Wall Street Journal on November 9. Mavria builds factory-built modular AI data centers — facilities assembled in a factory and deployed as modules, rather than constructed entirely on site.
Avantus raised $1.28 billion. Avantus is a large-scale solar photovoltaic and energy storage developer. Its business is generating and storing the power that data centers and the wider grid consume.
EdgeCore raised $521 million. EdgeCore operates in cloud, data center and AI infrastructure.
The context around these rounds is just as important as the rounds themselves. Anthropic reportedly committed $3.4 billion to gas generators for future data centers, according to The Neuron on November 10. Google expanded its TPU portfolio with TPU 8tx and TPU 8i Max — eighth-generation, high-memory and confidential-computing variants of its tensor processing units, the chips that power much of Google's AI work.
Why modular data centers
A traditional data center is a construction project: land, permits, concrete, cooling, power connections, and a timeline measured in years. Demand for AI compute is growing faster than that timeline allows.
Factory-built modular data centers are an attempt to compress it. By building modules in a factory — where conditions are controlled, labour is concentrated and quality can be standardised — and then shipping and assembling them on site, a developer can in principle deliver capacity faster and more predictably.
Mavria's reported $1 billion valuation, on $130 million raised cumulatively, suggests investors believe this approach has a durable advantage. The risk is execution: modular construction still requires land, power and permits, and the factory itself must be funded and operated at scale. A modular design does not remove the power constraint; it only changes how the building is delivered.
Power is the real bottleneck
Avantus's $1.28 billion raise and Anthropic's reported $3.4 billion gas-generator commitment point to the same constraint from opposite directions. Avantus develops solar generation and storage — adding new clean power to the system. Anthropic's gas generators are a way to secure power on site, independently of the grid.
Both reflect a single fact: AI data centers consume enormous amounts of electricity, and in many places the grid cannot supply it quickly enough. A data center without reliable power is an expensive empty building. The companies that can secure power — by generating it, storing it, or bringing their own generation on site — hold a decisive advantage.
Chips: Google's TPU expansion
Google's expansion of its TPU portfolio with TPU 8tx and TPU 8i Max adds a further layer. TPUs are Google's own AI chips, an alternative to buying accelerators from external suppliers. The eighth generation, with high-memory and confidential variants, is designed for the largest models and for workloads where data privacy matters.
The confidential-computing variant is worth noting. As AI moves into regulated industries — finance, healthcare, government — customers increasingly need guarantees that their data is protected even while it is being processed. Hardware-level confidentiality is one way to provide that assurance.
For the broader market, Google's continued investment in its own chips signals that the compute layer will remain competitive. No single supplier will set the terms unchallenged, and buyers of compute will have more than one source to negotiate with.
What this means for founders
If you are building an AI company, your infrastructure costs are not a detail. They are a strategic variable. The price and availability of compute, the location of data centers, and the cost of power will shape your margins and your product decisions.
Founders who depend on a single compute provider should understand that provider's capacity constraints and pricing power. Founders who can design their products to run efficiently — using less compute for the same result — gain a cost advantage that compounds as they scale.
There is also an opportunity in the infrastructure layer itself. The wave of capital into modular construction, power generation and storage suggests that investors are looking for the picks-and-shovels businesses of the AI era, not only the model builders.
What this means for investors
The valuations in this wave require careful underwriting. A reported $1 billion valuation for Mavria, on $130 million raised, prices in rapid growth and successful execution of a capital-intensive manufacturing model. Avantus's $1.28 billion and EdgeCore's $521 million are large commitments to businesses whose returns depend on power prices, construction timelines and customer demand holding up over many years.
Diligence should focus on contracted demand: who has committed to buy the capacity, for how long, and on what terms? A data center with signed, long-term customers is a different investment from one built on speculation that demand will arrive.
Angeled.com connects founders with investors.
The Neuron, November 10, 2026
Read Next
General Translation Raises $27 Million Series A Led by a16z to Build Agent-First Localization Infrastructure
LangChain Raises $300 Million Led by CapitalG at a $3 Billion Valuation to Build the Agent Stack
Mount Thor Raises $45 Million to Run Cloud Fleets of Macs for AI Agents
OpenAI Reportedly Raises $15 Billion at a $1.25 Trillion Valuation
Q-CTRL Raises $290 Million Series C Led by Salesforce Ventures, Reaching Unicorn Status
SE3 Raises $30 Million Series A Led by Accel



