Volta
by Volta Infra Holdings
AI factories financed, built and operated like a utility — long-term GPU capacity for labs without hyperscaler balance sheets
Volta is a vertically integrated AI infrastructure company that finances, develops, builds and operates gigascale GPU campuses, then sells the capacity to frontier AI labs and AI-native companies on multi-year contracts. It is aimed at organisations that need dedicated, contracted compute at hyperscaler terms but lack the investment-grade balance sheet that neoclouds normally require before signing.
Volta is a vertically integrated AI infrastructure company that finances, develops, builds and operates 'AI factories' — purpose-built GPU campuses — and sells the resulting capacity to frontier AI labs and AI-native companies under long-term contracts. Founded in January 2026 by Ricard Boada and Sofia Gumuzio, who previously built Brookfield Asset Management's AI infrastructure platform, the company exited stealth on 4 August 2026 having raised roughly $300 million across combined seed and Series A rounds at a $2.4 billion valuation, co-led by Azora, Andreessen Horowitz, Altimeter and NVIDIA, with participation from Michael Dell's family office. Its distinguishing move is financial rather than purely technical: most neoclouds contract with hyperscalers because their credit supports long-dated project debt, whereas Volta assembles credit support, project equity and senior infrastructure debt behind each individual deployment so AI companies without investment-grade balance sheets can still sign multi-year capacity agreements — a16z, which co-led the Series A, described the model as building 'the neocloud for Little Tech.' Volta also acquired the operations of Genesis Cloud, which had run one of Europe's earliest GPU-first clouds since 2018 and served more than 20,000 users, giving it working software for operating a public AI cloud and managing bare-metal clusters rather than having to build that layer. Its first disclosed site is a 133 MW hydro-powered facility in Tydal, Norway, developed alongside Bitdeer Technologies Group on NVIDIA Vera Rubin systems and anchored by a $10 billion six-year agreement; Volta announced the customer only as an unnamed AI lab, and Bloomberg reported it as Anthropic. Campuses are built on NVIDIA's DSX platform for extreme GPU density with direct liquid cooling. A separate $5 billion AI Infrastructure Program with asset manager Azora funds the pipeline, which Volta says exceeds 1 GW of near-term power capacity across North America and Europe, targeting several gigawatts deployed by 2030. The team is roughly 100 people across London, Palo Alto and New York.
The head of infrastructure at a well-funded AI lab or AI-native scale-up that needs tens to hundreds of megawatts of dedicated GPU capacity on a multi-year contract but cannot clear the credit bar hyperscaler-oriented neoclouds apply.
Contracted access to purpose-built NVIDIA Vera Rubin capacity with the financing structured by the provider, rather than being turned away for lack of an investment-grade balance sheet.
At a Glance
- Category
- Infrastructure & Cloud
- Pricing
- Contact for pricing
- Target Market
- CTOs, CIOs, VPs of Infrastructure, ML Platform Leaders, AI Research Leaders
- Deployment
- Cloud-first, Multi-cloud
- Founded
- 2026
- Headquarters
- London, United Kingdom
- Team Size
- 51-200
Key Features
- ✓Vertically integrated AI factories
Volta controls institutional capital, powered land, the data centre, compute, software and operations in one platform, removing the handoffs that delay AI capacity delivery
- ✓Project-finance structuring for the customer
Credit support, project equity and senior infrastructure debt are assembled behind each deployment so customers without investment-grade balance sheets can sign long-term contracts
- ✓NVIDIA Vera Rubin deployments
Sites run NVL72-class racks pairing 72 Rubin GPUs with 36 Vera CPUs, giving customers access to a current-generation NVIDIA architecture at scale
- ✓NVIDIA DSX gigascale campus design
Campuses are engineered from the ground up for extreme GPU density with direct liquid cooling rather than retrofitted into general-purpose colocation space
- ✓Genesis Cloud operating software
The acquired platform provides public AI cloud services and bare-metal cluster management, proven across more than 20,000 users since 2018
- ✓$5 billion Azora infrastructure program
A dedicated project-equity vehicle backed by senior infrastructure debt finances new campuses, lowering cost of capital versus funding builds from venture equity
Use Cases
- •Multi-year frontier model training capacity
A $10 billion six-year agreement anchors the Norway site, giving an AI lab dedicated capacity across the full training and serving cycle
- •Low-carbon European AI compute
The 133 MW Tydal, Norway facility runs on hydroelectric power, addressing both sustainability targets and European siting requirements for AI workloads
- •Scaling inference for AI-native products
AI-native scale-ups contract dedicated capacity instead of competing for on-demand GPUs, making unit economics predictable as usage grows
- •Converting crypto-mining sites to AI capacity
The Bitdeer partnership repurposes an existing powered data centre estate for AI, compressing the timeline versus greenfield construction
- •Diversifying away from hyperscaler compute concentration
Labs already contracted with AWS, Google or Microsoft add a fourth independent supplier to reduce single-vendor dependency on capacity
Ideal For
Best For
- ✓Frontier model training runs that need contiguous, dedicated capacity for years rather than burst access to a shared pool
- ✓AI-native companies with strong venture backing but no investment-grade credit rating seeking long-term compute commitments
- ✓Workloads with European data-sovereignty or low-carbon requirements — the first site is hydro-powered in Norway
- ✓Organisations that want the newest NVIDIA silicon generation early, via Volta's NVIDIA Cloud Partner status and Vera Rubin deployments
- ✓Buyers who want a counterparty that owns land, power, the data centre and operations rather than reselling someone else's capacity
Not Ideal For
- ✗Teams needing on-demand, self-serve or short-term GPU rental — Volta's disclosed business is multi-year contracted capacity, and no public pricing or signup exists
- ✗Buyers who require an operating track record: the company was founded in January 2026 and its first announced site is still being delivered with a partner, so nothing has been run at gigascale yet
- ✗Enterprises that need compliance attestations today — no SOC 2, ISO 27001 or similar certification is publicly documented
- ✗Smaller inference workloads better served by per-hour GPU clouds or a serverless inference provider
- ✗Organisations uncomfortable with concentration risk, given that a single anchor agreement represents the bulk of disclosed revenue
Deployment
Market Analysis
Pros
- ✓Financing model opens long-term contracted capacity to AI companies that hyperscaler-oriented neoclouds would decline on credit grounds
- ✓Exceptional investor and supplier alignment — NVIDIA is both backer and chip supplier, alongside a16z, Altimeter and Azora
- ✓Founders built Brookfield's AI infrastructure platform, so the project-finance and power-siting expertise is genuine rather than aspirational
- ✓First site is hydro-powered in Norway, which addresses European siting and low-carbon requirements at the same time
- ✓Acquiring Genesis Cloud's operations shortcut the control-plane problem that has slowed other new entrants
Cons
- ✗Founded in January 2026 with no operating track record at gigascale — the first announced site is still being delivered alongside Bitdeer
- ✗Extreme customer concentration: a single anchor agreement accounts for essentially all disclosed contracted revenue
- ✗Reuters could not independently verify that the $10 billion counterparty is Anthropic; Volta itself named only an unnamed AI lab, so the headline customer remains a press report
- ✗No published pricing, no self-serve access and no compliance certifications documented publicly, which makes evaluation slow for enterprise procurement
- ✗The neocloud segment is consolidating around a few heavily contracted providers (CoreWeave's $99.4B backlog, Nebius's Meta and Microsoft deals), raising dependency risk for buyers
- ✗Delivery depends on partners, power availability and construction schedules rather than on software the company controls
Pricing
Dedicated AI factory capacity
Contact for pricing
- ✓Multi-year contracted GPU capacity
- ✓NVIDIA Vera Rubin systems
- ✓Direct liquid cooled gigascale campuses
- ✓Volta-arranged project financing
- ✓Operated infrastructure with bare-metal cluster management
Volta publishes no list pricing and there is no self-serve signup. Capacity is sold through negotiated multi-year contracts — the one disclosed deal is $10 billion over six years for 133 MW — so the commercial unit is megawatts of contracted power over a term, not GPU-hours. A distinctive part of the commercial model is that Volta arranges the financing behind each deployment (project equity from the $5 billion Azora program plus senior infrastructure debt led by international banks) rather than requiring the customer to bring investment-grade credit, which is the practical difference for buyers who would otherwise be declined. Expect enterprise procurement, long lead times tied to construction schedules, and terms driven by power availability rather than a published rate card.
Security & Compliance
Sources
This page was written from 6 sources, 6 on domains other than volta.com.
- 1.techcrunch.com — anthropic signs 10 billion deal with ai cloud startup volta
- 2.a16z.com — investing in volta
- 3.cloudcomputing-news.net — volta ai cloud deal
- 4.finance.yahoo.com — anthropic signs 10 billion computing 180641716
- 5.pymnts.com — volta exits stealth 2 billion dollar valuation build ai infr
- 6.finsmes.com — volta infrastructure raises 300m in combined seed and series
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Redis Iris
Real-time context engine giving AI agents governed retrieval, live operational data and durable memory
Kong AI Gateway
One gateway for LLM, MCP and agent-to-agent traffic, built on the Kong data plane you already run
NVIDIA NeMo Switchyard
Open-source model router that sends each agent step to the cheapest model that can still do the job
WEKA NeuralMesh
Microservices storage and memory fabric built for AI training and inference at exabyte scale