Private AI: Xference launches its European infrastructure from Aruba data centers

• Italian startup Xference begins operations at Aruba data centers in Ponte San Pietro (Bergamo) and officially opens the invitation-only beta phase to the first 100 selected early adopters. It is the first European inference provider supported by Aruba to date.
• Xference chose Aruba data centers, which guarantee high per-rack energy availability, scalability, and operational management of machines, networking, and security.
• A European private AI inference infrastructure that will ensure users' digital autonomy: data ownership and sovereignty, the highest privacy standards, reduced energy consumption, and a protected environment.

Italian startup Xference inaugurates its European AI inference infrastructure at the Aruba data center in Ponte San Pietro and officially opens the invitation-only beta phase to the first 100 selected early adopters.

The initiative marks a concrete step in building a distributed infrastructure across European territory dedicated to private AI inference, designed to ensure data sovereignty, digital autonomy, regulatory compliance, and high energy-efficiency performance. The service enters its invitation-only beta phase today, with the first 100 selected users testing the new technology developed in Italy.

A European infrastructure for private and sovereign AI

Founded in 2025 in Cagliari, Xference was created with the goal of democratizing access to artificial intelligence by offering an inference service based on open source models, running in private environments with data that always remains under client control. To support this vision, the startup chose Aruba as its technology partner for colocation and server provisioning through a Hardware-as-a-Service model, launching infrastructure operations at the IT3 data center in Ponte San Pietro (Bergamo). Aruba, with its European infrastructure and longstanding pioneering role in innovation, strengthens its connection with deep tech startups by contributing to the development of European AI-driven platforms.

"Launching our production infrastructure on Aruba represents a key milestone: we enter the beta phase with the first users who will be able to harness the power of generative AI without their data ever leaving their perimeter," says Andrea Pili, Co-Founder and CEO of Xference. "Aruba has proven to be a partner fully aligned with our mission, supporting us from the earliest stages with responsiveness, per-rack scalability, and an operational model perfectly suited to the needs of a deep-tech startup. This is a first step toward real European digital autonomy."

High-density data centers for the new generation of private AI with open source LLM models

Xference's infrastructure requires energy density levels significantly higher than the industry average, with demands exceeding 20 kW per rack, along with linear growth capacity to support future expansion. Aruba provided a solution capable of supporting the startup's growth trajectory, handling server provisioning and operational activation according to the required configurations, as well as managing networking and security components.

"The operational launch of Xference's infrastructure as the first European Inference Provider in our data centers demonstrates how our network of European infrastructures can deliver state-of-the-art AI-ready systems and support the development of new high-performance AI platforms," says Giancarlo Giacomello, Head of Data Center Offering & Colocation Services at Aruba. "Through the combination of colocation and Hardware-as-a-Service, Xference can scale rapidly by converting capital investments into operational costs, while maintaining the highest standards of security, compliance, and reliability, and ensuring full data governance and localization, security and compliance by design, with energy from renewable sources."

Beta phase kicks off with the first 100 users

The beta phase, launched by invitation, will involve the first 100 early adopters who will help test and optimize the service in real-world usage scenarios. Upon completion of this phase, Xference plans a progressive and controlled rollout, with the goal of expanding the infrastructure and developing a distributed edge architecture capable of bringing compute nodes closer to end users, further reducing latency and consumption.

"This beta phase represents the beginning of a gradual and sustainable growth path," concludes Pili.
"We will monitor workloads in real time to scale the infrastructure efficiently and build, step by step, a leading independent European Inference Provider."

For more information: https://enterprise.aruba.it/case-study/per-xference.aspx