Summary

• Polarise launched Rise and Drive on 22 October, adding applications to its European AI infrastructure.
• Rise is available to teams; Drive uses controlled enrolment and an OpenAI-compatible model API.
• Increso is an early production customer, but customer reach and supplier sovereignty claims require careful attribution.

German AI infrastructure provider Polarise has launched an enterprise assistant and a managed inference service running on its European computing infrastructure, extending its business beyond supplying AI capacity into applications and developer services.

Announced at NVIDIA GTC Berlin on 22 October, Polarise AI Cloud includes Rise, a shared workspace for business users, and Drive, which provides access to open-weight models through an application programming interface. Rise is available to enterprises and teams from the launch date, while Drive is being introduced through a controlled application-based rollout.

Rise can be procured immediately, while general developer access to Drive has not been announced. Polarise also plans an Agentic Engine for configuring custom autonomous applications and support for Model Context Protocol integrations, but neither should be presented as an established capability of the initial Drive release.

Polarise says its computing and application stack is operated through European entities, with facilities in Germany and Norway, including a site at Munich’s Tucherpark. It is positioning local ownership, operational responsibility and control over processing as alternatives to consuming AI exclusively from international public cloud providers.

These are supplier descriptions of the service architecture, not an independent assessment that every customer configuration satisfies European regulatory or sovereignty requirements. Buyers still need to establish the precise location, contractual parties, access arrangements and model dependencies relevant to their workloads.

One service for employees, another for developers

Rise provides collaborative workspaces where employees can write, research and examine company documents. Polarise describes it as an enterprise assistant designed to keep customer information within the provider’s infrastructure rather than sending documents to a separate public chatbot service.

Such an application can reduce the need for employees to move information manually between systems, but governance depends on the permissions applied to underlying records. A document uploaded by an authorised employee should not automatically become accessible to every user in the same organisation.

Retrieval-augmented generation can improve responses by consulting relevant internal material, although it does not guarantee that the retrieved information is correct or current. Organisations also need rules for retention, document updates, sharing and the review of output used in consequential decisions.

Drive is aimed at software teams that want to call models through a managed service rather than operate their own accelerator hardware. Its API is described as compatible with the OpenAI development kit, allowing some existing applications to change service endpoints without a complete rewrite.

Compatibility reduces switching friction at the interface, but it does not mean every application will behave identically. Models vary in capability, context limits, output structure and supported tooling. Developers will need to test responses, latency, reliability and security before moving production workflows.

Polarise’s announced catalogue includes GLM, Kimi and NVIDIA Nemotron models, with payment based on token consumption. Usage billing can suit businesses whose demand fluctuates, since they need not reserve dedicated processors, although repeated inference calls and long conversations may still create substantial running costs.

The company did not provide independently verified price comparisons showing that its service is cheaper than alternatives for comparable workloads. Its claim that customers avoid a sovereignty premium should therefore be treated as a commercial position rather than a demonstrated market-wide result.

An early deployment does not establish general availability

Italian customer experience provider Increso is among the first organisations using Drive in production. Increso operates the Inxide platform, which supports customer service operations with AI agents and associated business processes.

Polarise says Increso’s services reach more than 10,000 end customers each day. That describes the customer service operations served by Increso, not a confirmed count of interactions processed through Polarise Drive or a measure of the new platform’s subscriber base.

Customer service provides a demanding implementation setting because agents may need to consult records, follow company policies and respond within tight service expectations. An assistant that answers questions is not necessarily entitled to alter an account, issue a refund or disclose private information.

Reliable deployment requires transaction limits, access controls, escalation routes and a way to investigate incorrect responses. These safeguards depend partly on the customer’s own applications and processes, not simply on where the language model runs.

Drive’s controlled rollout means other developers may need approval before gaining access. The existence of a production customer demonstrates a functioning service for that deployment, but does not establish broad general availability, typical performance or successful adoption across multiple industries.

Local infrastructure still has external dependencies

Polarise describes an approach in which existing buildings with power connections and planning permissions are converted into modular AI facilities. It says its Munich hub progressed from concept to operation in six months, a company-reported example that does not guarantee equivalent timelines at future sites.

Grid connections, cooling, equipment procurement and regulatory approvals can constrain infrastructure delivery even where buildings already exist. Hardware choices can also affect operating economics, particularly as customers demand more powerful models and longer sessions.

The service runs on NVIDIA Blackwell technology, with Polarise identifying Kerno as a principal supplier of its computing servers. That relationship illustrates how a European operated cloud can retain dependencies on hardware and software developed by international companies.

Polarise states that customer data stays within the EU and is not exposed to the US CLOUD Act. Those are the company’s legal and operational representations, not conclusions independently verified for every possible customer arrangement. Norway is part of the European Economic Area but is not itself an EU member state, so buyers should establish where their particular processing occurs.

The company reports ISO/IEC 27001 and ISO 9001 certifications. It says BSI C5 and SOC 2 Type I and Type II attestations are still in preparation, and they must not be described as completed certifications.

For Polarise, Rise and Drive open potential revenue streams from subscription software and metered inference rather than infrastructure alone. The initial launch is concrete, but its commercial scale remains undisclosed: the company has not supplied independently audited revenue, customer retention or comparative performance figures for the new services.

Customers assessing the offering will need to test service reliability, the actual range of available models and the provider’s contractual controls. Those results, together with delivery of the planned agent features, will determine whether local operation becomes a compelling enterprise alternative rather than a sovereignty claim attached to another cloud interface.