Bunq adopts Orq.ai router amid Europe AI sovereignty push

Bunq has replaced its in-house large language model routing infrastructure with a newly launched standalone product from Amsterdam-based Orq.ai, as European businesses face rising AI costs and growing pressure over data and infrastructure sovereignty.

The deal comes as Orq.ai releases its Router as a standalone product, rather than only as part of a broader agent lifecycle platform. Orq.ai positions the Router as an entry point for engineering teams that need to manage how requests are sent to different AI models through a single gateway.

Shift in focus

Companies building generative AI services are increasingly using multiple models from different providers. This can reduce dependence on a single vendor and give teams greater flexibility in pricing and performance. It also adds operational overhead, especially when teams need consistent monitoring, auditing, and reliability across services.

In regulated sectors such as financial services and healthcare, that overhead can become a governance issue. Teams need clarity on where inference runs, which vendors handle prompts and outputs, and how data moves between regions. They also need to control spending as usage grows and providers change pricing.

Bunq felt those pressures while running an in-house routing system. "We built our own LLM routing infrastructure, but maintaining it became increasingly expensive and time-consuming, while still leaving gaps in observability and performance," said Benjamin Kleppe, GenAI Lead at bunq. "We chose to work with Orq.ai to replace that internal setup with a production-ready AI Router that meets our governance, scalability, and cost-monitoring requirements."

Routing tools sit between an application and the model providers that process requests. They can direct traffic to different models based on rules such as cost, response time, or where the computing occurs. They can also centralise logging and tracking for troubleshooting and compliance.

Orq.ai is positioning routing as a control layer rather than a background integration. "As soon as AI systems move beyond a single model, routing turns from plumbing into a production bottleneck. Making the router standalone lets teams regain control early, with a single line of code," said Sohrab Hosseini, Orq.ai's co-founder.

Sovereignty pressures

Orq.ai is also using the release to address European concerns about sovereignty, as organisations seek more choice over infrastructure location and ownership. Many enterprises are asking how quickly they can switch AI suppliers or change deployment regions if regulations tighten or geopolitical disruptions disrupt supply chains.

Hosseini tied those concerns to the routing layer. "In Europe, AI sovereignty is no longer an abstract policy debate; it's a direct consequence of today's geopolitical reality," he said. "Enterprises need to know where AI inference runs, who controls the infrastructure, and how quickly they can adapt as conditions change. Those decisions are enforced at the routing layer, which is why we made it available as a standalone product."

Customers can define routing policies based on geography, latency, cost, or custom constraints. Orq.ai says the Router can run within a customer's infrastructure, enabling deployments with public, private, or both public and private models.

Pricing approach

Costs have become a central issue for businesses scaling AI features into production. Many AI gateways and routing services charge fees that rise with traffic volume. Some providers also add a percentage markup on top of underlying model costs, which can compound as usage grows.

Orq.ai states that its Router does not add platform markup for routing. Instead, it charges for tracing and logging of processed data based on volume, separating routing from usage-based observability services.

The standalone Router can be deployed on its own and later expanded into Orq.ai's wider product set, which includes tools for experimentation, evaluation, and monitoring of AI agents. Orq.ai also says its platform supports access to more than 300 models through an AI gateway and includes an agent runtime for production deployments.

Customer signal

Bunq's move from an internal system to an external routing product reflects a broader question facing engineering teams. Many firms began with custom integrations during early experimentation with large language models. Production rollouts have increased demand for consistent controls across providers and for clearer cost visibility.

For vendors, the routing layer has become a competitive battleground. It can influence which models get used, how quickly teams adopt new providers, and how data governance is enforced. Tools in this position can also serve as a common point of reference for observability and policy decisions across multiple AI applications.

Orq.ai says the Router is available immediately as a standalone product. It expects teams to start with routing and cost control, then expand into broader agent lifecycle functions as operational requirements grow.