DiscreteStack is building for European businesses that want powerful AI without handing control of their data and infrastructure to an external cloud provider. The Sofia-based company lets enterprises deploy open-weight AI models on infrastructure they choose, including their own data centers, private clouds and fully isolated environments where sensitive information never needs to leave the organization.
The startup has raised €800,000 in seed funding led by CleverPine Ventures, joined by strategic angel investors Milen Manev and Stoil Vasilev, alongside smaller investors. The capital will help DiscreteStack expand into finance, insurance, and the public sector while increasing access to GPU capacity for further inference-optimization work.
DiscreteStack was founded in 2025 by Hristo Todorov, a software engineer and entrepreneur whose background spans telecommunications, fintech, aviation, and enterprise software. His earlier companies included Upnetix and CleverPine, while work with large organizations exposed a recurring problem: enterprises wanted advanced AI but were increasingly uncomfortable surrendering control over critical data and infrastructure.
Instead of training another foundation model, DiscreteStack focuses on making existing open-weight models practical for enterprise deployment. Its patent-pending optimizations are designed to run even large models on a single standalone server, while deployments are tuned to the specific GPUs being used. An intelligent runtime manages workloads, caching, and compute requirements as usage changes.
The business model is equally important. Rather than charging by tokens consumed, DiscreteStack uses a fixed annual per-server license, giving enterprises greater cost predictability. The company says certain deployments can reduce AI costs by as much as 90%, although that figure remains company-reported.
DiscreteStack’s larger bet is that European AI sovereignty will eventually become an enterprise purchasing decision, not just a policy discussion. If AI becomes critical business infrastructure, the advantage may belong to companies that can decide where their intelligence runs, who can access it, and what it costs without asking another platform for permission.
