SPRIND launches application process for €125 million competition.

      The Next Frontier AI Challenge, introduced at EurIPS in December, instructs applicants not to attempt to catch up with OpenAI but instead to leapfrog into the next architectural S-curve. There is up to €1 billion in additional funding offered for the three winning labs.

      Today, SPRIND, Germany’s federal agency for breakthrough innovation, opened applications for its Next Frontier AI Challenge, which features a €125 million, 24-month structured competition aimed at establishing up to three European frontier AI labs from the ground up.

      The application period remains open until 1 June 2026, with jury presentations slated for 24–25 June, and the first ten funded teams expected to start in July.

      The competition was unveiled at EurIPS in Copenhagen on 3 December 2025, a prestigious European conference officially backed by NeurIPS, the foremost AI research conference worldwide. The message is clear about Europe’s current standing:

      “Europe’s competitiveness in AI innovation lags significantly behind that of the USA and China,” states the challenge brief. “If Europe does not develop its own models, it risks becoming increasingly dependent on these technologies.”

      The goal is not to narrow the gap based on the current trajectory but to bypass it completely by aiming for what SPRIND terms the next S-curve—the upcoming architectural and paradigmatic leap beyond transformer-based systems.

      How does the competition operate?

      The €125 million is allocated across three phases with decreasing selection at each stage. In Stage 1, up to ten teams will each receive up to €3 million over seven months, with the main deliverable being initial technological proof points for their frontier hypothesis, which could include technical reports, preprints, experimental artifacts, or evidence of a new scaling dimension or emergent phenomena.

      Up to six teams can advance to Stage 2, where they will receive up to €8 million each over eight months, at which point the focus shifts to production-ready engineering processes, validated scaling dimensions, and the initial identification of what SPRIND refers to as ‘technical secrets,’ which are proprietary insights offering significant performance advantages.

      Up to three winners will then move on to Stage 3, obtaining up to €15.5 million each over nine months, with the objective of developing a working frontier system prototype, user-facing applications in testing, and an investment-quality data room prepared for the next capital round.

      The maximum total non-dilutive funding per team across all three phases amounts to €26.5 million, a substantial initial investment for a serious AI lab, although it pales in comparison to the amounts raised by frontier labs such as Anthropic or Mistral.

      The real incentive is what comes afterward: SPRIND has structured the program with a target of a €1 billion scale-up round for each winning lab following the 24-month competition, positioning that capital as equivalent to a mega Series A round in the US, propelling a ‘serious seed lab’ into becoming a ‘real frontier player.’

      The €1 billion sum is not included in the challenge budget; instead, the labs are expected to secure it from external investors. SPRIND’s Financing Workstream is designed to assist teams in creating investment-quality data rooms to support that fundraising effort.

      The architectural gamble

      The challenge is deliberately technology-agnostic regarding submission requirements but is explicit about what it does not want.

      SPRIND’s disqualification criteria provide insight: they exclude incremental transformer optimization that lacks fundamentally new capabilities, reproducing or deriving established models such as rebuilding OpenAI, Llama, or Qwen, incremental efficiency enhancements like improved quantization or streamlined MoE routing, conventional agent architectures without systemic innovation, domain-specific fine-tuning without foundational innovation, and purely brute-force scaling as the main innovation strategy.

      What is being sought is less straightforward to define, which SPRIND acknowledges candidly. Examples of potential directions include alternative model architectures (e.g., state-space models, energy-based transformers, diffusion LLMs, JEPA-style objectives, Titans architectures, or ‘entirely novel frameworks’), agentic systems with fundamentally new orchestration theories as opposed to conventional tool-use wrappers; embodied AI and world models; neuro-symbolic and hybrid methods; scientific foundation models for protein design, material science, or drug discovery; and innovative training paradigms that replace the pre-train plus RLHF framework.

      The S-curve framing is an intentional strategic decision. SPRIND’s argument, articulated in the challenge materials and statements from the agency’s leadership, holds that if European teams attempt to replicate the current generation of frontier labs with the accompanying European constraints, they will inevitably fall behind in terms of cost and speed.

      The current S-curve is dominated by extensive transformer and diffusion models, and the capital needed to compete at scale on that curve exceeds what European public funding can provide.

      However, transitioning to a new architectural paradigm, whatever form it takes, signifies a moment where early entry and accumulated knowledge outweigh capital