Lyceum

Active

Overview

Lyceum is a German AI infrastructure startup founded in 2025 that provides GPU cloud services and hosted inference for open-source models. The company positions itself as a sovereign European AI infrastructure provider, emphasizing European data residency, usage-based billing, and alternatives to large hyperscale cloud providers.

Its platform covers serverless model inference, dedicated GPU virtual machines, cloud notebooks, and larger training or inference deployments. The company is notable for combining open-source model hosting with per-token inference pricing and per-second GPU billing, while operating from a Berlin-centered business with infrastructure distributed across European data centers.

History

Lyceum was founded in 2025 by Magnus Grünewald and Maximilian Niroomand. Grünewald serves as co-founder and CEO, with prior experience in energy and infrastructure. The company was established as a Berlin- and Zurich-based startup focused on sovereign European GPU infrastructure and open-source AI inference. It later secured €10.3 million in reported pre-seed funding to expand its cloud platform and European data-center operations.

Product Lines

Product Line	Positioning	Price Range
Serverless inference	Per-token hosting for open-source AI models	Approximately $0.06 - $0.20 per 1 million tokens
Dedicated GPU compute	On-demand virtual machines for training and inference	Approximately $1.19 - $3.59 per GPU hour
Cloud notebooks	Interactive GPU environments for development	Approximately $0.39 - $5.89 per GPU hour
Managed inference endpoints	Dedicated production serving for open-source models	Usage-based; custom pricing may apply
Large-scale GPU clusters	Cluster infrastructure for demanding training workloads	Quote-based

Manufacturing

Lyceum does not manufacture hardware. It operates a cloud infrastructure platform using third-party NVIDIA GPUs hosted in European data centers; public materials identify locations including the Netherlands, Germany, Finland, France, and the United States, with availability varying by product.

Notable Products

Lyceum Inference - Serverless inference service for running open-source language models with usage-based, per-token billing.
GPU VMs - Dedicated GPU virtual machines for fine-tuning, training, and production inference, with per-second billing.
Cloud Notebooks - Browser-based GPU notebook environments for interactive AI development and experimentation.
Inference Studio - Managed catalog of open-source models exposed through usage-based inference endpoints.

Reputation

Lyceum is perceived as an early-stage European alternative to hyperscale GPU clouds, particularly among teams interested in open-source models, data sovereignty, and transparent usage-based pricing. Its reported strengths include per-second billing, European hosting options, and access to dedicated NVIDIA GPUs. Because the company is young, its scale, geographic coverage, operational track record, and long-term pricing stability remain less established than those of major cloud providers. Publicly available information about employee count and customer adoption is limited.

Sources (10)