As demand for artificial intelligence soars, the race to secure reliable compute power has grown into one of the industry’s biggest bottlenecks. With cloud providers now facing months-long waitlists for high-end hardware, infrastructure costs have climbed to extremes. Amazon Web Services, for example, has been charging around 98 dollars per hour for an eight-GPU server equipped with Nvidia’s flagship H100 chips, while some decentralized platforms offer similar setups for just a few dollars an hour.
Mohitakshi Agrawal
Into this landscape comes Singularity Compute, the infrastructure arm of SingularityNET, which has begun phase one of its first enterprise-grade Nvidia GPU cluster in Sweden. The deployment is hosted in a modern Stockholm data centre operated by Conapto and powered entirely by renewable energy. The cluster uses next-generation Nvidia hardware including H200 and L40S GPUs, marking a significant addition to the European AI compute market.
Anna Ekman
Singularity’s cluster is built for high density performance. It supports traditional enterprise needs as well as the work of the Artificial Superintelligence Alliance, a decentralized AI ecosystem led by SingularityNET. Developers can rent full bare metal machines, launch GPU-backed virtual machines, or access dedicated API endpoints for inference. This flexibility allows teams to train full-scale models, fine-tune existing ones, or run heavy inference workloads such as generative AI applications.
Operations will be managed by Cudo Compute, a cloud provider and Nvidia partner. Cudo is responsible for maintaining enterprise-level reliability and supporting customers building mission-critical AI systems. Dr. Ben Goertzel, founder of SingularityNET and co-chair of the ASI Alliance, said the launch represents a meaningful step toward building open and globally accessible AI. He noted that as AI moves closer to general intelligence, the question of who can access powerful compute will shape the future of the field.
Singularity Compute CEO Joe Honan echoed that view, describing the Swedish cluster as more than extra capacity. He framed it as part of a broader push to redefine how AI infrastructure is delivered, with a focus on openness, security, and digital sovereignty.
The new cluster will anchor ASI:Cloud, Singularity’s AI inference service built with Cudo. ASI:Cloud gives developers wallet-based access to an OpenAI-compatible API that can scale from serverless tasks to dedicated GPU servers. According to Singularity Compute, early customers are already running workloads on the system, and more hardware deployments across new regions are planned.
The launch arrives at a moment when investment in AI infrastructure is surging. This year alone, more than one trillion dollars worth of data centre projects have been announced worldwide. Governments are getting involved too, with France recently unveiling a national plan exceeding 100 billion euros to grow its AI infrastructure footprint.
Not every company or country has the resources to build large-scale data centres, which has helped drive interest in decentralized GPU networks and distributed compute models. In the coming years, control over compute capacity may matter as much as control over data once did.
Ryan Browne
Singularity Compute’s Swedish cluster reflects a broader push to make high-performance AI resources more accessible and globally distributed. As the competition for compute heats up, efforts like this highlight how the industry is shifting toward more open and flexible infrastructure models.
