RAW
RAW provides API-first dedicated CPU and NVIDIA GPU servers for AI workloads, allowing customers to create, scale, resize, rebuild and destroy infrastructure over REST instead of navigating a hyperscaler console. Its stated goal is to make dedicated AI infrastructure simpler and substantially cheaper than AWS by combining bare-metal hardware, transparent flat pricing and unlimited bandwidth.
At a Glance
- AI labs and model developers
- Startups and SaaS companies
- Developers and indie builders
- Production inference fleets and AI-agent operators
- +2 more
AI Tools by RAW
(1)
RAW
Bare Metal Cloud for AI Workloads
Discussions
No discussions yet
Be the first to start a discussion about RAW
Latest News
Products & Services
A dedicated bare-metal CPU and NVIDIA GPU infrastructure service. Users select a server type and region, then provision with a REST API; servers include root SSH access, NVMe storage and unlimited bandwidth.
Bearer-token JSON API for signup, billing setup, server deployment, resizing, rebuilding, metrics, volumes, SSH keys and destruction. The product page describes 24 REST endpoints.
Open-source-oriented command-line interface distributed through npm, with commands for initialization, deployment, listing, SSH, logs, restart, start/stop, rebuild, resize and removal.
Web interface for metrics, graphs, networking, volumes, backups, snapshots, terminal access, rescaling, rebuilding and power management; it wraps the same servers managed by the API.
Market Position
RAW positions itself as a simpler, lower-cost alternative to AWS, SageMaker, Bedrock and other hyperscaler GPU offerings by using dedicated bare metal, one API, root access, flat pricing and no egress fees. Its published comparisons also position it against DigitalOcean, Hetzner, Fly.io, Render, Railway, Vercel, GCP, Azure, Lambda Labs, CoreWeave, RunPod and Vast.ai. The trade-off it acknowledges is less built-in global distribution, autoscaling and managed-service breadth than hyperscalers or platforms such as Fly.io and AWS.
Leadership
Founders
Michael C. Jakob
Founder of RAW and builder of Eulerpool, a financial data and analytics platform. His biography says he studied computer science and management at MIT, ETH Zurich and Karlsruhe Institute of Technology, and previously worked at UBS and McKinsey & Company.
Executive Team
Michael C. Jakob
Founder
Built Eulerpool from zero to millions of monthly page views and led its migration from managed cloud infrastructure to bare metal; previously worked at UBS and McKinsey & Company and studied at MIT, ETH Zurich and Karlsruhe Institute of Technology.
Founding Story
RAW began as an internal infrastructure tool for Eulerpool after the financial-data platform's monthly cloud bill exceeded $7,000. The team migrated from managed cloud services to dedicated bare metal, reducing the reported monthly cost from $7,370 to $460, and turned the provisioning and operations playbook into a product: a 100% API for creating and scaling servers. The initial vision was to offer the economics and control of real hardware without the operational complexity of doing the migration manually.
Business Model
Revenue Model
Monthly flat-price sales of dedicated CPU and GPU servers, with per-second billing described for server usage; RAW also charges for add-ons such as persistent NVMe volumes and additional IPv4 addresses. The company describes its model as buying hardware, automating provisioning and selling access with a thin transparent margin.
Pricing Tiers
2 vCPU, 4 GB RAM, 40 GB NVMe; EU (Germany and Ireland); unlimited bandwidth.
4 vCPU, 8 GB RAM, 80 GB NVMe; EU; unlimited bandwidth.
8 vCPU, 16 GB RAM, 160 GB NVMe; EU; unlimited bandwidth.
16 vCPU, 32 GB RAM, 320 GB NVMe; EU; unlimited bandwidth.
20 GB VRAM inference GPU configuration; dedicated NVIDIA GPU with CUDA.
96 GB VRAM training configuration, 24 CPU cores, 256 GB RAM and 1.9 TB NVMe.
96 GB VRAM maximum configuration, 24 CPU cores, 768 GB RAM and 15 TB NVMe.
Target Markets
- AI labs and model developers
- Startups and SaaS companies
- Developers and indie builders
- Production inference fleets and AI-agent operators
- Enterprises seeking private or GDPR-oriented AI infrastructure
- Teams migrating from AWS, Fly.io, Vercel, DigitalOcean, Hetzner and other cloud/PaaS providers
- LLM inference with vLLM, Ollama or TGI
- AI model training and LoRA fine-tuning
- Always-on AI agents and worker fleets
- Vector databases such as Qdrant, pgvector and Milvus
- Self-hosted SaaS backends, APIs and web applications
- Databases, queues, CI/CD and general bare-metal workloads
- Eulerpool, Michael C. Jakob's financial data and analytics platform, is the internal origin and public migration case study for RAW.