machine0 is a CLI-first cloud VM platform by Fundamental Software (YC S26) for developers and AI researchers who need persistent, reproducible VMs. Nine sizes run from $0.013/hr (small, 1 vCPU) to $3.714/hr (6xl, 60 vCPU) across US, UK, EU, and Asia with per-minute billing, suspend/resume, NixOS support, and optional H100/H200 GPUs. No free tier; $5 minimum top-up, unused credits refunded.
machine0 is a CLI-first cloud VM service by Fundamental Software (YC S26) offering 9 KVM/QEMU instance sizes from $0.013/hr (1 vCPU, 1 GB RAM) to $3.714/hr (60 vCPU, 240 GB RAM), billed per minute across US, UK, EU, and Asia regions. Suspended VMs cost $0.078/GB/month for image storage only. NixOS flake provisioning and optional H100, H200, L40S, and MI300X GPUs are included. Uptime SLA is 99.99%.
Frequently Asked Questions
What is machine0 and what does it offer?
machine0 is a cloud VM platform operated by Fundamental Software (Y Combinator S26, San Francisco) that gives developers persistent, real KVM/QEMU virtual machines controlled entirely from the command line. Founded in 2026 by Barnaby Malet, previously co-founder and CTO of Upflow (YC W20), machine0 targets the gap between heavyweight cloud consoles and ephemeral container platforms. The platform offers 9 named VM sizes from a 1 vCPU small instance to a 60 vCPU 6xl, covering $0.013/hr to $3.714/hr with per-minute billing across US, UK, EU, and Asia. Every VM gets a static public IP and an HTTPS endpoint at the vm.mac0.io subdomain automatically, with no NAT, tunnels, or reverse proxy configuration required. machine0 has first-class NixOS support with flake provisioning and also supports Ubuntu with Ansible, eliminating environment drift between development and production. Pre-installed tooling includes Claude Code, Codex, OpenClaw, and Hermes, and a remote MCP server lets AI agents drive VM lifecycle programmatically. VMs can be suspended to pause compute billing, costing only $0.078/GB/month for image storage until resumed.
How much does machine0 cost in 2026?
machine0 uses per-minute billing funded via a Stripe wallet, with a $5 minimum top-up and fully refundable unused credits. There are 9 named sizes: small (1 vCPU, 1 GB RAM, 25 GB) at $0.013/hr; medium (2 vCPU, 2 GB, 60 GB) at $0.034/hr; large (2 vCPU, 4 GB, 80 GB) at $0.052/hr; xl (4 vCPU, 8 GB, 160 GB) at $0.104/hr; xxl (8 vCPU, 16 GB, 320 GB) at $0.208/hr; xxxl (16 vCPU, 64 GB, 200 GB) at $0.825/hr; 4xl (32 vCPU, 128 GB, 480 GB) at $1.980/hr; 5xl (48 vCPU, 192 GB, 720 GB) at $2.970/hr; 6xl (60 vCPU, 240 GB, 900 GB) at $3.714/hr. Pricing is identical across all four regions (US, UK, EU, Asia) with no regional surcharge. Suspended VMs stop compute billing and cost only $0.078/GB/month for image storage. GPU instance pricing (H100, H200, L40S, MI300X) is not listed on the public pricing page and requires account creation to view. There is no free tier.
Where are machine0 data centers located?
machine0 operates across four regions: United States, United Kingdom, European Union, and Asia, with pricing identical across all four. The 99.99% uptime SLA applies to all regions, which is stronger than Hetzner's 99.9% guarantee. Every VM in any region gets a static public IP and an HTTPS endpoint at the vm.mac0.io subdomain that resolves globally. For latency-sensitive workloads, the UK and EU regions serve Western Europe, while the Asia region serves Southeast Asia and East Asia traffic. Specific datacenter cities within each region are not published on the public website as of June 2026. The four-region footprint is smaller than AWS (30+ regions) or Hetzner (7 named locations), but it covers all major developer geographies. Geographic expansion is expected as Fundamental Software scales beyond its YC S26 launch.
Does machine0 offer GPU servers?
Yes, machine0 offers optional GPU instances on top of its 9 standard VM sizes. Supported GPU models include the NVIDIA L40S (48 GB VRAM, suited to inference and mixed rendering workloads), NVIDIA H100 (80 GB VRAM, for large-scale AI training), NVIDIA H200 (141 GB VRAM, for very large memory inference and training), and AMD MI300X (192 GB VRAM, for LLM fine-tuning on AMD hardware). GPU instances share the same per-minute billing and suspend/resume model as standard VMs, so you pay only for active compute time and can freeze a GPU session between runs. GPU pricing is not published on the public pricing page and requires account creation to view exact per-hour rates. Pre-installed agents including Claude Code and OpenClaw can drive GPU-backed VMs via the MCP server, enabling AI agents to run training or inference jobs autonomously. This makes machine0 one of the few cloud VM providers where an AI agent can provision and run GPU workloads on the same machine without human intervention.
What are the best alternatives to machine0?
Fly.io is the closest developer-focused alternative, offering persistent Fly Machines with a global network and a simple deploy model, but it uses Firecracker microVMs rather than full KVM instances and lacks first-class NixOS support. Modal is an alternative for GPU-heavy AI workloads but is serverless and ephemeral, which suits batch jobs but not long-horizon agents that need a persistent OS environment between runs. Hetzner Cloud is significantly cheaper for always-on workloads (CX22 starts at about $4.15/month for 2 vCPU and 4 GB RAM) but has no CLI-first persistent VM model, NixOS flake provisioning, or pre-installed agent tooling. Lambda Labs and RunPod offer GPU cloud access with published per-hour rates but neither provides a full KVM OS environment, suspend billing, or NixOS support. DigitalOcean Droplets are more polished for web app hosting with a large managed services catalog, but bill hourly to a monthly cap with no per-minute granularity. For developers who specifically need NixOS VMs or AI agent infrastructure with a terminal-first workflow, machine0 has no direct equivalent in 2026. Teams needing managed databases, Kubernetes, or a CDN alongside their VMs should evaluate Hetzner or DigitalOcean instead.
Who is machine0 best for?
machine0 is best for developers running long-horizon AI agents that need a persistent, GPU-optional environment: an autonomous coding agent debugging a complex codebase for four to eight hours needs the VM to stay alive and accessible throughout. Backend engineers who prefer the terminal over cloud consoles and want a stable, static-IP VM they can SSH into without clicking through an AWS wizard are the second core audience. NixOS practitioners who need reproducible cloud environments provisioned from a flake.nix will find first-class support that no other cloud provider matches at this level in 2026. AI researchers who want Claude Code, Codex, or OpenClaw on a clean VM without manual tool installation benefit directly from the pre-installed agent environment. machine0 is not suitable for teams who need managed databases, Kubernetes, CDN, or a full cloud services portfolio from a single provider. Enterprises requiring SOC 2 or ISO 27001 compliance certifications should wait for Fundamental Software to publish its compliance posture, which was not confirmed at launch. Windows-only shops and teams relying on Azure AD or domain-joined infrastructure have no supported path on machine0.
How do you get started with machine0?
Visit machine0.io and sign up for an account, which requires adding funds to a Stripe wallet with a minimum $5 top-up (fully refundable if unused). After signing up, install the machine0 CLI with a single curl command provided in the onboarding documentation. Run machine0 new my-vm to provision your first VM, selecting one of the 9 sizes (small through 6xl), a base image (NixOS or Ubuntu), and whether you want an optional GPU. The VM spins up with a static public IP and an HTTPS endpoint at the vm.mac0.io subdomain within seconds. SSH into the VM directly using the provided key, or run machine0 shell my-vm to open an interactive terminal session from your local machine. To use NixOS flakes, include your flake.nix in the provisioning step and machine0 applies it automatically on first boot for a fully declarative environment. To suspend billing, run machine0 suspend my-vm; to resume from the same state, run machine0 start my-vm.
How does machine0 compare to Hetzner Cloud in 2026?
Hetzner Cloud is a European infrastructure provider with 20+ years of history, seven global locations, and a full managed services catalog including Kubernetes, object storage, managed databases, load balancers, and CDN, while machine0 is a YC S26 startup launched in 2026 with four regions and no managed services beyond raw VMs. Hetzner's CX22 costs about $4.15/month for 2 vCPU and 4 GB RAM billed hourly to a monthly cap, versus machine0's medium at $0.034/hr ($25/month) or large at $0.052/hr ($38/month) for similar specs with per-minute billing. Hetzner is cheaper for always-on workloads running full months; machine0 is cheaper for intermittent workloads where suspend billing drops costs to $0.078/GB/month during idle periods. Hetzner offers GPU instances (RTX 4000 Ada, A40) while machine0 offers H100, H200, L40S, and MI300X, covering more current AI training hardware. For NixOS VMs and AI agent infrastructure, machine0 is clearly ahead: Hetzner does not offer first-class NixOS images, pre-installed agent tooling, or an MCP server for programmatic VM control. Choose Hetzner for production web apps, managed databases, or cost-sensitive long-running services; choose machine0 for AI agent infrastructure, NixOS-first environments, or intermittent GPU workloads where per-minute billing reduces cost.
Top Alternatives
- Hetzner Cloud: Pick machine0 if you need NixOS VMs, per-minute suspend billing, or AI agent infrastructure; pick Hetzner if you need managed databases, Kubernetes, or cheaper always-on servers from $4.15/month.