{ "features_resolved": { "snapshot_any": true, "snapshot_mem": true, "fork": true, "pause_resume": true, "persistent_disk": true, "whole_disk": null, "persist_scope": null, "volumes": true, "auto_stop_idle": false, "long_sessions": true, "custom_image": true, "image_snapshot": true, "image_oci": false, "vm_isolation": true, "docker_inside": true, "nested_virt": false, "root": true, "systemd": true, "browser": false, "desktop": false, "computer_use": false, "code_interpreter": false, "browser_control": null, "desktop_control": null, "browser_and_desktop": false, "anti_bot": false, "captcha_solving": false, "residential_ip": false, "agent_harnesses": false, "ssh": true, "public_ipv4": false, "inbound_https": false, "custom_domain": true, "raw_tcp_inbound": true, "egress_allowlist": true, "open_internet": true, "static_egress_ip": true, "private_network": true, "self_host": true, "byoc": true, "open_source": true, "soc2": false, "hipaa": false, "sso": false, "gpu": false, "arm64": true, "windows": false, "macos": false, "lang_python": false, "lang_node": false, "lang_go": null, "lang_rust": null, "lang_java": null, "gpu_desktop": false, "credential_injection": null, "egress_http_rules": false, "wake_on_request": false, "live_resize": false, "fork_running": false, "memory_snapshot_fork": true, "webhooks": null, "scheduled_wakeups": null, "audit_logs": null, "eu_data_residency": true, "zero_data_retention": null, "scoped_api_keys": null, "spend_limits": null, "mcp_server": false, "usage_api": false, "snapshot_auto": false, "snapshot_on_demand": false, "harness_api": false, "own_agent_api": false, "fast_boot": null, "strong_isolation": true, "ingress_rules": true, "guest_firewall": true, "secret_proxy": false, "secret_proxy_any": false, "volume_attach": true, "volume_shared": true, "sidecar": null, "fuse": true, "kernel_features": true, "burst": null }, "company": { "id": "self-host-baremetal", "checked": "2026-09-30", "company": "N/A — DIY reference architecture (not a vendor/company)", "ownership": { "value": null, "parent": null, "source": "", "quote": "" }, "stage": { "value": "unknown", "source": "", "quote": "", "confidence": "low", "note": "This card ('self-host-baremetal') is not a company or product — it is a hypothetical DIY reference architecture (a modeled fleet of self-managed Firecracker microVMs on rented bare-metal servers from Hetzner, OVH, Latitude, and AWS). There is no legal entity, funding history, accelerator batch, or revenue to research for 'it'; the card itself says 'DIY infrastructure only... No managed service, support or operations included' and 'not a managed-product offer or full operating budget.' Funding/backing questions apply to the underlying hardware vendors (Hetzner is a bootstrapped/private German company, OVHcloud is a public company on Euronext Paris, Latitude.sh and AWS are respectively a funded startup and an Amazon subsidiary) — not to this synthetic sandbox-fleet construct, which has no ownership, funding, or revenue of its own." }, "accelerators": [], "total_funding_usd": { "value": null, "status": "unknown" }, "last_round": { "type": null, "amount_usd": null, "date": null, "leads": [], "status": "unknown", "source": "", "quote": "" }, "investors": [], "portfolio_listings": [], "revenue": { "value": null, "status": "unknown" }, "employees": { "value": null, "source": "" }, "note": "This card is a hardware-cost floor model, not a real vendor: it computes what it would cost to run your own Firecracker microVM sandbox fleet on rented dedicated servers from Hetzner, OVH, Latitude, and AWS (bare-metal), explicitly excluding ops labour, support, and any managed control plane ('OPS LABOUR EXCLUDED. Hardware-only theoretical floor, not a managed-product offer or full operating budget.'). Buyers evaluating 'who is behind it' should instead evaluate the underlying hardware suppliers individually: Hetzner AG (privately held, bootstrapped/profitable, Germany), OVHcloud (public company, Euronext Paris, ticker OVH), Latitude.sh (funded bare-metal-as-a-service startup), and AWS EC2 Metal (an Amazon.com, Inc. subsidiary product). None of those funding/ownership facts apply to a 'self-host-baremetal' entity because no such entity exists — it is purely a pricing construct in this dataset." }, "id": "self-host-baremetal", "name": "Self-hosted bare-metal sandbox fleet (hardware floor)", "url": "https://www.hetzner.com/dedicated-rootserver/ax42/", "category": "self-host", "isolation": "firecracker", "modes": [ { "key": "hetzner-ax42-p50", "label": "Hetzner AX42 / 50% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "eu" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.017220898683078554, "ram_gib_h": 0.005535288862418106, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 14, "max_ram_gib": 57.6, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 3 simultaneous4vCPU/8GiB/50GB guests per host and 50% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "hetzner-ax42-p70", "label": "Hetzner AX42 / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "eu" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.012300641916484683, "ram_gib_h": 0.003953777758870076, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 14, "max_ram_gib": 57.6, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 3 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "hetzner-ax42-p90", "label": "Hetzner AX42 / 90% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "eu" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.00956716593504364, "ram_gib_h": 0.0030751604791211693, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 14, "max_ram_gib": 57.6, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 3 simultaneous4vCPU/8GiB/50GB guests per host and 90% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "hetzner-ex44-p70", "label": "Hetzner EX44 / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "eu" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.0055092316855271, "ram_gib_h": 0.0017708244703479963, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 18, "max_ram_gib": 57.6, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 4 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "hetzner-ax102-p70", "label": "Hetzner AX102 / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "eu" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.013461102129952638, "ram_gib_h": 0.004326782827484776, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 30, "max_ram_gib": 115.2, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 7 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "hetzner-ax162-p70", "label": "Hetzner AX162 / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "eu" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.01603668566890723, "ram_gib_h": 0.005154648965005894, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 94, "max_ram_gib": 115.2, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 14 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "ovh-rise1-p70", "label": "OVHcloud RISE-1 US / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.011291443319436174, "ram_gib_h": 0.0036293924955330555, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 10, "max_ram_gib": 28.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 2 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "latitude-c3-month-p70", "label": "Latitude c3.small monthly / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": null, "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.028290649195949976, "ram_gib_h": 0.009093422955841061, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 10, "max_ram_gib": 28.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 2 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "latitude-c3-hour-p70", "label": "Latitude c3.small hourly,730h / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": null, "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.056521739130434796, "ram_gib_h": 0.01816770186335404, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 10, "max_ram_gib": 28.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 2 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "latitude-m4-month-p70", "label": "Latitude m4.metal.medium monthly / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": null, "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.019399302305794265, "ram_gib_h": 0.006235490026862442, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 30, "max_ram_gib": 115.2, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 7 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "aws-c7i-metal-24xl-p70", "label": "AWS c7i.metal-24xl PAYG / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.046722964349527774, "ram_gib_h": 0.015018095683776782, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 94, "max_ram_gib": 172.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 21 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "aws-m7i-metal-24xl-p70", "label": "AWS m7i.metal-24xl PAYG / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.04790017349871819, "ram_gib_h": 0.0153964843388737, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 94, "max_ram_gib": 345.6, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 23 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "aws-c6g-metal-p70", "label": "AWS c6g.metal PAYG / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": true }, "pricing": "resource", "vcpu_h": 0.03735152726963329, "ram_gib_h": 0.012005848050953555, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 62, "max_ram_gib": 115.2, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 14 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "aws-m7g-metal-p70", "label": "AWS m7g.metal PAYG / 70% occupancy", "flags": [ "alt" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": true }, "pricing": "resource", "vcpu_h": 0.04116867182846934, "ram_gib_h": 0.013232787373436571, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 62, "max_ram_gib": 230.4, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 15 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "aws-c7i-metal-1y-p70", "label": "AWS c7i.metal24xl 1y RI / 70% occupancy", "flags": [ "alt", "sales" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.0317112666269605, "ram_gib_h": 0.010192907130094443, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 94, "max_ram_gib": 172.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 21 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit.", "cpu_class": "dedicated" }, { "key": "ovh-rise1-12m-p70", "label": "OVHcloud RISE-1 US 12-month upfront / 70% occupancy", "flags": [ "alt", "commit" ], "os": [ "linux" ], "regions": [ "us" ], "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.009902162478471654, "ram_gib_h": 0.0031828379395087455, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 10, "max_ram_gib": 28.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": "$798 paid upfront for 12 months ($66.50/month), installation $0 (OVH US catalog upfront12).", "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 2 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit. Yearly term: $798 paid upfront for 12 months ($66.50/month), installation $0 (OVH US catalog upfront12).", "commit_term": "1-year", "cpu_class": "dedicated" }, { "key": "latitude-c3-year-p70", "label": "Latitude c3.small paid yearly / 70% occupancy", "flags": [ "alt", "commit" ], "os": [ "linux" ], "regions": null, "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.019803454437164983, "ram_gib_h": 0.006365396069088742, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 10, "max_ram_gib": 28.8, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": "$1,596/yr ($133/month) vs $190 monthly (latitude.sh/pricing).", "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 2 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit. Yearly term: $1,596/yr ($133/month) vs $190 monthly (latitude.sh/pricing).", "commit_term": "1-year", "cpu_class": "dedicated" }, { "key": "latitude-m4-year-p70", "label": "Latitude m4.metal.medium paid yearly / 70% occupancy", "flags": [ "alt", "commit" ], "os": [ "linux" ], "regions": null, "gpu_only": false, "features": { "arm64": false }, "pricing": "resource", "vcpu_h": 0.013578235344167446, "ram_gib_h": 0.004364432789196679, "cpu_basis": "alloc", "ram_basis": "alloc", "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": 2, "vcpu_options": null, "min_vcpu": 1, "max_vcpu": 30, "max_ram_gib": 115.2, "sizes": [], "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": "$3,830/yr ($319.17/month) vs $456 monthly (latitude.sh/pricing).", "min_billed_seconds": 0, "granularity_s": 1, "start_fee": 0, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": {}, "note": "Derived allocation, NOT provider per-second billing. Calibrated to 7 simultaneous4vCPU/8GiB/50GB guests per host and 70% temporal occupancy. Host rent + setup/12 + specifiedIP/disk included; ops labour, separate control plane and spare hosts excluded. Other guest sizes are linearized approximations; use integer fleet model for invoices. No CPU overcommit. Yearly term: $3,830/yr ($319.17/month) vs $456 monthly (latitude.sh/pricing).", "commit_term": "1-year", "cpu_class": "dedicated" } ], "storage": { "included_disk_gib": 50, "disk_gib_month": null, "disk_billed_when_stopped": true, "snapshot_gib_month": null, "paused_ram_billed": false }, "network": { "egress_free_gib": null, "egress_gib": null, "ipv4_month": null, "ipv4_available": false, "internet": { "default": "full" } }, "plans": [ { "name": "DIY infrastructure only", "fee": 0, "flags": [], "fee_is_credit": false, "included_usd": 0, "per_seat": 0, "concurrency": null, "max_session_h": null, "max_vcpu": null, "max_ram_gib": null, "sizes_allowed": null, "note": "No managed service, support or operations included. Host-levelIP cost already included; per-guest publicIPv4 not included." } ], "free": { "monthly_credit": 0, "one_time_credit": 0 }, "features": { "linux": true, "windows": false, "macos": false, "arm64": true, "gpu": false, "gpu_types": [], "snapshot": "mem", "pause_resume": true, "fork": true, "persistent_disk": true, "volumes": true, "auto_stop_idle": false, "max_session_h": null, "custom_image": true, "docker_inside": true, "nested_virt": false, "root": true, "systemd": true, "browser": false, "desktop": false, "computer_use": false, "code_interpreter": false, "agent_harnesses": false, "ssh": true, "public_ipv4": false, "inbound_https": false, "custom_domain": true, "raw_tcp_inbound": true, "egress_allowlist": true, "static_egress_ip": true, "private_network": true, "self_host": true, "byoc": true, "open_source": true, "regions": [ "eu", "us", "asia", "other" ], "soc2": false, "hipaa": false, "sso": false, "lang_python": false, "lang_node": false, "lang_go": null, "lang_rust": null, "lang_java": null, "gpu_desktop": false, "credential_injection": null, "egress_http_rules": false, "wake_on_request": false, "live_resize": false, "fork_running": false, "memory_snapshot_fork": true, "webhooks": null, "scheduled_wakeups": null, "audit_logs": null, "eu_data_residency": true, "zero_data_retention": null, "scoped_api_keys": null, "spend_limits": null, "mcp_server": false, "snapshot_auto": false, "snapshot_on_demand": false, "harness_api": false, "egress_allowlist_level": "ip", "ingress_rules": true, "guest_firewall": true, "secret_proxy": false, "volume_attach": true, "volume_shared": true, "isolation": "firecracker", "image_oci": false, "anti_bot": false, "captcha_solving": false, "egress_ip_type": "datacenter", "own_agent_api": false, "max_session_unlimited": true, "open_internet": true, "usage_api": false }, "perf": { "cpu_runs_s": null, "cold_start_ms": null, "burst_tti_ms": null, "dax_s": null }, "knobs": [], "caveats": [ "OPS LABOUR EXCLUDED. Hardware-only theoretical floor, not a managed-product offer or full operating budget.", "Packing50/70/90% is temporal paid-slot occupancy AFTER host reserves and integer4/8/50GB packing. The synchronized daytime workload alone reaches only about24% occupancy; higher percentages require backfill.", "Rates allocate ONE host cost between CPU andRAM using CPUshare0.2016/0.3312; they are not independently purchasable CPU/RAM prices. Other guest sizes require repacking.", "Setup amortized over12months; first-month cash larger. Reserve2 hardwarethreads,10%RAM,10%disk per host; these are scenario assumptions, not measurements.", "No separate control plane, N+1 standby, multi-region, durable snapshot storage or network overage in headline rates. Co-locating control plane requires enough host reserve.", "50GiB snapshots are not automatically free: local leftover disk is non-durable on host loss; optional1TB StorageBox4USD/month can hold the example off-host. AWS EBS snapshots0.05USD/GB-month are an external storage scenario, not a Firecracker-managed product.", "Primary hostIPv4 included once in hardware model, NOT oneIP per sandbox. Guests need NAT/proxy/IPv6 or paidIP allocations.", "Hetzner1Gbps traffic unlimited policy; OVH regional plan and Latitude20TB allowances differ; AWS egress billed after account allowance. Provider-global network fields intentionally null to avoid cross-mode false free egress.", "OnlyLinuxKVM-capable hardware assumed. gVisor is userspace-kernel isolation, not a full guest VM; Kata/CloudHypervisor have different overhead profiles. Firecracker guest nested virtualization not offered by this model.", "No performance benchmark equivalence: SMT threads, hybridEcores, Arm cores and dedicated physicalcores differ. CPUovercommit2x table is optimistic and may hurt tail latency.", "Hetzner primary IPv4 is a separate add-on (live_data ipPrice $1.90/month per server); the card's derived Hetzner host costs (AX42 $119 + $59/12 = $123.92) do not add it. If the headline price excludes it, add $1.90/host/month (+1.5% AX42, +2.6% EX44). Product pages still list '1 Dedicated IPv4 address' under basic features, so inclusion is ambiguous." ], "sources": [ "https://www.hetzner.com/_resources/app/data/app/live_data_en_USD.json", "https://api.us.ovhcloud.com/1.0/order/catalog/public/eco?ovhSubsidiary=US", "https://www.latitude.sh/pricing", "https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonEC2/current/us-east-1/index.csv", "https://www.hetzner.com/storage/storage-box/", "https://aws.amazon.com/ebs/pricing/", "https://aws.amazon.com/vpc/pricing/", "https://aws.amazon.com/ec2/pricing/on-demand/", "https://github.com/firecracker-microvm/firecracker", "https://github.com/e2b-dev/infra", "https://github.com/superradcompany/microsandbox", "https://github.com/kata-containers/kata-containers", "https://github.com/google/gvisor", "https://github.com/cloud-hypervisor/cloud-hypervisor" ], "fleet": { "default_max_instances": null, "default_max_vcpu": null, "default_max_gpus": null, "raise": null, "raise_time_days": null, "verification": "Fraud checks may request government-issued ID or advance payment with own credit card; unpaid invoices/order risk may block purchases.", "tiers": null, "notes": "DIY reference architecture, not a managed account or sandbox quota. Guest VM count depends on purchased hosts, packing and operator configuration. Hetzner ordering/verification applies to the priced hardware; do not confuse host core count with guest fleet quota.", "sources": [ "https://docs.hetzner.com/general/security-and-identify/fraud-prevention-faq/", "https://www.hetzner.com/dedicated-rootserver/ax42/" ], "researched": "2026-09-28" }, "audit_2026_09_29": { "reviewed": 5 }, "feature_evidence": { "snapshot_auto": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Rescue-System VNC installation Image installer Traffic statistics Reverse DNS administration Domain Registration Robot DDoS protection Firewall", "note": "Hetzner/OVH/Latitude dedicated-server product pages list no automatic guest state save-on-stop/idle/timeout feature; nothing in their basic/optional feature lists mentions checkpoint-and-restore. An idle-stop hook calling Firecracker's snapshot API would be code the operator writes themselves, not a provider feature.", "checked": "2026-09-30" }, "snapshot_on_demand": { "value": false, "source": "https://github.com/firecracker-microvm/firecracker/blob/main/docs/snapshotting/snapshot-support.md", "quote": "In order to make restoring possible, Firecracker snapshots save the full state of the following resources: the guest memory, the emulated HW state (both KVM and Firecracker emulated HW).", "note": "Not a contradiction once the card's own model is read: Hetzner/OVH/Latitude/AWS bare-metal rentals have no first-party server snapshot product at all (the card explicitly notes this). snapshot='mem' and fork=true describe the *self-built* layer this DIY reference architecture assumes you would code yourself by calling the Firecracker VMM's own snapshot API (which does support on-demand full/diff snapshots of guest memory+disk) - but that API call is your own automation, not something Hetzner/OVH/Latitude/AWS sell as a managed on-demand-snapshot feature. So snapshot_on_demand=false is correct for 'is this offered to you by the provider', consistent with snapshot='mem' describing what the self-built layer is capable of.", "checked": "2026-09-30" }, "harness_api": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Full root access", "note": "Hetzner/OVH/Latitude/AWS sell bare compute (root access to a server) with no coding-agent-harness product whatsoever; any harness API would be entirely self-built.", "checked": "2026-09-30" }, "egress_allowlist": { "value": true, "source": "https://docs.hetzner.com/robot/dedicated-server/firewall/", "quote": "On the customer administration interface Robot, you can use the firewall feature to define your own filtering settings for incoming and outgoing traffic.", "note": "Hetzner Robot stateless firewall (free), per physical server, IP/port/protocol rules; guests only get distinct treatment if they have their own additional IPs. OVH Edge Network Firewall handles incoming only; Latitude Firewall is an agent managing UFW/iptables on the host itself (not off-box). Real per-guest egress control (default-deny, domain allowlists via proxy) is self-built on the host with nftables, which a root agent inside a Firecracker guest cannot undo.", "checked": "2026-09-29", "level": "ip" }, "egress_http_rules": { "value": false, "source": "https://docs.hetzner.com/robot/dedicated-server/firewall/", "quote": "With our stateless firewall, or static firewall, data packets are not unpacked; rather, the head of each individual packet is inspected", "note": "Provider firewalls are L3/L4 packet filters. HTTP-level rules would be a self-built egress proxy.", "checked": "2026-09-29" }, "ingress_rules": { "value": true, "source": "https://docs.hetzner.com/robot/dedicated-server/firewall/", "quote": "If you activate the firewall before entering any firewall rules, you will block all incoming traffic.", "note": "Hetzner Robot firewall incoming rules (IP/port); OVH Edge Network Firewall per public IP ('only handles network traffic coming from outside the OVHcloud network'); Latitude host-agent firewall with inbound/outbound rules. Per-guest exposure, auth'd preview URLs etc. are self-built (host reverse proxy / NAT).", "checked": "2026-09-29" }, "guest_firewall": { "value": true, "source": null, "quote": null, "note": "Firecracker/Cloud Hypervisor microVMs with root and their own kernel on hardware you control; nftables/iptables inside guests work (guest kernel config permitting). You also have root on the host itself.", "checked": "2026-09-29", "derived": true }, "secret_proxy": { "value": false, "source": "https://docs.hetzner.com/robot/dedicated-server/firewall/", "quote": "With our stateless firewall, or static firewall, data packets are not unpacked; rather, the head of each individual packet is inspected", "note": "Provider network features are plain L3/L4 packet firewalls; there is no credential-injecting egress proxy product. A secret proxy for the guests would have to be self-built on the host.", "checked": "2026-09-30" }, "volume_attach": { "value": true, "source": "https://www.latitude.sh/docs/storage/block-storage.md", "quote": "Block storage gives you persistent, high performance volumes for bare metal servers.", "note": "Latitude block storage over NVMe/TCP, mapped to one host (not to guests directly). Hetzner dedicated servers have no attachable block volumes (Hetzner Cloud Volumes only), only Storage Box network mounts; guest-level volumes are self-built (e.g. extra Firecracker drives on host storage).", "checked": "2026-09-29" }, "volume_shared": { "value": true, "source": "https://www.latitude.sh/docs/storage/filesystem.md", "quote": "File storage gives you persistent filesystems that several servers can mount at the same time.", "note": "Latitude file storage = NFSv3/v4, read-write, private preview. Hetzner Storage Box: SMB/CIFS mountable read-write from several hosts ('Up to 10 simultaneous connections per Storage Box account'), on a single shared host server, not mirrored. Shared between hosts; exposing to guests (virtiofs/NFS re-export) is self-built.", "checked": "2026-09-29" }, "isolation": { "value": "firecracker", "source": "https://github.com/firecracker-microvm/firecracker", "quote": "Firecracker is an open source virtualization technology that is purpose-built for creating and managing secure, multi-tenant container and function-based services.", "note": "Card models a DIY Firecracker-on-bare-metal fleet (Hetzner/OVH/Latitude/AWS metal), matching the card's own top-level isolation field.", "checked": "2026-09-30" }, "image_oci": { "value": false, "source": "https://raw.githubusercontent.com/firecracker-microvm/firecracker/main/docs/snapshotting/snapshot-support.md", "quote": "A Firecracker microVM snapshot can be used for loading it later in a different Firecracker process, and the original guest workload is being simply resumed.", "note": "Firecracker boots a kernel + raw rootfs image via its own boot-source/drives API, not a Docker/OCI image directly; converting an OCI image into a bootable rootfs (e.g. via firecracker-containerd/ignite tooling) would be self-built.", "checked": "2026-09-30" }, "anti_bot": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope (this is a DIY bare-metal/microVM hardware fleet, not a browser/stealth product)", "checked": "2026-09-30" }, "captcha_solving": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope (bare-metal hardware rental, no browser/CAPTCHA product)", "checked": "2026-09-30" }, "egress_ip_type": { "value": "datacenter", "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Available Locations HEL 1 FSN 1", "note": "Rented datacenter hardware (Hetzner/OVH/Latitude/AWS metal) with datacenter-assigned IPv4/IPv6; no residential/mobile IP option.", "checked": "2026-09-30" }, "fork_running": { "value": false, "source": "https://raw.githubusercontent.com/firecracker-microvm/firecracker/main/docs/snapshotting/snapshot-support.md", "quote": "To create a snapshot, first you have to pause the running microVM and its vCPUs ... all vCPU threads are paused, and all device emulation stops.", "note": "Firecracker requires pausing the microVM before a snapshot (fork) can be taken; there is no clone-while-running path.", "checked": "2026-09-30" }, "memory_snapshot_fork": { "value": true, "source": "https://raw.githubusercontent.com/firecracker-microvm/firecracker/main/docs/snapshotting/snapshot-support.md", "quote": "A Firecracker microVM snapshot can be used for loading it later in a different Firecracker process, and the original guest workload is being simply resumed.", "note": "Full snapshots capture guest memory plus device/vCPU state, so a fork resumed from one keeps processes running; this is a self-built capability on top of the Firecracker API, not a managed provider feature (derived).", "checked": "2026-09-30" }, "wake_on_request": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "No provider product wakes a stopped guest on incoming request; a wake-on-request proxy would be self-built on top of the host/Firecracker API.", "checked": "2026-09-30" }, "auto_stop_idle": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "Rented dedicated hardware and self-run microVMs are not stopped by the provider on idleness; an idle-stop hook would be self-built.", "checked": "2026-09-30" }, "live_resize": { "value": false, "source": "https://raw.githubusercontent.com/firecracker-microvm/firecracker/main/src/firecracker/swagger/firecracker.yaml", "quote": "operationId: putMachineConfiguration ... Machine Configuration Parameters", "note": "PutMachineConfiguration (vCPU count / mem_size_mib) is a pre-boot configuration call; Firecracker does not support changing CPU/RAM of a running microVM.", "checked": "2026-09-30" }, "own_agent_api": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope (bare hardware/microVM fleet, no first-party task/browser/computer-use agent)", "checked": "2026-09-30" }, "mcp_server": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope; no MCP server is shipped for this DIY hardware/Firecracker fleet", "checked": "2026-09-30" }, "lang_python": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope. This is bare hardware / a self-built rootfs image, with no provider-supplied language runtime preinstalled; you must build your own guest image.", "checked": "2026-09-30" }, "lang_node": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope. No provider-supplied Node.js runtime; you must build your own guest image.", "checked": "2026-09-30" }, "soc2": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope. This card models DIY hardware rental (Hetzner/OVH/Latitude/AWS metal); no SOC 2 attestation for this self-built sandbox layer was found, and compliance of the guest environment is the operator's own responsibility.", "checked": "2026-09-30" }, "hipaa": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope. No BAA/HIPAA-eligible offering found for bare-metal rental; HIPAA compliance of the self-built sandbox stack is the operator's own responsibility.", "checked": "2026-09-30" }, "sso": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "not offered: outside the product's scope. No SAML/OIDC SSO found for the Hetzner Robot/console account used to rent this hardware, and there is no managed sandbox platform login to apply SSO to.", "checked": "2026-09-30" }, "eu_data_residency": { "value": true, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Available Locations HEL 1 FSN 1", "note": "AX42 hardware is only offered in Hetzner's EU datacenters (Falkenstein, Germany and Helsinki, Finland), so machines/data can be kept in the EU by construction.", "checked": "2026-09-30" }, "max_session_h": { "value": "none", "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AX42 ... Dedicated Root Server", "note": "Rented dedicated/bare-metal server billed monthly/hourly until cancelled; there is no provider-imposed maximum session length, matching the card's own max_session_h: null (unbounded) for a DIY fleet you control.", "checked": "2026-09-30" }, "nested_virt": { "value": false, "source": "https://github.com/firecracker-microvm/firecracker/blob/main/FAQ.md", "quote": "Firecracker supports Linux host and guest operating systems as well as OSv guests.", "note": "Firecracker's own docs describe only Linux/OSv guest support with a minimal 6-device virtio model; there is no mention of exposing /dev/kvm or nested-virtualization support to the guest, and Firecracker microVMs are not designed to run a second layer of virtualization inside them. The card's own caveats state this directly: 'Firecracker guest nested virtualization not offered by this model.'", "checked": "2026-09-30" }, "desktop": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Full root access 1 Dedicated IPv4 address optional IPv6 subnet (/64) Administration software Webmin available for LAMP 24h reset service via web interface Rescue-System VNC installation Image installer", "note": "The product is bare dedicated hardware/DIY sandbox fleet: you get root SSH access and an OS installer; there is no managed graphical desktop/VNC-to-your-sandbox product (the 'VNC' listed is only for the Rescue-System OS installer, not a per-guest desktop feature). A desktop would have to be self-built (e.g. installing a VNC server yourself inside a guest).", "checked": "2026-09-30" }, "browser": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Available operating systems Debian Ubuntu Linux AlmaLinux Arch Linux Open SUSE Cent OS Windows", "note": "The product ships bare OS images only (Debian/Ubuntu/etc.); there is no managed/pre-installed browser offered as part of the server or the Firecracker microVM tooling. Installing and managing a browser inside a guest would be self-built.", "checked": "2026-09-30" }, "computer_use": { "value": false, "source": "https://github.com/firecracker-microvm/firecracker/blob/main/docs/design.md", "quote": "Firecracker is a new virtualization technology that enables customers to deploy lightweight micro Virtual Machines or microVMs... They provide a secure, trusted environment for multi-tenant services, while maintaining minimal overhead.", "note": "Firecracker (the isolation layer this card models) and the underlying bare-metal renters offer only a VM/API surface for compute isolation, not any computer-use/browser-driving API. No product; entirely self-built.", "checked": "2026-09-30" }, "inbound_https": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "1 Dedicated IPv4 address optional IPv6 subnet (/64)", "note": "The host itself gets one dedicated IPv4 (and optional IPv6), but there is no first-party per-sandbox HTTPS ingress/routing product for the individual Firecracker guests - the card's own caveat states 'Primary host IPv4 included once in hardware model, NOT one IP per sandbox. Guests need NAT/proxy/IPv6 or paid IP allocations.' Exposing a guest over HTTPS requires self-built NAT/reverse-proxy/TLS termination on the host.", "checked": "2026-09-30" }, "public_ipv4": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "1 Dedicated IPv4 address", "note": "Confirmed at host level (each Hetzner/OVH/Latitude server includes one IPv4), but that IP belongs to the physical host, not to individual Firecracker guest sandboxes; the card models a fleet of many guests per host, so per-sandbox public IPv4 is not included and would require the operator to build NAT/proxying or purchase extra IP blocks (Hetzner's 'Additional Single IP'/'Additional IP subnet' add-ons). public_ipv4=false is correct at the guest/sandbox scope this card evaluates.", "checked": "2026-09-30" }, "agent_harnesses": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Available operating systems Debian Ubuntu Linux AlmaLinux Arch Linux Open SUSE Cent OS Windows", "note": "Only base OS images are offered; no coding-agent harness (Claude Code, Codex, OpenCode, etc.) is preinstalled or offered as a template by Hetzner/OVH/Latitude/AWS.", "checked": "2026-09-30" }, "gpu": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AMD Ryzen™ 7 PRO 8700GE Simultaneous Multithreading RAM: 64 GB up to 64 GB DDR5 ECC Disk: 2x 512 GB NVMe", "note": "The specific SKUs priced in this card's modes (Hetzner AX42/EX44/AX102/AX162, OVH RISE-1, Latitude c3/m4, AWS c7i/m7i/c6g/m7g metal) are all CPU-only bare metal with no GPU in the spec sheet; Hetzner's separate GPU-Line was deliberately not selected as a mode, matching features.gpu_types=[] and gpu:{} on every mode.", "checked": "2026-09-30" }, "windows": { "value": false, "source": "https://github.com/firecracker-microvm/firecracker/blob/main/FAQ.md", "quote": "Firecracker supports Linux host and guest operating systems as well as OSv guests.", "note": "Although Hetzner's own AX42 order page does list 'Windows' as a selectable host OS for the bare-metal server itself, this card explicitly models a Firecracker-based sandbox fleet running on top of that hardware (isolation='firecracker', caveat 'Only Linux KVM-capable hardware assumed'), and Firecracker's own docs state it only supports Linux (and OSv) guest operating systems - no Windows guest support. So windows=false is correct for the sandboxed-guest product this card evaluates, even though the underlying bare-metal host could itself boot Windows.", "checked": "2026-09-30" }, "macos": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "AMD Ryzen™ 7 PRO 8700GE Simultaneous Multithreading", "note": "All priced hardware (Hetzner, OVH, Latitude, AWS metal instances) is x86_64 (or Arm Neoverse for the arm64 AWS modes); none of it is Apple hardware, and Firecracker/KVM cannot run macOS guests on non-Apple silicon. macOS is plainly outside what this product is.", "checked": "2026-09-30" }, "cpu_class": { "value": { "hetzner-ax42-p50": "dedicated", "hetzner-ax42-p70": "dedicated", "hetzner-ax42-p90": "dedicated", "hetzner-ex44-p70": "dedicated", "hetzner-ax102-p70": "dedicated", "hetzner-ax162-p70": "dedicated", "ovh-rise1-p70": "dedicated", "latitude-c3-month-p70": "dedicated", "latitude-c3-hour-p70": "dedicated", "latitude-m4-month-p70": "dedicated", "aws-c7i-metal-24xl-p70": "dedicated", "aws-m7i-metal-24xl-p70": "dedicated", "aws-c6g-metal-p70": "dedicated", "aws-m7g-metal-p70": "dedicated", "aws-c7i-metal-1y-p70": "dedicated", "ovh-rise1-12m-p70": "dedicated", "latitude-c3-year-p70": "dedicated", "latitude-m4-year-p70": "dedicated" }, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "Derived allocation, NOT provider per-second billing. Calibrated to 3 simultaneous 4vCPU/8GiB/50GB guests per host and 50% temporal occupancy. ... No CPU overcommit.", "note": "Every mode in this card is a DIY Firecracker-on-bare-metal fleet (Hetzner, OVH, Latitude, AWS metal instances) where the card's own allocation model explicitly assumes 'No CPU overcommit' (stated verbatim in each mode's note field): the modeled vCPU count per guest is packed 1:1 from the physical host's dedicated cores/threads, with no sharing of a core's cycles across guests beyond that fixed integer packing. Bare-metal rental itself gives the whole physical machine's cores exclusively to the renter (Hetzner/OVH/Latitude/AWS metal instances are by definition not shared with other customers), and the operator-controlled Firecracker layer on top allocates whole vCPUs (pinned host threads) to each guest rather than oversubscribing them. This applies uniformly across all 18 pricing modes (different hosts, occupancy percentages, and commit terms), since they all share this same DIY packing methodology and hardware model.", "checked": "2026-09-30" }, "usage_api": { "value": false, "source": "https://www.hetzner.com/dedicated-rootserver/ax42/", "quote": "DIY reference architecture, not a managed account or sandbox quota. Guest VM count depends on purchased hosts, packing and operator configuration.", "note": "granularity: account-only; cost: no; Self-host-baremetal is a DIY bare-metal fleet using Firecracker on rented hardware (Hetzner, OVH, Latitude, AWS). There is no provider-managed usage API for individual sandboxes. You rent hardware by the server, not by guest sandbox; usage metering of individual guests is self-built by the operator. The underlying hardware providers (Hetzner, OVH, etc.) bill monthly/hourly for the whole server, with no per-sandbox breakdown or API endpoint.", "confidence": "high", "checked": "2026-10-01" } }, "locations": [ { "country": "DE", "state": null, "city": "Falkenstein", "provider_region": "FSN1", "self_serve": true, "choosable": true }, { "country": "FI", "state": null, "city": "Helsinki", "provider_region": "HEL1", "self_serve": true, "choosable": true }, { "country": "US", "state": "NY", "city": "New York", "provider_region": "NY", "self_serve": true, "choosable": true }, { "country": "US", "state": "CA", "city": "Los Angeles", "provider_region": "LA", "self_serve": true, "choosable": true }, { "country": "US", "state": "CA", "city": "Silicon Valley", "provider_region": "SV", "self_serve": true, "choosable": true }, { "country": "GB", "state": null, "city": "London", "provider_region": "LON", "self_serve": true, "choosable": true }, { "country": "DE", "state": null, "city": "Frankfurt", "provider_region": "FRA", "self_serve": true, "choosable": true }, { "country": "SG", "state": null, "city": "Singapore", "provider_region": "SG", "self_serve": true, "choosable": true }, { "country": "AU", "state": null, "city": "Sydney", "provider_region": "SYD", "self_serve": true, "choosable": true }, { "country": "JP", "state": null, "city": "Tokyo", "provider_region": "TYO", "self_serve": true, "choosable": true } ], "locations_source": "https://www.hetzner.com/_resources/app/data/app/live_data_en_USD.json (Hetzner) and https://www.latitude.sh/pricing (Latitude)", "locations_quote": "Hetzner: 'FSN1' (Falkenstein, Germany), 'HEL1' (Helsinki, Finland); Latitude: 'NY', 'LA', 'Silicon Valley', 'London', 'Frankfurt', 'Singapore', 'Sydney', 'Tokyo'", "locations_checked": "2026-10-01" }