# Vultr Cloud GPU Official pricing: https://api.vultr.com/v2/plans?type=vcg Category: gpu-cloud · Isolation: vm ## Pricing regimes (raw) - **vcg-a16-2c-8g-2vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a16-2c-16g-4vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a40-1c-5g-2vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a16-3c-32g-8vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a40-2c-10g-4vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a40-4c-20g-8vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a40-6c-30g-12vram, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h [alt] - **vcg-a16-12c-128g-32vram, on-demand (×2 node)** (resource): $0/vCPU-h, $0/GiB-h [alt] ## Features Yes: snapshots, persistent disk, volumes, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, full VM (own kernel), Docker inside, SSH, public IPv4, HTTPS preview URLs, open internet, static egress IP, SOC 2, HIPAA, SSO, GPU, Python, Node.js, live resize, EU data residency, snapshots on demand, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables), extra volumes, shared volumes, FUSE mounts, kernel features (eBPF, mounts, cgroups) No: memory snapshots, fork/clone, pause/resume, idle auto-stop, your own Docker/OCI image, nested virtualization, browser, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, egress allowlist, self-hosting, BYOC, open source, arm64, Windows, macOS, HTTP method/path egress rules, wake on request, live fork (no pause), memory fork, MCP server, usage per sandbox (API), automatic snapshots, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/vultr-gpu.json → feature_evidence ## Company - Ownership: independent - Stage: series-d+ [source](https://hyperframeresearch.com/is-vultrs-333-million-gamble-a-winning-bet-in-the-ai-cloud/) - Accelerators: none found - Total raised: $333M (confirmed: Sole external financing round to date, reported by multiple sources including Wikipedia and HyperFRAME Research, sourced to a Vultr/BusinessWire press release.) [source](https://www.businesswire.com/news/home/20241218657527/en/Vultr-Completes-Financing-With-LuminArx-and-AMD-Ventures-at-3.5-Billion-Valuation-Accelerating-Growth-in-AI-Infrastructure) - Last round: growth financing, $333M, 2024-12, led by AMD Ventures, LuminArx Capital Management (confirmed) [source](https://hyperframeresearch.com/is-vultrs-333-million-gamble-a-winning-bet-in-the-ai-cloud/) - Investors: AMD Ventures, LuminArx Capital Management - Revenue: $37.2M (estimated: Growjo third-party estimate; no primary source confirms a revenue figure, and this estimate looks low relative to Vultr's stated $3.5B valuation and reported hundreds of thousands of customers, so treat with caution.) [source](https://growjo.com/company/Vultr) annual revenue, 2025 Vultr (legal entity The Constant Company, LLC) was founded in 2014 by David Aninowsky and was bootstrapped/profitable for roughly a decade before taking its first-ever outside capital: a $333M growth financing co-led by AMD Ventures and LuminArx Capital Management, closed December 2024, valuing the company at $3.5B. In 2025 it also secured $329M in separate credit financing (debt, not equity) from banks including Bank of America, Citi, and Goldman Sachs, and is building a >$1B, 50MW AMD GPU data center cluster in Ohio. No accelerator/incubator affiliation found. Revenue/employee figures here are unverified third-party (Growjo) estimates only. ## Caveats - Website blocked (403), use public Vultr plans API. Fractional GPU slices not equivalent to a full GPU. Listed plan may have no currently offered regions. API hourly vs monthly costs retained without assuming a cap policy. - Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free. - This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null. - GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards. - Modes with gpu=null are intentionally excluded, not free GPU modes. Never substitute 0 for a null rate. Flags alt indicate DIY/ML infrastructure, not the same thing as managed untrusted-code sandboxes. - Unknown fee, storage or bandwidth means full monthly total may be unavailable. Rates exclude taxes. - SKU check 2026-09-28 (API /v2/plans?type=all + vultr.com/pricing): Vultr sells NO single-GPU H100 VM - H100 is only the 8x HGX VM ($23.92/h in the API; website "starting at $2.300/GPU/hr") or 8x bare metal. 1x L40S ($1.671/h), 2x/4x L40S and 1x A100 80 GB PCIe ($2.397/h) exist but the API lists no deployable locations for any of them (flag stock). - Full-GPU VM rates are all-in (vCPU, RAM and local disk included) and encoded as the per-GPU price with vcpu_h = ram_gib_h = 0; disk beyond 0 GiB shows as unknown although each plan bundles 1.2-13 TB. - Billing caps: cloud GPU instances are billed up to 730 h/month, bare metal and other server types up to 672 h/month; stopped instances keep billing until destroyed. ## How this provider charges As of 2026-09-28. Cloud GPU virtual machines and GPU bare metal. Website blocked (403), use public Vultr plans API. Fractional GPU slices not equivalent to a full GPU. Listed plan may have no currently offered regions. API hourly vs monthly costs retained without assuming a cap policy. | Regime | When applicable | Billing | Numbers ($/GPU-hour unless slice) | Source | |---|---|---|---|---| | on-demand | vcg-a16-2c-8g-2vram ; count [1]; ewr,ord,atl,fra,sjc,nrt,sgp,blr | per-hour; min unknowns | 0.059; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a16-2c-16g-4vram ; count [1]; ewr,sjc,nrt,sgp,blr | per-hour; min unknowns | 0.118; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a40-1c-5g-2vram ; count [1]; ewr,lax | per-hour; min unknowns | 0.075; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a16-3c-32g-8vram ; count [1]; sjc,sgp,blr | per-hour; min unknowns | 0.236; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a40-2c-10g-4vram ; count [1]; No available locations returned | per-hour; min unknowns | 0.144; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a40-4c-20g-8vram ; count [1]; No available locations returned | per-hour; min unknowns | 0.288; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a40-6c-30g-12vram ; count [1]; lax | per-hour; min unknowns | 0.432; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a16-12c-128g-32vram ; count [2]; sjc | per-hour; min unknowns | 0.471; list; per_gpu_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a40-8c-40g-16vram ; count [1]; No available locations returned | per-hour; min unknowns | 0.575; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a40-12c-60g-24vram ; count [1]; No available locations returned | per-hour; min unknowns | 0.856; list; per_slice_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-l40s-16c-180g-48vram ; count [1]; No available locations returned | per-hour; min unknowns | 1.671; list; per_gpu_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a16-24c-256g-64vram ; count [4]; No available locations returned | per-hour; min unknowns | 0.471; list; per_gpu_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-a16-48c-496g-128vram ; count [8]; No available locations returned | per-hour; min unknowns | 0.470875; list; per_gpu_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-l40s-32c-375g-96vram ; count [2]; No available locations returned | per-hour; min unknowns | 1.671; list; per_gpu_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vcg-l40s-64c-750g-192vram ; count [4]; No available locations returned | per-hour; min unknowns | 1.671; list; per_gpu_hour | [source](https://api.vultr.com/v2/plans?type=vcg) | | on-demand | vbm-48c-1024gb-4-a100-gpu NVIDIA_A100; count [4]; No available locations returned | per-hour; min unknowns | 2.39725; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-48c-1024gb-4-a100-gpu NVIDIA_A100; count [4]; No available locations returned | per-hour; min unknowns | 2.39725; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-112c-2048gb-8-h100-gpu NVIDIA_H100; count [8]; No available locations returned | per-hour; min unknowns | 2.3; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-112c-2048gb-8-a100-gpu NVIDIA_A100_SXM; count [8]; ewr | per-hour; min unknowns | 1.49; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-64c-2048gb-8-l40-gpu NVIDIA_L40S; count [8]; No available locations returned | per-hour; min unknowns | 1.49; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-72c-480gb-gh200-gpu NVIDIA_GH200; count [1]; No available locations returned | per-hour; min unknowns | 1.99; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-256c-2048gb-8-mi300x-gpu AMD_MI300X; count [8]; No available locations returned | per-hour; min unknowns | 1.85; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-256c-3072gb-8-mi325x-gpu AMD_MI325X; count [8]; No available locations returned | per-hour; min unknowns | 2.0; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-256c-3072gb-8-b200-gpu NVIDIA_B200; count [8]; No available locations returned | per-hour; min unknowns | 3.2; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | | spot | vbm-256c-3072gb-8-mi355x-gpu AMD_MI355X; count [8]; No available locations returned | per-hour; min unknowns | 2.59; dynamic-snapshot; per_gpu_hour | [source](https://api.vultr.com/v2/plans-metal) | ## Gotchas - Website blocked (403), use public Vultr plans API. Fractional GPU slices not equivalent to a full GPU. Listed plan may have no currently offered regions. API hourly vs monthly costs retained without assuming a cap policy. - Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free. - This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null. - GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards. ## Worked example (required protocol workload) 4 vCPU / 8 GiB, 50 concurrent × 8 hours/day × 22 days = **8,800 instance-hours/month**. 30% CPU utilization does not reduce allocated GPU uptime. 50 GiB snapshots and 100 GiB egress are additional. The requested CPU-only workload has no GPU type/count, so its full GPU-provider total is **null**, not a fictitious CPU equivalent. Explicit GPU sensitivity example: add **1 × L40S** to each of those 50 workers at the vcg-l40s-16c-180g-48vram quoted shape. Compute component = 8,800 × $1.671 = **$14,704.8000**. CPU/RAM are included in that SKU (not independently resizable). This is compute-only, before platform fees, storage, egress, setup, minimum rounding, cold-start/idle tails and capacity quotas. 50 concurrent GPUs are not promised by a unit rate. ## Evidence scope Sources read are listed in the provider/features JSON. Raw pricing pages, relevant docs and API payloads are retained under `../raw/`. All unknown feature toggles remain null.