Thunder Compute
Thunder Compute is a San Francisco-based GPU cloud provider offering on-demand NVIDIA RTX A6000, A100 80GB, and H100 PCIe instances at 80% below AWS. It targets ML engineers, AI researchers, and AI startups using proprietary GPU-over-TCP virtualization and IDE integrations.
- Company typePrivate
- Founded2022
- HeadquartersSan Francisco, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Thunder Compute does
Thunder Compute is a San Francisco-based private cloud GPU provider that offers on-demand access to NVIDIA RTX A6000, A100 80GB, and H100 PCIe instances at prices it advertises as 80% below AWS. The company operates as a Y Combinator W25-backed startup that emerged from approximately four years of stealth research in March 2026 to commercialize its GPU-over-TCP virtualization technology — a novel architecture in which GPUs sit across a high-speed network rather than PCIe slots, communicating over TCP while appearing local to the virtual machine.
The core product, Thunder Cloud, is accessed through a self-serve web console, a `tnr` command-line tool, official IDE extensions for VS Code/Cursor/Windsurf, pre-configured templates for Ollama (LLM inference), ComfyUI (image generation), Unsloth (fine-tuning), PyTorch, and CUDA Kernels, plus a REST API and an MCP server that lets AI agents like Claude Code and Codex provision GPUs programmatically. Supporting features include instance snapshots, port forwarding with DDoS-protected HTTPS URLs, sole-tenant GPU allocation with memory reset between users, and LLM-friendly error messages. Underlying GPU-over-TCP allows GPUs to detach from servers when idle and transparently reattach when needed, with ~10-20ms initial connection latency.
Thunder Compute runs a product-led go-to-market targeting individual ML engineers, AI researchers, PhD students, and small AI startups, with usage-based per-minute billing and no minimum commitments. A referral program pays 3% lifetime credits on referred customer spend. The team is small (1-10 employees) with systems and infrastructure pedigree from Citadel Securities, Aquatic, and AWS, and the company raised $4.5M in seed funding led by Matrix Partners in early 2026.
Thunder Compute firmographics
Firmographics- Name
- Thunder Compute
- Legal name
- Thunder Compute
- Website
- https://thundercompute.com
- Company type
- Private
- Founded year
- 2022
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Thunder Compute is a San Francisco-based GPU cloud provider offering on-demand NVIDIA RTX A6000, A100 80GB, and H100 PCIe instances at 80% below AWS. It targets ML engineers, AI researchers, and AI startups using proprietary GPU-over-TCP virtualization and IDE integrations.
- Ownership category
- akta.pro rank
Thunder Compute industry classification
Industry- Product category
- GPU Cloud Infrastructure
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)
- SIC
- Services-Computer Programming, Data Processing, Etc. (7370)
- akta.pro industry
- AI Compute Cloud & GPU-as-a-ServiceHardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers)Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
Keywords
Where Thunder Compute is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Thunder Compute business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Infrastructure, Technology or R&D, Personnel, Marketing or Sales, Operations
Revenue model
- GPU Instance Rental: Usage-based rental of GPU compute instances billed per minute while running. Customers pay per GPU-hour based on GPU type (RTX A6000, A100, H100). No minimum commitments, no upfront costs. Preloaded credit never expires and is consumed before auto-pay charges.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Usage-based | Pay-as-you-go | RTX A6000 On-demand GPU |
| Usage-based | Pay-as-you-go | A100 80GB On-demand GPU (1-8 GPUs) |
| Usage-based | Pay-as-you-go | H100 PCIe On-demand GPU (1-8 GPUs) |
| Other | Pay-as-you-go | Referral Program |
Go-to-market motion1 record
Distribution channels4 records
Marketing channels5 records
Thunder Compute product offering
Product offeringCore offering
Thunder Compute provides an on-demand GPU cloud platform that gives developers and AI/ML teams instant access to NVIDIA RTX A6000, A100 80GB, and H100 PCIe instances via proprietary GPU-over-TCP network-attached virtualization. Customers spin up dedicated, sole-tenant GPU servers in seconds and are billed per minute, with pricing positioned at roughly 80% less than hyperscalers such as AWS. The service is consumed through a web console, a tnr CLI, IDE extensions (VS Code, Cursor, Windsurf), and a REST API, and ships with pre-built templates for common AI workflows (Ollama, ComfyUI, Unsloth, PyTorch).
Product overview
Thunder Compute is a GPU cloud platform providing on-demand access to NVIDIA GPUs (RTX A6000, A100 80GB, H100 PCIe) through GPU-over-TCP technology for AI/ML workloads. The core product Thunder Cloud is accessed via multiple interfaces: the tnr CLI for terminal users, VS Code/Cursor/Windsurf extensions for integrated cloud development, and a web console for visual management. Pre-built templates (base, Ollama, ComfyUI, Unsloth) accelerate common AI workflows. Additional features include instance snapshots, port forwarding with HTTPS, a referral program, and REST API access for programmatic management. The platform targets developers and enterprises seeking cost-effective GPU compute, emphasizing 80% savings versus hyperscalers with per-minute billing.
Differentiator
Problem solved
Functional benefit
Products and services
- Thunder Cloud On-demand GPU cloud service providing dedicated, sole-tenant access to NVIDIA RTX A6000, A100 80GB, and H100 PCIe instances via GPU-over-TCP network-attached virtualization. Targets ML engineers, AI developers, and AI/ML teams needing affordable GPU compute for training, fine-tuning, inference, and research workloads, billed per minute with no minimum commitments.
- tnr CLI Standalone command-line interface that lets developers create, connect to, modify, snapshot, transfer files to, and tear down Thunder Compute GPU instances from a terminal. Available for Windows, macOS, and Linux; targeted at developers who prefer terminal workflows over the web console.
Quantifiable outcome
- 80% cost savings vs AWS for equivalent GPU instances
- +3 more outcomes
Companies that use Thunder Compute
Customer profileNamed customers6 records
Segments3 records
Ideal customer profiles2 records
Thunder Compute technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration10 records
AI capability5 records
Feature5 records
Thunder Compute partnerships and signals
Strategic signalScale indicators2 records
Recent moves4 records
Expansion highlights6 records
Thunder Compute competitors and assessment
Company assessmentDirect peers
- CoreWeave: CoreWeave is a leading specialized GPU cloud provider offering NVIDIA GPU compute (H100, A100, etc.) optimized for AI workloads, directly comparable to Thunder Compute in product category and target market. It is significantly larger and more capitalized, serving as the principal reference competitor.
- Lambda Labs: Lambda provides on-demand GPU cloud instances and GPU workstations specifically priced for AI/ML developers and researchers, with a similar cost-leadership positioning to Thunder Compute. Direct overlap in customer segment (individual developers, AI startups) and pricing-led value proposition.
- RunPod: RunPod offers on-demand and spot GPU cloud instances plus serverless GPU compute for AI/ML developers with self-serve signup and competitive pricing. Highly comparable developer-first GTM and per-hour GPU rental model.
- Vast.ai: Vast.ai operates a marketplace model for renting GPU compute (including consumer and data-center GPUs) at discounted rates versus hyperscalers, directly competing with Thunder Compute on affordability for AI training and inference workloads.
- Modal: Modal provides serverless compute for AI/ML workloads with developer-friendly deployment, auto-scaling, and per-second billing. Directly comparable developer-experience focus and PLG motion targeting ML engineers.
- Together AI: Together AI offers a cloud platform for training, fine-tuning, and serving open-source AI models with GPU-backed infrastructure, including per-token inference APIs. Directly comparable in serving AI developers and startups with cost-optimized GPU compute.
- Paperspace (DigitalOcean): Paperspace (acquired by DigitalOcean) provides GPU-powered virtual machines and notebooks targeted at developers and ML practitioners. Similar offering portfolio (GPU VMs, notebooks, containers) targeting the same developer persona.
- Fluidstack: Fluidstack provides large-scale GPU cloud infrastructure for AI training and inference, with significant enterprise focus. Comparable GPU cloud product category and AI workload specialization.
Broad incumbents
- AWS EC2 (P-series GPU instances): AWS offers GPU-accelerated EC2 instances (P4, P5, G5) at premium pricing and is the principal incumbent whose pricing Thunder Compute benchmarks against. Broader portfolio across general compute, storage, and services, but the explicit competitive reference point for the company.
Emerging players
- Crusoe Energy: Crusoe builds large-scale GPU data centers powered by stranded/renewable energy, providing AI cloud compute with cost and sustainability differentiation. Comparable AI compute infrastructure focus, though at much larger scale and with different go-to-market.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks7 records
Key highlights7 records
Customer concentration
Thunder Compute social profiles
Digital presenceThunder Compute financial estimates
Financial estimateRevenue estimate
Valuation estimate
Thunder Compute leadership team
Management profileNumber of profiles
Profiles2 records
Thunder Compute funding detail
Funding detailFunding overview
Funding rounds5 records
Investors9 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Thunder Compute M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Thunder Compute
What does Thunder Compute do?
Thunder Compute provides an on-demand GPU cloud platform that gives developers and AI/ML teams instant access to NVIDIA RTX A6000, A100 80GB, and H100 PCIe instances via proprietary GPU-over-TCP network-attached virtualization. Customers spin up dedicated, sole-tenant GPU servers in seconds and are billed per minute, with pricing positioned at roughly 80% less than hyperscalers such as AWS. The service is consumed through a web console, a tnr CLI, IDE extensions (VS Code, Cursor, Windsurf), and a REST API, and ships with pre-built templates for common AI workflows (Ollama, ComfyUI, Unsloth, PyTorch).
Is Thunder Compute a public or private company?
Thunder Compute is a private company. It is classified as venture growth investor backed and is currently operating.
When was Thunder Compute founded?
Thunder Compute was founded in 2022. It employs 11 to 50 people.
Where is Thunder Compute based?
Thunder Compute is headquartered in San Francisco, United States, in the North America region.
How does Thunder Compute make money?
One revenue line is on record: GPU Instance Rental.
Who are Thunder Compute's main competitors?
Direct peers on record are CoreWeave, Lambda Labs, RunPod, Vast.ai, Modal, Together AI, Paperspace (DigitalOcean) and Fluidstack. AWS EC2 (P-series GPU instances) is listed as a broad incumbent. Crusoe Energy is listed as an emerging player.
Does Thunder Compute have an API?
Yes. REST API for programmatic GPU instance management. Endpoints include instance creation (POST /instances/create), listing (GET /instances/list), modification (POST /instances/{id}/modify), deletion (POST /instances/{id}/delete), SSH key management (POST /instances/{id}/add_key), and port forwarding (PATCH /instances/{id}/ports). Authentication via Bearer token (API key). API base URL: https://api.thundercompute.com:8443/v1/. Developer documentation is at www.thundercompute.com/docs/api-reference.
What industry is Thunder Compute in?
Thunder Compute's product category is GPU Cloud Infrastructure. Its primary akta.pro industry code is HDAAAAAK, AI Compute Cloud & GPU-as-a-Service, with a secondary code of HDAAAAAG, AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers). Its NAICS code is 5182 and its SIC code is 7370.