# TensorPool

> TensorPool (YC W25) is a San Francisco startup that turns cloud GPUs into a git-style command line. Instead of wrestling with SSH sessions, provider dashboards, and idle-instance billing, machine learning engineers describe a training job in a config file and run 'tp job push' - TensorPool spins up a multi-node cluster across partner clouds, streams logs, saves outputs, and shuts everything down, billing only for runtime. It advertises GPU access at roughly half the cost of major cloud providers, with H100s starting around $1.99/hr. Founded by three Stanford dormmates, the company was accepted into Y Combinator's Winter 2025 batch.

- **Founded:** 2024
- **Headquarters:** San Francisco, California, United States
- **Founders:** Tycho Svoboda (CEO & Co-Founder), Joshua Martinez (CTO & Co-Founder), Hlumelo Notshe (CPO & Co-Founder)
- **Team size:** 3-5
- **Products:** TensorPool CLI, TensorPool Jobs, On-Demand GPU Clusters
- **Notable:** Accepted into Y Combinator's Winter 2025 batch, Launched publicly in March 2025 with a CLI that reached PyPI distribution, Advertises GPU access at roughly half the cost of major cloud providers (H100s around $1.99/hr)

## Products & services

- **TensorPool CLI** — A git-style command line interface (pip install tensorpool) for running ML training jobs and accessing on-demand multi-node GPU clusters. Commands cover clusters, storage volumes, jobs, SSH, and account management.
- **TensorPool Jobs** — A job-based workflow where users define a training run in a tp.config.toml file and submit it with 'tp job push'. TensorPool provisions the cluster, runs the command, streams logs (tp job listen), saves and returns outputs (tp job pull), then deallocates - billing only for active runtime.
- **On-Demand GPU Clusters** — Multi-node clusters with NVIDIA H100, H200, and B200 GPUs (and L40S), paired with high-performance NVMe/NFS storage, scheduled automatically across partner compute providers at competitive prices.

## Achievements

- Accepted into Y Combinator's Winter 2025 batch
- Launched publicly in March 2025 with a CLI that reached PyPI distribution
- Advertises GPU access at roughly half the cost of major cloud providers (H100s around $1.99/hr)
- Shipped TensorPool Jobs, a git-style job workflow, in October 2025
- Published technical analyses on GPU price-performance (e.g., B200 vs H100 for large-scale training)

## Latest updates

- **2025-10** — Introduced TensorPool Jobs - a git-style GPU workflow where jobs are defined in tp.config.toml and pushed with 'tp job push', with runtime-only billing (announced Oct 31, 2025 by CEO Tycho Svoboda).
- **2025-03** — Public launch as a YC W25 company; reported ~$130K seed and initial GPU CLI availability.

## Links

- Website: https://tensorpool.dev
- LinkedIn: http://www.linkedin.com/company/tensorpool
- Twitter/X: https://x.com/TensorPool
- GitHub: https://github.com/tensorpool/tensorpool

---

Profile page: https://yespress.io/tensorpool-yc-w25
Published by YesPress — https://yespress.io
Last updated: 2026-07-30