DEPLOYDatabase

Company

Replicate

Replicate is a San Francisco AI inference platform that lets developers run, fine-tune, and deploy machine-learning models with a single API call, and is best…

Founded
2019
HQ
San Francisco, California, United States
Status
active

The verified answer

As of Sep 2026, DEPLOY verifies Replicate: $57.5M raised. 4 sources back the record.

Overview

Replicate is a San Francisco AI inference platform that lets developers run, fine-tune, and deploy machine-learning models with a single API call, and is best known for Cog, its open-source packaging format that turns any model into a production-grade container with an auto-generated HTTP interface. Founded in 2019 by Ben Firshman (previously product lead on Docker Compose at Docker) and Andreas Jansson (ex-Spotify ML engineer), the company sits at the long-tail end of the neocloud spectrum: rather than selling raw GPU capacity by the hour like CoreWeave or Lambda, it fans out per-second, scale-to-zero inference across a marketplace of thousands of community-contributed models, including image generators (Flux, Imagen), video (Seedance, Grok Imagine), and open language models.\n\nInfrastructure-wise Replicate is not a datacenter operator.

It brokers Nvidia GPU capacity (T4, L40S, A100 80GB, H100, H200, and multi-GPU 2x/4x/8x configurations) from underlying cloud providers and bills customers per second of run time, with published rates starting at $0.81 per hour for a T4 and $5.49 per hour for a single H100 or H200. Its commercial pitch is speed to production for individual developers and product teams: bring a model, wrap it in Cog, get an autoscaling endpoint.

That positioning is what attracted roughly $57M in venture funding across a 2021 seed and a 2022 Series B led by Andreessen Horowitz, with participation from Sequoia and Y Combinator.\n\nIn November 2025 Replicate announced it was being acquired by Cloudflare, folding the model catalog and Cog runtime into Cloudflare's Workers AI stack. Terms were not disclosed.

The team continues to operate the replicate.com product under the Cloudflare umbrella, and the About page now routes hiring through Cloudflare careers, marking the effective end of Replicate's independent-company phase.

Verified record

Verified deployments
None on file
Active incidents
None on file

Key facts

Positioning

Inference-focused model marketplace and serverless deployment platform; long-tail developer segment rather than hyperscale training

Cog runtime

Author of Cog, an open-source containerization format for ML models that generates a production HTTP API and is now the packaging layer for the platform

Announced hourly pricing

Nvidia T4 $0.81/hr; L40S $3.51/hr; A100 80GB $5.04/hr; H100 $5.49/hr; H200 $5.49/hr. Multi-GPU 2x/4x/8x priced proportionally; 8x H100/H200 available via committed spend

Billing granularity

Per-second billing with scale-to-zero for public models; private Cog deployments charge for all online time unless configured as fast-booting fine-tunes

Silicon partnerships

Nvidia-only fleet (T4, L40S, A100, H100, H200). No disclosed AMD, Cerebras, Intel, or custom-silicon capacity

Datacenter footprint

None owned or disclosed. Replicate brokers GPU capacity from underlying cloud providers rather than operating datacenters, so it is a software-layer neocloud rather than a facility operator

Funding history

~$17.8M seed in 2021; ~$40M Series B in December 2022 led by Andreessen Horowitz, with Sequoia and Y Combinator participating. No priced round disclosed after 2022

Acquisition by Cloudflare

Announced November 17, 2025 on the Replicate blog as "Replicate is joining Cloudflare." Financial terms were not disclosed. Hiring now routes through the Cloudflare careers page

Founders

Ben Firshman (previously Docker Compose product lead) and Andreas Jansson (previously Spotify ML engineer), co-founded in 2019

Model catalog

Thousands of community-contributed models spanning image generation (Flux, Imagen, GPT image), video (Seedance, Grok Imagine, Happy Horse), music (Minimax Music), and open language models

Public / private status

Private company through Nov 2025; now a wholly-owned unit of Cloudflare (NYSE: NET)

GPU count

Not disclosed. Replicate has never published fleet size and, as a broker of underlying cloud capacity, likely does not own a fixed inventory

Capacity in MW

Not applicable; no owned datacenter or colocation footprint disclosed

Data & sources

Web sources

4

4 sources backing this record.View all →

Replicate on the deployment map

View the global map

Replicate has no verified deployments on the DEPLOY registry yet. Explore the deployment map by place and type.

Free to use on your site, with a link back to DEPLOY.

No verified deployments for Replicate on the DEPLOY record yet. Track Replicate to know the moment the first one is verified.

Current leadership (2)

Safety record

No incidents on record for Replicate.

Only active incidents are counted. Retracted incidents are excluded from this summary but remain reachable at their canonical URLs.

Full safety record: incidents, sourcing, and exposure data →

Recent coverage

Replicate in third-party press