AI Infrastructure Clouds

CoreWeave

Roseland, United States

CoreWeave (NASDAQ: CRWV) is one of the largest specialized AI cloud computing providers, founded and headquartered in Roseland, New Jersey. Founded by Michael Intrator, Brannin McBee, and Brian Venturo (originally as a cryptocurrency mining operation), CoreWeave operates over 30 hyperscale data centers powered by hundreds of thousands of Nvidia H100, H200, and Blackwell B200 GPUs serving major AI customers including Microsoft, Meta, OpenAI, IBM, and Mistral AI. The company went public in March in one of the largest tech IPOs of the year and has signed multi-billion dollar capacity deals including a major partnership with Meta to support its AI infrastructure expansion through.

Fireworks AI

Redwood City, United States

Fireworks AI is a leading production-grade AI inference platform, founded and headquartered in Redwood City, California. Founded by Lin Qiao (former leader of PyTorch infrastructure at Meta), Dmytro Dzhulgakov, and Pawel Garbacki, Fireworks provides ultra-fast inference for over 100 open-source models including Llama, Mistral, DeepSeek, Qwen, and proprietary FireFunction models for function calling. The platform offers serverless GPU inference, dedicated deployments, on-demand fine-tuning, and the FireOptimizer for inference optimization. Fireworks serves customers including Cresta, Cursor, Notion, Quora, Uber, and Verizon with sub-second inference latency. The company raised 52 million Series B at a 552 million valuation.

Lambda

San Jose, United States

Lambda is a leading AI cloud infrastructure provider specializing in GPU compute for AI training and inference, founded and headquartered in San Jose, California. Founded by Stephen Balaban and Michael Balaban, Lambda operates Lambda Cloud providing on-demand and reserved Nvidia H100, H200, and Blackwell B200 GPU instances at competitive pricing, plus the Lambda Stack development environment, Lambda 1-Click Clusters, and the Hyperplane line of on-premise GPU servers. Lambda serves over 5,000 research institutions, AI startups, Fortune 500 enterprises, and government agencies including Stanford, MIT, Microsoft, and Sony. The company has raised over 320 million in funding at a 2.5 billion valuation.

Modal Labs

New York, United States

Modal is a serverless cloud infrastructure platform optimized for AI and machine learning workloads, founded and headquartered in New York City. Founded by Erik Bernhardsson (former CTO of Better and creator of the Annoy library at Spotify), Modal provides a Python-native serverless GPU compute platform with sub-second cold starts, autoscaling from zero to thousands of GPU containers, instant model deployment, batch inference, scheduled jobs, and the Modal Sandboxes for running untrusted code. Customers include Suno, Substack, Ramp, ScaleAI, and Cursor for AI inference, fine-tuning, batch data processing, and AI agent execution. Modal raised 80 million Series B at a 1.1 billion valuation from Lux Capital, Redpoint, and Definition Capital.

Replicate

San Francisco, United States

Replicate is a leading AI model hosting and inference platform that lets developers run open-source machine learning models with a single API, founded and headquartered in San Francisco, California. Founded by Ben Firshman and Andreas Jansson, Replicate operates a marketplace of thousands of community-hosted AI models including FLUX, SDXL, Whisper, Llama, and many specialized image, video, audio, and language models. The platform features Cog (an open-source containerization tool for ML), automatic GPU autoscaling, billing only for compute time used, and the Replicate API serving millions of developers building AI applications without managing infrastructure or model deployment.

Together AI

San Francisco, United States

Together AI is a leading AI cloud platform specializing in open-source model training, fine-tuning, and inference, founded and headquartered in San Francisco, California. Founded by Vipul Ved Prakash, Ce Zhang, Chris Re, and Percy Liang, Together AI operates the Together GPU Cloud, the Together Inference platform serving over 200 open-source models including Llama, DeepSeek, Mistral, and Qwen, the Together Fine-Tuning service, and the Together Custom Models program for enterprise model development. The company has built one of the largest dedicated GPU clusters for open-source AI development and raised over 528 million in funding at a 3.3 billion valuation from Salesforce Ventures, NVIDIA, and Lux Capital.

Vast.ai

Los Angeles, United States

Vast.ai is a decentralized GPU rental marketplace connecting AI developers with affordable GPU compute from a global network of providers, founded and headquartered in Los Angeles, California. The platform allows users to rent on-demand GPU instances from individual operators, data centers, and crypto mining facilities at significantly lower prices than hyperscale clouds, supporting Nvidia consumer GPUs (RTX 4090, 5090), enterprise GPUs (H100, H200), and AMD Instinct chips. Vast.ai features real-time bidding, persistent storage, instant template deployment, Jupyter notebook integration, and automated GPU price comparison. Popular among AI researchers, indie ML developers, and cost-conscious AI startups seeking budget GPU compute.