# Crusoe

> An energy-first AI cloud that builds its own data centers and serves open models on a cluster-wide KV cache.

Canonical: https://www.subconscious.dev/providers/crusoe · By The Subconscious Team · Updated September 30, 2026

- Founded: 2018
- Example models: DeepSeek V4 Pro, GLM 5.3, Kimi K2.6
- Website: https://crusoe.ai

## Overview

Crusoe started in 2018 turning wasted natural gas into power for computing and has since become a vertically integrated AI infrastructure company: it sources energy, builds data centers and rents GPUs through Crusoe Cloud. It designed and built the Abilene, Texas campus behind the OpenAI and Oracle Stargate project, planned at 1.2 GW, and in March 2026 announced an adjacent 900 MW campus for Microsoft. On September 17, 2026 it closed the first part of a $3.9B Series F at a $30.9B post-money valuation, and it reports over 6 GW of contracted capacity. Crusoe Cloud lists GB200 NVL72, B200 and AMD MI355X by quote, with H100 at $3.90 and H200 at $4.29 per GPU-hour on demand.

Its managed AI platform, Intelligence Foundry, launched in late 2025 and offers three tiers: Serverless Inference with per-token pricing, Self-Serve Deployments billed per GPU-hour (H100 at $5.50, B200 at $9.65), and Tailored Deployments with SLAs. Models run on Crusoe's own engine with MemoryAlloy, a KV cache shared across the cluster with cache-aware routing, so a prefix computed on one node can be reused on another. Crusoe claims up to 9.9x faster time to first token and 5x throughput versus vLLM on prefix-heavy work. The serverless catalog covers DeepSeek, GLM, Kimi, Gemma, gpt-oss and Nemotron behind an OpenAI-compatible API, from $0.05 in and $0.20 out per million tokens. LoRA fine-tuning launched in July 2026.

## Upsides

- Cluster-wide KV cache reuse suits multi-turn chat, agents and long shared prefixes, with cached input billed well below list.
- Owns the full stack from power to GPUs, which gives it large capacity and newer NVIDIA and AMD hardware.
- One account covers serverless tokens, dedicated endpoints, fine-tuning and raw GPU clusters with Kubernetes or Slurm.

## Core use cases

- Agents and chat products that resend long, repeated context.
- Teams that want to fine-tune an open model and deploy it in one place.
- Large training or inference clusters on GB200 and B200 capacity.

## Downsides

- The serverless catalog is small, and GB200, B200 and MI355X instances need a sales conversation.
- Headline speed figures are Crusoe's own benchmarks against vLLM, measured on workloads with heavy prefix reuse.

## At a glance

| Attribute | Value |
|---|---|
| Model access | Open weights |
| Flagship models | DeepSeek V4, GLM 5.3, Kimi K2.6, Nemotron 3 |
| Speed | Up to 9.9x faster TTFT vs vLLM (vendor claim) |
| Price | $0.05–$1.74 in, $0.20–$4.40 out per 1M |
| Customization | Serverless LoRA fine-tuning |
| Deployment | Serverless, self-serve and tailored dedicated, raw GPUs |
| Long context | Varies by model; cluster-wide KV cache |

## FAQ

### What is Crusoe?

Crusoe started in 2018 turning wasted natural gas into power for computing and has since become a vertically integrated AI infrastructure company: it sources energy, builds data centers and rents GPUs through Crusoe Cloud. It designed and built the Abilene, Texas campus behind the OpenAI and Oracle Stargate project, planned at 1.2 GW, and in March 2026 announced an adjacent 900 MW campus for Microsoft. On September 17, 2026 it closed the first part of a $3.9B Series F at a $30.9B post-money valuation, and it reports over 6 GW of contracted capacity. Crusoe Cloud lists GB200 NVL72, B200 and AMD MI355X by quote, with H100 at $3.90 and H200 at $4.29 per GPU-hour on demand.

### What is Crusoe best for?

Agents and chat products that resend long, repeated context; Teams that want to fine-tune an open model and deploy it in one place; Large training or inference clusters on GB200 and B200 capacity.

### How much does Crusoe cost?

Crusoe pricing at a glance: $0.05–$1.74 in, $0.20–$4.40 out per 1M. Rates change often, so check Crusoe's pricing page before committing.

### How much context does Crusoe support?

Crusoe's long-context support: Varies by model; cluster-wide KV cache.

### What are the downsides of Crusoe?

The serverless catalog is small, and GB200, B200 and MI355X instances need a sales conversation; Headline speed figures are Crusoe's own benchmarks against vLLM, measured on workloads with heavy prefix reuse.

### What are the best alternatives to Crusoe?

Common alternatives include Subconscious, OpenAI, Anthropic, Google Vertex AI, Amazon Bedrock. Each has a head-to-head comparison with Crusoe on this site.

## Comparisons

- [Subconscious vs Crusoe](https://www.subconscious.dev/compare/subconscious-vs-crusoe.md)
- [OpenAI vs Crusoe](https://www.subconscious.dev/compare/openai-vs-crusoe.md)
- [Anthropic vs Crusoe](https://www.subconscious.dev/compare/anthropic-vs-crusoe.md)
- [Google Vertex AI vs Crusoe](https://www.subconscious.dev/compare/google-vertex-vs-crusoe.md)
- [Amazon Bedrock vs Crusoe](https://www.subconscious.dev/compare/aws-bedrock-vs-crusoe.md)
- [Together AI vs Crusoe](https://www.subconscious.dev/compare/together-ai-vs-crusoe.md)
- [Fireworks AI vs Crusoe](https://www.subconscious.dev/compare/fireworks-vs-crusoe.md)
- [Baseten vs Crusoe](https://www.subconscious.dev/compare/baseten-vs-crusoe.md)
- [Groq vs Crusoe](https://www.subconscious.dev/compare/groq-vs-crusoe.md)
- [Cerebras vs Crusoe](https://www.subconscious.dev/compare/cerebras-vs-crusoe.md)
- [DeepInfra vs Crusoe](https://www.subconscious.dev/compare/deepinfra-vs-crusoe.md)
- [Hugging Face Inference Providers vs Crusoe](https://www.subconscious.dev/compare/hugging-face-vs-crusoe.md)
- [Modal vs Crusoe](https://www.subconscious.dev/compare/modal-vs-crusoe.md)
- [Cloudflare Workers AI vs Crusoe](https://www.subconscious.dev/compare/cloudflare-workers-ai-vs-crusoe.md)
- [xAI vs Crusoe](https://www.subconscious.dev/compare/xai-vs-crusoe.md)
- [Mistral AI vs Crusoe](https://www.subconscious.dev/compare/mistral-ai-vs-crusoe.md)
- [DeepSeek vs Crusoe](https://www.subconscious.dev/compare/deepseek-vs-crusoe.md)
- [Moonshot AI vs Crusoe](https://www.subconscious.dev/compare/moonshot-ai-vs-crusoe.md)
- [Z.ai vs Crusoe](https://www.subconscious.dev/compare/z-ai-vs-crusoe.md)
- [Alibaba Cloud vs Crusoe](https://www.subconscious.dev/compare/alibaba-cloud-vs-crusoe.md)
- [Meta vs Crusoe](https://www.subconscious.dev/compare/meta-vs-crusoe.md)
- [Cohere vs Crusoe](https://www.subconscious.dev/compare/cohere-vs-crusoe.md)
- [SambaNova vs Crusoe](https://www.subconscious.dev/compare/sambanova-vs-crusoe.md)
- [Nebius vs Crusoe](https://www.subconscious.dev/compare/nebius-vs-crusoe.md)
- [Crusoe vs fal](https://www.subconscious.dev/compare/crusoe-vs-fal.md)
- [Crusoe vs Novita AI](https://www.subconscious.dev/compare/crusoe-vs-novita-ai.md)
- [Crusoe vs Venice](https://www.subconscious.dev/compare/crusoe-vs-venice.md)
- [Crusoe vs Parasail](https://www.subconscious.dev/compare/crusoe-vs-parasail.md)
- [Crusoe vs Inference.net](https://www.subconscious.dev/compare/crusoe-vs-inference-net.md)
- [Crusoe vs GMI Cloud](https://www.subconscious.dev/compare/crusoe-vs-gmi-cloud.md)
- [Crusoe vs Thinking Machines](https://www.subconscious.dev/compare/crusoe-vs-thinking-machines.md)
- [Crusoe vs Sail Research](https://www.subconscious.dev/compare/crusoe-vs-sail-research.md)
- [Crusoe vs Morph](https://www.subconscious.dev/compare/crusoe-vs-morph.md)
- [Crusoe vs Relace](https://www.subconscious.dev/compare/crusoe-vs-relace.md)
- [Crusoe vs TypeSafe AI](https://www.subconscious.dev/compare/crusoe-vs-typesafe-ai.md)
- [Crusoe vs StepFun](https://www.subconscious.dev/compare/crusoe-vs-stepfun.md)
- [Crusoe vs Runware](https://www.subconscious.dev/compare/crusoe-vs-runware.md)
- [Crusoe vs StreamLake](https://www.subconscious.dev/compare/crusoe-vs-streamlake.md)
- [Crusoe vs Wafer](https://www.subconscious.dev/compare/crusoe-vs-wafer.md)
- [Crusoe vs RunInfra](https://www.subconscious.dev/compare/crusoe-vs-runinfra.md)
- [Crusoe vs Particle.AI](https://www.subconscious.dev/compare/crusoe-vs-particle-ai.md)

## Sources

- [Crusoe Cloud pricing](https://www.crusoe.ai/cloud/pricing)
- [Crusoe Series F announcement](https://www.crusoe.ai/resources/newsroom/crusoe-announces-series-f-funding)
- [Crusoe Managed Inference launch](https://www.crusoe.ai/resources/newsroom/crusoe-launches-managed-inference-delivering-breakthrough-speed-for-production-ai)
- [Crusoe Serverless Inference docs](https://docs.crusoecloud.com/managed-inference/overview/index.html)

Pricing and model lineups change often; figures are a snapshot.