Crusoe raises $3B at $30B: your compute has a landlord
Crusoe tripled its valuation to $30B in ten months on AI data center demand. What compute concentration means for the price you pay per token.
AI compute keeps repricing upward, and the latest mark is not a chip vendor. Crusoe raised over $3 billion at roughly a $30 billion valuation, Bloomberg reported on September 3. Ten months ago the same company was worth about $10 billion. Nothing about the model you call changed. The floor under it did.
What actually happened
Atreides Management and Valor Equity Partners co-led the round, with Mubadala Capital participating, per Bloomberg. TechCrunch puts the prior mark at $1.38 billion raised at $10 billion in October 2025 — roughly a 3x step-up in under a year.
Crusoe builds and operates AI data centers. Named customers include Meta, Microsoft, OpenAI, Oracle, and Jane Street, and the company signed a $13 billion, five-year infrastructure contract with Jane Street. The company started in 2018 flaring-gas crypto mining and pivoted into AI capacity. TechCrunch reports leadership has met with Goldman Sachs, Morgan Stanley, JPMorgan Chase, and Bank of America about a possible near-term IPO.
Why compute concentration matters for your business
You do not buy from Crusoe. You buy from someone who does — or from someone who buys from someone who does. That is the point. The stack between your API key and a rack in a building is now four or five parties deep, and each one has its own margin, its own contract term, and its own refinancing schedule.
What that means in practice: your per-token price is not set by your vendor's generosity or by Moore's law. It is set by a lease signed years ago and by whoever is willing to outbid you for the same capacity. A single trading firm just committed $13 billion over five years. Your annual spend is a rounding error against that, which is exactly why you should never assume today's rate card is durable.
Three things to do, none of them dramatic. Keep contract terms short — twelve months maximum on anything AI-priced, so a market move is an inconvenience rather than a crisis. Keep a second provider wired up and tested quarterly, not theoretically available. And track your cost per completed task, not your cost per million tokens, because that is the number that survives a model swap and tells you whether a price increase actually hurt.
Compute is real estate now. Prices go up, landlords change, and the tenant with a portable setup is the one who negotiates.
Key takeaways
- Crusoe raised over $3B at ~$30B on September 3, roughly tripling its October 2025 mark of $10B
- Atreides Management and Valor Equity Partners co-led; Mubadala Capital participated; an IPO has been discussed with major banks
- Named customers include Meta, Microsoft, OpenAI, Oracle, and Jane Street — the last on a $13B five-year contract
- Your token price tracks data center economics you have no visibility into, so cap AI contract terms at twelve months
- Measure cost per completed task, not cost per million tokens — it survives model swaps and shows real impact
The compute market reprices faster than your annual contract. We build AI systems with a swappable model layer, a tested fallback provider, and cost telemetry measured per completed task instead of per token. See how we build for provider churn, or send us your vendor contracts and we'll flag the terms that will hurt.
Sources: Bloomberg, TechCrunch.
- #ai-infrastructure
- #data-centers
- #funding
- #vendor-risk
- #inference
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
XDOF at $1.2B: the training data is the product
XDOF is in talks at a $1.2B valuation for collecting robot training data, three months out of stealth. Proprietary training data is the asset labs cannot self-serve.
Read itResect AI takes $25M to fix hallucinations in-model
Resect AI launched with $25M to intercept LLM hallucinations at runtime instead of catching them after. How to evaluate the claim before you buy it.
Read it