CoreWeave Plans to Spend Up to $39 Billion in One Year as AI Compute Demand Explodes

Imagine running a company where increasing your annual spending plan to $39 billion is considered good news.
Welcome to AI infrastructure. 🤣
CoreWeave raised its 2026 capital-expenditure forecast from $31–35 billion to $35–39 billion after stronger-than-expected demand for AI cloud computing. Its shares rose more than 14% in extended trading following the results. �
Reuters
The company reported a second-quarter revenue of $2.58 billion, while its revenue backlog increased to $104.2 billion. More than $25 billion in new customer commitments were added during the quarter. �
Reuters
Its customers include meta, Anthropic, and caterpillar. �
Reuters
What Is a Neocloud?
CoreWeave belongs to a newer class of cloud providers specialising heavily in AI computing.
Instead of attempting to provide every imaginable internet service like a traditional hyperscaler, a neocloud concentrates on access to high-end GPUs and AI infrastructure.
The model benefits from one enormous market problem:
Everybody wants powerful AI computing.
There isn’t enough convenient capacity.
The Risk Hidden Inside the Growth
Huge demand justifies huge construction.
But construction creates debt, leases, and long-term commitments.
If AI computing prices fall rapidly or today’s expensive hardware depreciates faster than expected, infrastructure companies must still service those obligations.
That’s why the CoreWeave story is simultaneously:
evidence of incredible AI demand
and
evidence of how capital-intensive the race has become.
Mary Chuks’ Perspective
This is the scaler economy.
The model companies need more intelligence.
That creates demand for GPUs.
GPUs require data centres.
Data centres require electricity.
Electricity requires infrastructure.
Every layer scales the next layer.
And every layer sends another invoice upstairs. 😭🤣
Conclusion
CoreWeave’s numbers show that businesses are not merely talking about adopting AI.
They are signing contracts worth tens of billions for the computing required to run it.
The AI boom has moved from software ambition to concrete, cables, and very serious money.
Primary source: Reuters, 11 August 2026. �
Reuters