CoreWeave debuts on the stock market and puts a price on the boom of AI-specialised cloud

What happened

CoreWeave, a cloud infrastructure provider specialising in AI workloads and backed by Nvidia, priced its stock market debut at 40 dollars per share on 28 March 2025, after downsizing the offering from its original plans.

The company sets itself apart from the major hyperscalers by focusing on GPU clusters for training and running AI models, a niche that has grown quickly as more businesses need computing capacity specific to these workloads.

What it means for a mid-sized business

Specialised AI providers widen the choice beyond the three big hyperscalers, but they also add another decision to the technology map: a niche provider can offer better performance or pricing, in exchange for less variety of complementary services and a shorter track record as a listed company.

For a mid-sized business in Spain or Catalonia starting to explore AI workloads, the relevant question isn't only the price per hour of compute, but what happens if that provider changes terms, suffers an outage, or if data and models need to be migrated elsewhere.

What to review

Before concentrating critical AI workloads on a specialised provider:

• What continuity and support guarantees it offers compared with traditional hyperscalers.
• How data, models and configurations are exported if you need to switch provider.
• What network connectivity and latency the actual workload requires.
• What security and compliance controls the provider applies to hosted data.

Frequently Asked Questions

How does CoreWeave differ from a hyperscaler like AWS or Azure?

CoreWeave focuses its infrastructure on GPU clusters for AI, while hyperscalers offer a much broader catalogue of general cloud services.

Why was the IPO downsized?

According to reporting at the time of the offering, the company adjusted its valuation and the number of shares on offer before setting the final price.

Should an SME contract this type of provider directly?

It depends on the workload. For intensive AI projects it can bring cost or performance advantages, but dependency and portability should be weighed before deciding.