APLD
CORZ
CRWV
GLXY
New AI toys, same old infrastructure race
CoreWeave is trying to do the classic Silicon Valley thing: ship the shiny AI product and build the giant backend to support it. On Thursday, the company launched Physical AI Field Engineering, a service that pairs its specialists with customer teams to build and deploy AI tools using existing engineering and testing data.
The pitch is pretty simple: instead of making customers wrestle with the infrastructure from scratch, CoreWeave wants to be the hands-on crew that gets AI into real engineering workflows faster. The company says the service was built using capabilities it picked up through its Monolith AI acquisition, and it’s already helped on more than 100 projects across automotive, aerospace, and robotics.
DARPA just got a faster lane
CoreWeave also teamed up with Parallel Works to provide a managed AI and computing environment for DARPA’s NODES program. In plain English: researchers get dedicated compute and support without having to spend weeks or months building the plumbing themselves.
That matters because in AI, time is money and compute is oxygen. If customers can start modeling in days instead of months, CoreWeave has a better shot at turning all that infrastructure spending into sticky usage — the kind investors love when they’re trying to justify a premium valuation.
The power race is starting to look friendlier
There’s also some good news on the boring-but-important side of the business: the data centers.
According to BNP Paribas, CoreWeave’s partner ecosystem is moving well enough that some capacity could come online earlier than expected. The note pointed to:
- Core Scientific delivering another 200 MW ahead of schedule
- Applied Digital completing about 75 MW at Polaris Forge 1
- Galaxy Digital delivering 133 MW at Helios
That kind of progress can matter a lot for late-2026 profitability, because more live capacity means more revenue potential. BNP even said the risk/reward still tilts in CoreWeave’s favor near term. Investors, apparently, still like a good “we’re building it fast” story.
But the stock still got smacked
Despite the upbeat setup, CRWV was down about 4% in premarket trading. Growth stocks were under pressure overall, with Nasdaq and S&P futures both red, so CoreWeave got dragged into the mood swing too.
Big picture: CoreWeave is trying to prove it’s more than just expensive GPU real estate. If it can turn infrastructure speed and niche AI services into durable demand, the story gets a lot more interesting — and maybe a lot less dependent on one more data-center delivery headline.