Nvidia just locked in a $6.3 billion agreement with CoreWeave, giving the New Jersey-based data center firm one hell of a guarantee.
Announced Monday, the deal ensures that Nvidia will cover any unused cloud capacity through April 13, 2032. If CoreWeave can’t sell it, Nvidia still pays. That’s the deal.
According to Reuters, this is an extension of a prior agreement signed in April 2023, and now it’s stretched out for another seven years.
CoreWeave shares jumped 8% after the announcement. The company runs AI-heavy data centers in both the U.S. and Europe, offering access to Nvidia’s GPUs, the same chips that power large-scale AI tools like ChatGPT.
The deal cements CoreWeave as one of Nvidia’s biggest infrastructure allies and gives them cover if demand dips. Right now, AI is hot. But Nvidia’s saying: if that changes, we’ve still got your back.
Nvidia commits to buying unused capacity
Barclays analysts said this updated contract acts as a safety net for CoreWeave. “The amended agreement serves as a backstop,” they explained, “ensuring capacity will be leveraged irrespective of the end customer.”
The analysts also pointed out that this $6.3 billion spend gives Nvidia more control over cloud operations without relying too much on giants like Microsoft or OpenAI.
That matters because CoreWeave is already tightly tied to OpenAI. Back in March, the two companies signed a five-year, $11.9 billion deal for CoreWeave to provide cloud infrastructure. On top of that, OpenAI committed to spending another $4 billion by April 2029.
The connection’s deep. Nvidia’s investment adds even more weight to that triangle: Nvidia, CoreWeave, OpenAI. Demand’s been surging. CoreWeave reported a major uptick in usage for Q2, driven by AI adoption.
But the bills are also climbing. The company’s operating expenses quadrupled to $1.19 billion in that quarter. The growth is real, but it’s burning through money fast. This Nvidia backstop helps keep the lights on.
China goes after Nvidia over Mellanox deal
While Nvidia is tightening its grip on U.S.-based AI infrastructure, China is turning up the heat. On Monday, the country’s State Administration for Market Regulation (SAMR) said Nvidia broke Chinese anti-monopoly laws, based on a preliminary investigation, as Cryptopolitan reported previously.
The probe ties back to Nvidia’s 2020 acquisition of Mellanox, an Israeli company that builds network solutions for servers and data centers. China had approved that deal, with some conditions. Now, they’re saying Nvidia didn’t meet them.
That investigation lands at a bad time. Trade talks between China and the U.S. restarted Sunday in Madrid, and this could blow things up. Over the weekend, Beijing launched two other investigations tied to semiconductors.
One is an anti-dumping case against chips from the U.S., and the other accuses America of discriminatory policies targeting China’s chip industry. This isn’t Nvidia’s first clash with Chinese regulators.
Earlier in the year, China blocked the export of Nvidia’s H20 chip, a product built specifically to comply with U.S. export laws. That rejection showed how geopolitical tensions are freezing out even the most careful attempts to play by the rules.
Jensen Huang, Nvidia’s CEO, hasn’t been quiet about any of this. He said American companies should be allowed to sell to China, estimating the Chinese AI market could reach $50 billion in just a few years. “If we’re not there,” Huang said, “Huawei will be.”
It didn’t fall on deaf ears. Last month, Nvidia cut a deal with Washington to restart chip sales to China. The catch? Nvidia has to give 15% of that revenue to the U.S. government. The company is now in talks to export even more advanced chips to China under tighter regulations, but nothing is finalized.
Get seen where it counts. Advertise in Cryptopolitan Research and reach crypto’s sharpest investors and builders.
Source: https://www.cryptopolitan.com/nvidia-to-spend-6-3b-on-coreweave-cloud/
