Another industry chain hard evidence against the "AI computing power surplus theory"! Leading the new cloud Nebius (NBIS.US) with a 10 billion US dollar long contract locked until 2029.

In recent years, the European technology leader Nebius Group NV (NBIS.US) has shifted its focus to providing cloud AI computing resources. They have agreed to sell AI computing infrastructure resources worth at least over 1 billion USD to the globally renowned AI model developer Reflection AI. Reflection AI was founded by two former Alphabet Inc. Class C DeepMind senior researchers. The two companies, Nebius and Reflection AI, stated in a joint statement on Tuesday that the agreement will run until 2029 and will enable Reflection to utilize high-performance AI server clusters built exclusively around the GB300 AI chip developed by NVIDIA Corporation. Before the New York market opened, Nebius' stock price showed a strong upward trend, rising by over 5% in pre-market trading. Nebius will sell AI computing resources worth 1 billion USD to Reflection, and coupled with news on Monday that Meta's investment in its Hyperion data center campus in Louisiana has expanded from an initial announcement of 10 billion USD to over 50 billion USD, it collectively counters the pessimistic views of some AI critics who simplistically interpret Meta's sale of idle AI computing resources as "oversupply of AI computing infrastructure." This transaction is an important part of the "endless narrative of AI computing demand," where AI startups and mature tech giants vie to obtain artificial intelligence computing infrastructure. Last month, Reflection signed a multi-billion dollar AI computing supply agreement with SpaceX to gain access to computing clusters based on the same model of AI chip. Reflection has received financial backing from NVIDIA Corporation and other large institutional investors in recent years. It was recently reported that the company was in discussions to raise 2.5 billion USD in funding based on a valuation of 25 billion USD. Nebius was spun off from the Russian internet services giant Yandex in 2024 and is one of the cloud computing technology companies known as "new cloud service providers" (Neocloud), renting out AI computing power. This Amsterdam-based company has entered into long-term computing infrastructure partnership agreements with Microsoft Corporation and Meta Platforms. Nebius and the U.S.-based AI cloud computing leader CoreWeave belong to the Neocloud race of "AI-specific cloud computing providers," focusing on faster delivery, optimized training/inference stacks, and flexible contracts to serve large AI model developers and enterprises with hefty AI workloads. Microsoft Corporation's collaboration with both companies helps alleviate supply bottlenecks. Nebius is not just a "colo data center" but an integrated AI Cloud leader covering hardware, networking, and scheduling stacks, emphasizing elasticity from single cards to "thousands of AI GPU pre-optimized clusters" to reduce integration costs and delivery uncertainties on Microsoft Corporation's end. The rapid delivery of regional capacity is also a key logic behind Microsoft Corporation's collaboration with Nebius, especially since Nebius has a large Neocloud infrastructure presence in both the EU and the U.S. "Selling AI computing power" is not a sign of oversupply but a way to turn idle capacity into strong cash flow. Nebius' billion-dollar order to Meta's Hyperion campus counters the simplistic view that Meta's...