AI infrastructure rush
Big tech is locking up raw compute: Meta agreed a fresh $21 billion cloud deal with CoreWeave to secure capacity for AI workloads, signalling that compute access is a strategic asset rather than a commodity. That kind of spending privileges engineers who build and optimise infrastructure, distributed systems and GPU platforms — the roles most likely to command durable pay and equity value right now. (reuters.com)(cnbc.com)
Meta just agreed to spend another $21 billion with CoreWeave for artificial intelligence cloud capacity, and the contract runs through December 2032. That is on top of an earlier $14.2 billion arrangement, which puts Meta’s total committed spend with CoreWeave at about $35 billion. (cnbc.com) (investors.coreweave.com) CoreWeave is not a consumer app or a model lab. It is a company that rents out huge clusters of Nvidia graphics processing units, which are the chips most companies use to train and run large artificial intelligence systems. (cnbc.com) (investors.coreweave.com) This deal is about access, not convenience. When Meta signs a multiyear contract this large, it is treating compute the way airlines treat airport gates or retailers treat warehouse space: something you lock up early because there may not be enough later. (reuters.com) (investors.coreweave.com) The timing tells you how tight the market is. CNBC reported that the new $21 billion commitment covers 2027 through 2032, which means Meta is reserving capacity years before it needs to use every machine. (cnbc.com) Meta is doing this while planning one of the biggest spending years in corporate tech history. In its January 29, 2026 earnings report, the company said 2026 capital expenditures should land between $115 billion and $135 billion, up from $72.22 billion in 2025. (investor.atmeta.com) That money is going into the physical layer of artificial intelligence. Data centers, power systems, networking gear, and chip-filled server racks now matter as much as clever model design, because a model cannot train or answer users without electricity, cooling, and thousands of linked processors. (investor.atmeta.com) (reuters.com) CoreWeave’s rise shows how fast that layer is being repriced. The company was established in 2017 and completed its public listing on Nasdaq in March 2025, and now Meta is willing to commit tens of billions of dollars to its infrastructure. (investors.coreweave.com) Nvidia sits in the middle of this buildout too. In January 2026, Nvidia bought $2 billion of CoreWeave stock in a private placement, tying the leading chip supplier even more closely to one of the fastest-growing renters of those chips. (sec.gov) Reuters said Meta is expanding infrastructure for more complex artificial intelligence workloads after falling behind rivals following an underwhelming model release last year. That helps explain why Meta is buying from outsiders even while building its own capacity at home. (reuters.com) (cnbc.com) The labor market signal is buried inside the contract. If the scarce thing is not just software talent but the ability to make giant clusters run reliably, then the people closest to distributed systems, graphics processing unit scheduling, networking, and data center efficiency are sitting nearer the money than they were two years ago. (reuters.com) (investor.atmeta.com)
