Core Compute

We’re the first cloud to bring up an NVIDIA Vera Rubin NVL72 system for validation, another big step in building the next generation of AI infrastructure with NVIDIA.

Mark Zuckerberg stood in front of the world and promised to build personal superintelligence for everyone. He said Meta had the resources, infrastructure, and billions to bring the future closer. That video aged like milk left in the sun. Meta’s flagship AI model, code named Avocado has been delayed. It reportedly failed to keep up with rivals on basic tasks like reasoning, coding and writing. The model was supposed to launch this month and it has been pushed to May at the earliest.​ Here is the part that should make every Meta investor sit up straight. Meta's own AI leadership is now discussing licensing Google's Gemini to power Meta AI products. The company that pledged to build superintelligence is now considering renting someone else's brain. This is happening after Meta committed between $115 billion and $135 billion in capital spending for 2026 alone nearly double what it spent last year. Zuckerberg told analysts this would be the year "AI fundamentally transforms our work processes." Instead, Avocado could not even beat Google's Gemini 3.0 from November. It outperformed Meta's older models. It edged past Google's Gemini 2.5 from March but against the current frontier, it fell short on every critical benchmark. The backstory makes this even worse. Zuckerberg personally poached over 20 researchers from OpenAI. He dropped $14.3 billion on Scale AI and installed its 28 year old founder, Alexandr Wang, as Chief AI Officer. He created an entirely new division called Meta Superintelligence Labs. Some recruits received signing bonuses reportedly as high as $100 million. The result of all that spending and all that talent was two models, Avocado for text and Mango for images and both kept slipping their deadlines. Avocado was first expected by late 2025, then early 2026, then mid-March, and now May at the earliest. Meanwhile, Google is becoming the default intelligence layer for the biggest companies on the planet. Apple already signed a multiyear deal for Gemini to power the new Siri and Apple Intelligence features. Analysts estimate that deal could be worth up to $5 billion for Google and Samsung's Galaxy AI already runs on Gemini. If Meta ends up licensing Gemini too, Google could effectively power the AI behind platforms serving well over five billion users combined. That includes Apple’s 1.5 billion devices, Samsung’s massive base, and Meta’s 3.6 billion monthly active users.

Oh and it works in all AI4-equipped cars, so your car can do office work for you when not driving. We’re also deploying millions of dedicated Digital Optimus units in the field at Superchargers where we have ~7 gigawatts of available power.