Imagine you\u2019re visiting a new city and want to find a restaurant that impresses your vegan in-laws. Using voice conversations on the Meta AI app, you ask, \u201cHey Meta, what are the best vegan options around?\u201d\u00a0<\/span><\/p>\n Within seconds, Meta AI \u2014 powered by <\/span>Muse Spark<\/span><\/a> \u2014 responds with a list of local vegan restaurants, a short description of each restaurant\u2019s vibe, and a map showing you exactly where the restaurants are. It\u2019s a quick and seamless interaction that feels effortless, but behind that brief exchange are layers of calculations enabled by compute power.<\/span><\/p>\n Simply put, compute power is the measure of how much work a computer chip can do and how fast it can do it \u2014 like horsepower in a car engine. Compute power is measured in FLOPS: floating-point operations per second, or the number of calculations that a chip can perform in one second. FLOPS measure the speed of compute and gigawatts measure the scale of it, or how many chips you can keep running at once.\u00a0<\/span><\/p>\n When you ask Meta AI to find a vegan restaurant, it runs billions of calculations in just a few seconds. Your voice is captured, converted from sound waves into text, and routed to computers or servers inside a <\/span>data center<\/span><\/a>. From there, a large language model (LLM) , and the result is delivered right to your ear.\u00a0<\/span><\/p>\n Even simple actions, searching for a local barbershop on Instagram, require layers of computation: understanding language, processing your query, scanning an index, generating results, and delivering it back to you, all before your thumb leaves the screen. All of this processing power is made possible by processing chips inside the servers inside our data centers.\u00a0\u00a0<\/span><\/p>\n So why does the future of AI depend on compute? Here\u2019s a closer look.<\/span><\/p>\n Compute is an abstract concept, but it\u2019s delivered by physical chips. Different chips are designed to handle different types of calculations and workloads.\u00a0<\/span><\/p>\n \u00a0<\/p>\n At Meta, we\u2019re building a global network of AI-optimized data centers, each designed with the flexibility to support both our AI workloads and the other workloads that are central to our apps and services. We believe that building at this scale requires a diversified approach to infrastructure.<\/span> T<\/span>hat\u2019s why we\u2019re sourcing silicon from a range of partners to ensure the right chips are matched with the right workload, allowing us to <\/span>build and deliver new AI experiences at a faster pace<\/span>.\u00a0<\/span><\/p>\n Our custom MTIA silicon is an essential part of our efforts. We\u2019re developing and deploying four new generations of chips within the next two years to support ranking, recommendations, and generative AI workloads. In April, <\/span>we announced<\/span><\/a> an expanded partnership with Broadcom to co-develop multiple generations of MTIA chips.<\/span><\/p>\n And earlier this year, we announced <\/span>a partnership with Arm<\/span><\/a> to co-develop the Arm AGI CPU, \u2014\u00a0the first data center processor specifically designed to handle the massive amount of data movement demanded by AI workloads. We\u2019ve also announced partnerships with industry leaders <\/span>AWS<\/span><\/a>, <\/span>AMD<\/span><\/a>, and <\/span>NVIDIA<\/span><\/a> to supply chips for our compute portfolio.\u00a0<\/span><\/p>\n These partnerships will enable us to continue innovating and building AI tools for the future. We recently announced<\/span> Muse Spark<\/span><\/a>, our most advanced AI model to date and the first LLM built by Meta Superintelligence Labs. Muse Spark is natively multimodal, processing voice, text, and images together. What makes it possible is compute at every level \u2014 from training models across thousands of GPUs to supporting billions of inferences each day on custom MTIA chips \u2014 and all of it running through efficient networks of servers at data centers around the world.\u00a0\u00a0<\/span><\/p>\n The demand for more powerful and efficient compute will only accelerate. As AI continues to become more capable, personal, and integrated into people\u2019s lives, we\u2019ll keep building the infrastructure needed to power it. <\/span><\/p>\n<\/p><\/div>\nWhat Is Compute Power?<\/h2>\n
The Building Blocks of Compute<\/h2>\n
\n
\n
![\"Helios](\"https:\/\/about.fb.com\/wp-content\/uploads\/2026\/06\/02_GPU.jpg?resize=960%2C836\")
<\/a><\/p>\n\n
How Does Compute Power Meta\u2019s AI?<\/h2>\n