Engineering at Meta

Introducing Instantaneous PowerLoss Storm: our new infrastructure testing paradigm designed to validate data center readiness against zero-notice, region-wide power failures. Key architecture & engineering highlights: 1️⃣ In-Memory Data Persistence: Leverages dedicated rack batteries and a Power Loss Siren protocol to safeguard volatile state data immediately upon de-energization. 2️⃣ Bootstrapping Loops: Re-starting a dead region introduces circular dependencies. We use Belljar tests in our CI/CD pipelines to catch these early, paired with a custom Twine recovery kit to jumpstart core orchestration services. 3️⃣ Validation: Verified via controlled fault injection in shadow regions, establishing the baseline to test live client traffic next. Read the full deep dive: engineering.fb.com/2026/06/03/dat…

Today we're announcing America's Workforce Academy (AWA), the largest private-sector commitment to the skilled trades in American history, beginning with a $115 million commitment in the first year alone. This cost-free program supports all participants while they learn and then guarantees a job for all graduates in high-demand fields such as electrical work, mechanical systems and plumbing. Every graduate will be guaranteed a job on a Meta construction site. Learn more: meta.com/AmericasWorkfo…

Want to see how we're pushing the limits of recommendation systems? Meet SilverTorch: our "Index as Model" retrieval paradigm that unifies microservices into a single PyTorch neural network to enable more sophisticated models while achieving latency and GPU efficiency step-function changes. The results: ⚡ Up to 23.7x higher throughput than state-of-the-art approaches 📉 20.9x more compute cost efficiency vs. CPU baselines 🎯 Better recommendations with sub-100ms latency Find a technical deep dive and research paper here: engineering.fb.com/2026/05/26/ml-…

Migrating data ingestion systems that process petabytes of social graph data every single day comes with zero room for error, especially when the migration has to happen live, under production traffic. We recently completed a 100% migration to a new hyperscale data warehouse service with zero downtime. To seamlessly transition tens of thousands of incremental ingestion jobs at scale, our teams designed the clear migration job lifecycle and built automated tooling to effectively manage the multi-phased rollout. Dive into the full architectural breakdown, data quality safeguards, and automation strategies on the Meta Engineering blog: engineering.fb.com/2026/05/12/dat…

We’re rolling out version 1.1 of Labyrinth, the encrypted storage system and protocol that secures messages and history on Messenger. Labyrinth 1.1 enhances the reliability of end-to-end encrypted backups with a new sub-protocol that helps messages survive the loss of a device, a switched device, and long gaps between sign-ins. Read our updated white paper, “The Labyrinth Encrypted Message Storage Protocol” for more details: go.meta.me/4c30af

We're sharing a deep dive into how we're hardening end-to-end encrypted backups for WhatsApp and Messenger: • HSM-based Backup Key Vault stores recovery codes in tamper-resistant hardware • Over-the-air key distribution decouples infrastructure rotation from app releases • Cloudflare-signed attestation bundles provide independent verification • Public deployment evidence for every new HSM fleet Read the full post and whitepaper here: engineering.fb.com/2026/05/01/sec…

These partnerships with Overview and Noon build on our approach to helping strengthen the grid and power our AI infrastructure through diverse and reliable solutions. Learn more: go.meta.me/635755

2/🔋 Next-Gen Storage: Deploying up to 1 GW/100 GWh of ultra-long-duration storage with Noon Energy, delivering 100+ hours of capacity.

We’re announcing two new partnerships to bring innovative energy generation and storage to our data centers: 1/ 🛰️ Space Solar: Partnering with Overview Energy to beam up to 1 GW of space solar power from orbit to Earth for around the clock power production.

Today we’re announcing an agreement with Amazon Web Services to bring tens of millions of AWS Graviton cores to our compute portfolio. This partnership marks an expansion of our diversified AI infrastructure and will help scale systems behind Meta AI and agentic experiences that serve billions of people. Learn more: go.meta.me/2bc5c5

We're breaking ground on a new AI-optimized data center in Tulsa, Oklahoma. A few things that make this facility interesting from an engineering perspective: 💧 Cooling: Closed-loop, liquid-cooled system that recirculates the same water + uses zero water for server cooling during the majority of the year. ⚡ Energy: 100% clean energy match. We're adding over 1,500 MW of clean energy to the grid in Oklahoma, and paying hundreds of millions through utility bills to fund grid infrastructure like substations + transmission lines. 🎓 Workforce pipeline: Partnering with Tulsa Tech + Tulsa Community College on a new cross-institutional program for digital infrastructure careers, targeting 200+ graduates annually in cooling simulation, fiber optics, structured cabling + AI/data analytics. 🌾 Water restoration: A 10-year partnership with Phytech to deploy plant-sensor technology across ~1,500 acres of commodity crops near Tulsa, saving 50M+ gallons of water per year. Learn more: go.meta.me/1bac0e

Today we're announcing LevelUp: a free, four-week training program that takes people with no prior experience and prepares them to work as fiber technicians on data center construction sites across the US. We built this program with CBRE because the fiber technician field, and the broader construction industry, is facing a nationwide shortage at a time when data center demand is higher than ever. How it works: 🔧 Classroom instruction, hands-on labs + team activities covering transferable technical skills 🎓 Graduates have the opportunity to work at Meta's US construction sites through our contractor network 🤝 Open to everyone from recent high school grads to mid-career professionals Since 2010, Meta's data center projects have supported 30,000+ skilled trade jobs during construction + 5,000+ permanent operational roles. LevelUp is about building the pipeline to keep that going. Learn more: go.meta.me/0eb3f6

@FAbnousi Great piece, @FAbnousi! Thanks for leading the conversation about how we need to rethink health data architecture in the world of AI.

AI knows medicine. It doesn’t know patients. Every day, we ask patients to reconstruct a life they were busy living, not recording. A patient moves through the health care system like a traveler through airports — each stop generates a new record, rarely connected to the rest of the journey. Medicine has learned to interpret records — but not the life that produces them. With @CelinaMYongMD in @statnews.

As the industry transitions to post-quantum standards, we’re sharing lessons from Meta’s multi-year PQC migration. By outlining our approach, from risk-based prioritization and building cryptographic inventories to deploying hybrid PQC-classical models, we aim to provide a roadmap for organizations strengthening their resilience against future quantum threats like "Store Now, Decrypt Later.” We’re also proposing a framework of PQC Migration Levels (from PQ-Unaware to PQ-Enabled) to help teams manage complexity across diverse use cases. Read the full technical deep dive: go.meta.me/e4b348

At Meta, WebRTC powers real-time audio and video across 50+ use cases. But forking a large open-source project within a monorepo presents a unique challenge — over time, an internal fork can drift behind upstream, cutting itself off from community upgrades. We built a dual-stack architecture that enabled safe A/B testing across all 50+ use cases, then built workflows that keep us continuously upgraded with upstream, improving performance, binary size, and security across the board. Here's how we escaped the "forking trap" ↓ engineering.fb.com/2026/04/09/dev…

Today we’re announcing an expanded partnership with @Broadcom to co-develop multiple generations of our next-generation MTIA chips. This custom silicon will help power AI across all of Meta's apps and services, ensuring we have the massive computing foundation needed to deliver personal superintelligence to billions Read more: go.meta.me/220372

We're open-sourcing BOxCrete, a new AI model for the construction industry. Using Bayesian optimization, BOxCrete helps producers rapidly design concrete mixes with domestic materials, bypassing months of lab work. The results from our data center build in Rosemount, MN: 🚀 43% faster time to full structural strength 🛠️ 10% reduction in cracking risk 🇺🇸 100% domestic material usage We are open-sourcing the model and the foundational data to empower producers everywhere. Check out the full technical deep dive on our Engineering blog: go.meta.me/90538a

Our next-gen data center in El Paso, Texas is officially growing to 1GW. Here are a few updates: 📐 The build: We're increasing our investment from $1.5B to more than $10B, supporting 4,000+ construction jobs at peak + 300 permanent operational roles once complete. 🎓 The community: A new $500K workforce development partnership with El Paso Public Schools will connect students with career paths in STEM + trades through real-world learning experiences. 💧 Water stewardship: The data center's closed-loop cooling system means zero operational water use most of the year. We're also partnering with @DigDeepH2O to bring clean water to 100+ homes for the first time and Bonneville Environmental Foundation to restore millions of gallons annually through new irrigation tech for local farmers. Learn more here: about.fb.com/news/2025/10/m…

We’re sharing the technical architecture behind friend bubbles in Facebook Reels – a recommendation system that highlights Reels your friends have liked or reacted to – including how machine learning estimates relationship strength and ranks content your friends have interacted with to create more opportunities for meaningful engagement and connection. Check out our blog post here: go.meta.me/691937

Today we’re announcing a new partnership with @Arm to collaborate on the development of multiple generations of purpose-built CPUs to support our compute and AI infrastructure. The first generation of the chip that we co-developed, the Arm AGI CPU, delivers more than 2x performance per rack compared with x86 platforms. The Arm AGI CPU will also be available to the broader AI ecosystem through Arm and Meta will be releasing our board and rack designs for this CPU under the Open Compute Project later this year. Learn more about our partnership: go.meta.me/15e3b8