'We have maybe 20 months' to rebuild for AI agents, Meta's infrastructure VP tells VB Transform 2026

2026-07-15 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Data Science & Analytics · Depth: Intermediate, short

Summary

Meta's VP of Engineering, Barak Yagour, highlighted at VB Transform 2026 the critical need for organizations to transform their enterprise infrastructure within approximately 20 months to accommodate AI agents. Yagour noted that agentic queries hitting Meta's data systems grew 30x in a single half, reflecting a broader trend where automated internet traffic surpassed human traffic in 2025, reaching 51% and growing eight times faster. This shift is breaking three core infrastructure assumptions: capacity, identity, and velocity. Meta is addressing this by implementing agent-aware infrastructure with dynamic controls and cost attribution, developing trusted data environments for governed agent autonomy, and overhauling its data layer. This includes transitioning to real-time streaming for ranking pipelines and schema-aware storage to optimize GPU utilization, aiming for 500 million queries per second and a petabyte per second throughput for training data reads.

Key takeaway

For AI Architects and MLOps Engineers planning future infrastructure, recognize the urgent 20-month window to re-architect systems for agentic AI. Your current capacity, identity management, and CI/CD pipelines are likely insufficient for the exponential growth of agent traffic. Prioritize implementing agent-aware infrastructure, robust data governance for agent autonomy, and migrating to real-time, schema-aware data processing. Failure to adapt risks significant operational bottlenecks and security vulnerabilities as agentic workloads scale.

Key insights

Enterprise infrastructure must rapidly adapt to agentic AI, which is fundamentally changing data consumption and system loads.

Principles

• Infrastructure must be agent-aware.
• Autonomy requires robust governance.
• Reasoning models demand real-time data.

Method

Implement agent-aware infrastructure with dynamic controls, cost attribution, and adaptive throttling. Establish trusted data environments for agent autonomy with real-time access evaluation and sensitive field masking.

In practice

• Transition to real-time data streaming.
• Adopt schema-aware data storage.
• Re-evaluate agent data interfaces.

Topics

• AI Agents
• Enterprise Infrastructure
• Data Governance
• Real-time Data Streaming
• Schema-aware Storage
• MLOps

Best for: CTO, VP of Engineering/Data, Investor, AI Architect, Director of AI/ML, MLOps Engineer

Related on AIssential

• How to build resilient agentic AI pipelines in a world of change — Resilient AI pipelines are essential for enterprise operations, ensuring continuous performance and compliance amidst constant change.
• D&B's database of 642 million businesses was built for humans, not AI agents. So they rebuilt it. — Enterprise data systems built for humans require fundamental redesign for AI agents.
• The Year AI Gets Serious — 2026 marks AI's transition from hype to practical deployment, driven by agentic systems executing complex tasks.
• Scality's Long Bet: How The 15-Year-Old French Storage Company Is Rewiring Itself for the Agentic AI Era — Established companies can reinvent themselves by integrating agentic AI to address evolving market demands and expand opportunities.
• Agents think in milliseconds, legacy infrastructure doesn't. LinkedIn, Walmart and Zendesk shared how they closed the gap at VB Transform 2026 — Legacy enterprise infrastructure, built for human speeds, is the primary bottleneck for AI agents in production, not the models themselves.
• Agents are ready but your architecture probably isn't — AI value stems from outcome-driven design, robust data infrastructure, and proactive governance, not just model selection.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.