
Meta's VP of Engineering Barak Yagour warned at VB Transform 2026 that agentic AI queries to Meta's data systems surged 30x in a half-year, exposing the fundamental inadequacy of infrastructure built over two decades for human use. Automated internet traffic has already surpassed human traffic at 51% of the total and is growing eight times faster, signaling an industry-wide inflection that enterprises have roughly 20 months to address through infrastructure overhaul.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Meta VP of Engineering Barak Yagour told VB Transform 2026 that agentic queries hitting Meta's data systems grew 30x in a single half-year, breaking assumptions the company built over two decades. He warned that enterprise infrastructure built for humans is starting to fail under the load of AI agents.
Why it matters
Automated traffic already overtook human traffic on the internet last year, reaching 51% of the total according to Imperva's 2025 Bad Bot Report, and is growing roughly eight times faster than human traffic per HUMAN Security's 2026 State of AI Traffic report. Organizations that do not rebuild their systems risk being unable to handle the shift to agentic AI workloads.
What to watch
Yagour framed the window to adapt as perhaps 20 months, suggesting Meta and other enterprises face an urgent deadline to overhaul infrastructure that was never designed for autonomous agent queries at scale.
At VB Transform 2026, Meta VP of Engineering Barak Yagour opened his presentation wearing a pair of Ray-Ban Meta AI glasses, a physical symbol of how deeply AI has already integrated into everyday life. But his core message was a warning: the infrastructure underpinning the internet was built for humans, not agents, and that assumption is failing. Yagour, who leads Meta's data infrastructure organization, shared a stark statistic: agentic queries hitting Meta's data systems grew 30x in a single half-year. That velocity of growth is breaking foundational assumptions the company spent two decades constructing. The problem extends far beyond Meta. Imperva's 2025 Bad Bot Report documented that automated traffic overtook human traffic on the internet last year, now comprising 51% of the total. More alarming, HUMAN Security's 2026 State of AI Traffic report found that this automated traffic is growing roughly eight times faster than human traffic. Yagour cited both figures to illustrate what he called an inflection point—a structural shift, not a temporary spike. His central argument was that organizations need to transform to meet the needs of agentic AI, and the window to do so is narrow. He told the audience that enterprises have maybe 20 months to rebuild systems originally engineered for human-speed interactions to handle autonomous agent workloads at machine speed.
The infrastructure gap Yagour identified reflects a fundamental mismatch between how enterprise systems were architected and the operational reality of AI agents. Over two decades, companies like Meta built data systems optimized for human-paced queries with predictable patterns; a 30x surge in agentic traffic in a single half-year invalidates the assumptions embedded in those designs. This is not an isolated Meta problem. Industry data from Imperva and HUMAN Security confirm that the shift from human-directed to automated traffic has already crossed a threshold—automated requests now constitute a majority of internet traffic and are expanding at an order of magnitude faster than human requests. Yagour's framing of this as an "inflection" underscores that the trend is not marginal or reversible; it is structural. The 20-month window he cited is not arbitrary speculation but a practical estimate of how long organizations have before the gap between current infrastructure and agentic demand becomes catastrophic.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion




Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack