$GOOGL Threat Intelligence just dropped a wild update: autonomous AI attacks are already here. Not theory. Not future. Now.

Key shift: attackers moved past manual prompting into full multi-agent offensive frameworks. AI is shrinking the gap between vulnerability disclosure and live exploitation — basically zero-day windows are collapsing fast.

Google's response? They're going full autonomous defense too. Gemini + frontier models for reasoning, smaller models for 24/7 scanning. Classic arms race setup.

But here's the real angle most people are missing:

The winners aren't necessarily the cybersecurity vendors. It's the AI compute and infra providers. Why? Because cybersecurity is about to become a massive, persistent inference consumer.

$AMZN CEO Andy Jassy threw out a number earlier: ~20% of AI tokens could eventually go toward cybersecurity workloads. That's not a side use case. That's structural demand.

So yeah, hostile state actors launching endless AI-driven attacks? That's not just a national security problem. It's a long-term bullish driver for compute infrastructure.

Cybersecurity might be the most underrated AI inference growth story right now. $GOOGL $AMZN $MSFT all positioned. Watch the capex.