Palo Alto Networks, the $130 billion cybersecurity behemoth, just agreed to acquire Portkey — a two-year-old Indian AI infrastructure startup — for a reported $120–140 million. That’s roughly 10x the total capital Portkey ever raised. Three months ago, the company closed a $15 million Series A. Now it’s being absorbed into the largest cybersecurity platform on Earth. If you want to understand where enterprise security is headed in the age of autonomous AI agents, this is the deal that tells the story.
What Portkey Actually Does — And Why Palo Alto Couldn’t Build It
Portkey, founded by Rohit Agarwal and Ayush Garg, built what the industry calls an “AI Gateway” — a centralized control plane that sits between enterprises and every AI model they use. Think of it as a traffic cop for large language models. Every API call, every token, every agent-to-agent interaction passes through Portkey’s infrastructure before reaching production.
The numbers are staggering for a company this young: trillions of tokens processed per month, 125 million daily requests, over $500,000 in AI spending managed daily, and more than 24,000 organizations on the platform — including Postman and Snorkel AI. When your platform becomes the chokepoint through which all enterprise AI traffic flows, you become either a target or an acquisition. Portkey became both.
The Real Story: AI Agents Need Security That Doesn’t Exist Yet
Here’s what Palo Alto Networks understands that most cybersecurity companies haven’t caught up to: the next wave of enterprise AI isn’t chatbots answering customer service tickets. It’s autonomous agents — systems that write code, execute transactions, access databases, and communicate with other agents without a human in the loop. Every single one of those agents is an attack surface that traditional security tools can’t see.
Nikesh Arora’s Palo Alto already has Prisma AIRS, its AI security product line. But Portkey fills a gap that no internal engineering team can easily replicate: real-time observability and governance over millions of concurrent AI agent interactions, with the low latency required for agent-to-agent communication. You can’t bolt that onto a firewall. It has to be purpose-built from the ground up, tested at scale, and battle-hardened across thousands of production environments.
That’s exactly what Portkey spent two years doing while most cybersecurity incumbents were still writing blog posts about “AI risk.”
The India Angle: Elevation Capital’s Fastest Exit Ever
Portkey raised its $15 million Series A from Elevation Capital and Lightspeed in February 2026. Less than three months later, Palo Alto swooped in at roughly double that post-money valuation — meaning Elevation Capital is looking at a near-immediate 2x return on an investment it made weeks ago. Before the Series A, Portkey had raised a $3 million seed from Lightspeed in 2023.
Total capital raised: $18 million. Exit price: $120–140 million. That’s a 7–8x return on total invested capital in under three years. In the current fundraising climate — where most Indian SaaS startups are struggling to raise flat rounds — this is a standout outcome. It also validates the thesis that infrastructure-layer AI companies can achieve liquidity much faster than application-layer ones, because the acquirers are massive incumbents who can’t afford to build from scratch.
Why This Matters More Than Another AI Wrapper Acquisition
Most AI acquisitions in 2025–2026 have been acqui-hires or “talent grabs” where the product gets killed on day one. This isn’t that. Palo Alto explicitly stated it will continue supporting existing Portkey customers while integrating the technology into Prisma AIRS. They’re buying the platform, not just the people.
The distinction matters because it signals where cybersecurity M&A is headed. The big players — CrowdStrike, Palo Alto, Zscaler — are all racing to own the “AI security” layer. But none of them have native infrastructure for governing autonomous agents at scale. Portkey’s acquisition gives Palo Alto a 12–18 month head start over every competitor that now has to build or buy their own AI gateway.
Consider the market timing: enterprises are deploying AI agents faster than they can secure them. Gartner estimates that by 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024. That’s millions of autonomous systems making decisions, accessing sensitive data, and interacting with external APIs — all without human oversight. Every one of those systems needs a governance layer. Palo Alto just bought the company that built it.
The Bottom Line
Palo Alto Networks didn’t pay $140 million for a startup. It paid $140 million to own the toll booth between every enterprise AI agent and the outside world. In cybersecurity, whoever controls the chokepoint controls the market. Portkey built that chokepoint while processing more tokens per month than most companies will process in a year — and they did it with 18 million dollars and a team in India.
For Indian AI startups watching from the sidelines: this is the playbook. Don’t build another chatbot. Build the infrastructure that sits underneath every chatbot, every agent, every autonomous system. Make yourself indispensable at the protocol layer. Then wait for the $100 billion incumbents to come knocking — because they will, and they’ll pay whatever you ask.
