Palo Alto Networks Acquires Portkey for $1.6B — Doubles Last Valuation in 6 Months

Q: What this means for AI infrastructure startups

Three implications for the broader market. First, AI gateway competitors are now in a strategic-bidding window. Helicone, LangFuse, and similar names should expect inbound from CrowdStrike, Cisco, Datadog, and Splunk over the next two quarters. The price ceiling has been set, and acquirers will move fast to lock in the next-best targets. Second, AI evaluation platforms (Patronus, Confident AI, etc.) are likely the next category to see strategic M&A — they're closely adjacent to gateway use cases and benefit from the same enterprise governance trend. Third, open-source AI infrastructure (LiteLLM, vLLM as a service offering) gets an interesting boost — when proprietary alternatives get acquired and integrated into security platforms, open-source becomes the only "neutral" option for enterprises that don't want vendor concentration. For Palo Alto specifically, the integration risk is real. The company has had mixed success absorbing acquisitions historically, and Portkey's product shape is meaningfully different from most of what Palo Alto sells today. The first 12 months of integration will tell us whether this is a transformative bet or a category-protection move that leaves Portkey running semi-independently as a "AI security business unit."

By Jaspal Singh May 1, 2026 Updated: May 4, 2026

Palo Alto Networks has acquired Portkey, an AI infrastructure and gateway startup, in a deal that doubles Portkey's last private valuation to roughly $1.6 billion. The transaction — announced this morning and expected to close in Q3 — folds Portkey's AI gateway and observability stack into Palo Alto's enterprise security platform, and signals an aggressive pivot in how legacy security vendors are absorbing the AI infrastructure category.

Portkey, founded in 2023 and based out of San Francisco and Bangalore, became one of the fastest-growing AI infrastructure plays of the past 18 months. Its core product is an AI gateway — a routing, caching, governance, and observability layer that sits between enterprise applications and the underlying LLM APIs (OpenAI, Anthropic, Cohere, Bedrock, Azure OpenAI, etc.). The pitch was straightforward: as enterprises adopt multiple LLM providers, they need a single chokepoint for governance, cost control, and audit logging. Portkey was the dominant pure-play in that emerging category by Q1 2026.

Why Palo Alto wanted this

Palo Alto's strategic rationale comes down to a single product gap. The company's enterprise security platform has been gradually extending into AI security — Prisma AIRS for AI risk and security, Cortex for AI-driven SOC operations — but it didn't have a native gateway to see and govern AI traffic at the application layer. Portkey closes that gap exactly.

The acquisition fits a pattern Palo Alto has been telegraphing for months: AI security is becoming an enterprise networking concern, not just an SOC concern. If you're a CISO at a Fortune 500, your AI exposure is no longer about whether someone uses ChatGPT in a browser — it's about whether your applications are calling LLM APIs in ways that violate data residency, leak PII, or get compromised through prompt injection. The visibility and control for that has to live at the gateway layer, and Palo Alto wants to own that layer the way it has historically owned the perimeter.

The deal economics

Public reporting puts the headline price at $1.6B, with roughly 65% in cash and the balance in stock. Portkey's last private round closed at an $800M valuation in October 2025, so this is roughly a 2x markup over six months — strong by any measure, especially in a market where down-rounds have been common.

Notable details: Portkey's founders reportedly negotiated retention packages tied to milestone integrations rather than time-based vesting alone, which suggests Palo Alto is treating this as a build-out acquisition rather than a pure tuck-in. The Bangalore engineering team — about 60 people — is expected to remain in place as a regional AI security R&D hub. That's a reasonable bet on Indian engineering depth in this category and an indirect endorsement of Bangalore as a serious AI infrastructure talent market.

My Take

This deal validates a thesis that I've been writing about for six months: AI gateway is the next hot category in enterprise security M&A, and the pure-plays (Portkey, Helicone, LangFuse, Lakera) were always going to get acquired before they could grow into independent IPO candidates. Portkey was the largest, so it gets the highest multiple. Expect at least two more deals in the same category over the next 9–12 months — probably one absorbed into Cisco/Splunk's expanded AI portfolio, and one into a CrowdStrike or Wiz-aligned acquisition.

The deeper takeaway is what this means for AI infrastructure category dynamics. We are watching infrastructure layers harden faster than the underlying models stabilize. Gateways, evaluation platforms, vector databases, and orchestration layers are all consolidating into the major security and cloud platforms while frontier model providers (OpenAI, Anthropic, Google, Meta) are still rotating their capabilities every six months. That dynamic favors platform-layer companies in the short run because they're easier to acquire, integrate, and price than the volatile model layer. Palo Alto's $1.6B for Portkey is a reasonable price for that strategic optionality.

What this means for AI infrastructure startups

Three implications for the broader market. First, AI gateway competitors are now in a strategic-bidding window. Helicone, LangFuse, and similar names should expect inbound from CrowdStrike, Cisco, Datadog, and Splunk over the next two quarters. The price ceiling has been set, and acquirers will move fast to lock in the next-best targets. Second, AI evaluation platforms (Patronus, Confident AI, etc.) are likely the next category to see strategic M&A — they're closely adjacent to gateway use cases and benefit from the same enterprise governance trend. Third, open-source AI infrastructure (LiteLLM, vLLM as a service offering) gets an interesting boost — when proprietary alternatives get acquired and integrated into security platforms, open-source becomes the only "neutral" option for enterprises that don't want vendor concentration.

For Palo Alto specifically, the integration risk is real. The company has had mixed success absorbing acquisitions historically, and Portkey's product shape is meaningfully different from most of what Palo Alto sells today. The first 12 months of integration will tell us whether this is a transformative bet or a category-protection move that leaves Portkey running semi-independently as a "AI security business unit."

Frequently Asked Questions

What does Portkey actually do?
Portkey provides an AI gateway — a routing, caching, governance, and observability layer that enterprise applications use to interact with LLM APIs across multiple providers. It handles cost control, audit logging, prompt injection defense, and governance policies in a single chokepoint.

How much did Palo Alto pay?
The reported deal value is $1.6 billion, with roughly 65% in cash and the balance in Palo Alto stock. That doubles Portkey's last private valuation of $800M from October 2025.

Will Portkey continue as a standalone product?
Initially yes — Palo Alto plans to integrate Portkey's gateway capabilities into its Prisma AIRS and Cortex platforms, but the standalone Portkey product will continue to be sold and supported during the integration window.

Who's the next AI gateway acquisition target?
Helicone, LangFuse, and similar pure-plays are the most likely next targets, with potential acquirers including CrowdStrike, Cisco/Splunk, Datadog, and Wiz/CNAPP-aligned vendors.

The Bottom Line

Palo Alto's $1.6B acquisition of Portkey is a clear signal: AI gateway has been validated as an enterprise security category, and the pure-play startups in this space are now bid targets, not IPO candidates. Expect rapid consolidation over the next 9–12 months and a meaningful re-rating of comparable AI infrastructure private valuations as the strategic floor is reset upward.