The 7.6x Step-Up: What Zenity's $125M Actually Buys

Hook

Three flags. No American ones.

SoftBank writes the check. Hitachi validates the industrial use case. LG tags the consumer-IoT frontier. For Zenity — an Israeli-founded AI Agent security platform — the $125M Series B investor list reads less like a venture round and more like a geopolitical statement.

The step-up compounds the anomaly. January 2024: $16.5M Series A. Mid-2025: $125M Series B. That is a 7.6x increase in eighteen months, during the steepest security-funding contraction in a decade. Capital is not rewarding momentum here. Capital is pre-positioning for a deployment wave that has not arrived yet.

Gartner says less than 1% of daily work decisions were agentic in 2024. It projects 15% by 2028. The security layer is being funded four years ahead of the actual workload. Liquidity didn't create this round. Procurement fear did.

Context

Zenity was founded in 2021 to secure something nobody took seriously at the time: low-code and no-code application sprawl. The January 2024 A-round, led by Insight Partners, framed it as a platform protecting low-code and no-code applications plus AI-generated content. The 2025 B-round expands that mandate to the thing that keeps CISOs awake: autonomous AI Agents.

The technical stack matters. Agent security spans three layers. MAAP — Machine Access and Behavior Protection. AI-SPM — AI Security Posture Management. AI-SSRM — AI Security Services and Risk Management. The defended subject changes from human operators to machine principals. Security no longer watches click logs. It watches behavior chains: what an Agent was permitted to access, what it actually accessed, and who inherits liability when those diverge.

The trigger is shadow AI. Microsoft Copilot Studio and Google Vertex AI Agent Builder allow non-technical employees to ship a working enterprise Agent in hours. My own work tracking autonomous wallets on Solana — five thousand AI-managed addresses executing micro-transactions without human sentiment — tells me the same pattern is emerging across every execution layer. Machine actors scale faster than governance can map them.

The shift from network perimeter to behavior perimeter has a concrete consequence. A firewall knows where traffic flows. A behavior-layer platform must know what a machine actor is permitted to intend — a question of semantics, not protocols. When I audited ICO contracts in 2017, the recurring flaw was admin keys: projects promised decentralization while retaining a master override. Agent security reprises that pattern at scale. Does the Agent inherit its creator's full permissions, or does it hold independent minimal privileges? That decision is the centralization question of the machine era. The security perimeter is sinking. It used to sit at the network edge. Now it sits at the behavior edge. Zenity is selling the fence around that new boundary. The open question is whether the fence is load-bearing or cosmetic.

Core

The Commercialization Math — Follow the Term Sheet

The B-round scale demands examination. In a normal security software cycle, a $125M Series B at a 10-15x revenue multiple implies annual recurring revenue between $60M and $85M. No B-stage startup carries that weight. The more plausible read: Zenity's ARR sits in the $20M-$30M range, and the round prices in an AI-native premium — a 20x-plus multiple that anticipates a category, not a company.

Enterprise contract values anchor the projection. CrowdStrike and Zscaler command $50K-$200K in annual contract value for mature platforms. Zenity, positioning as AI Agent security infrastructure, likely lands in the $100K-$300K band. That arithmetic produces a wide funnel: 150 to 300 enterprise logos would generate $15M to $90M in ARR. Too wide for conviction. The round's job is to narrow it.

The round also marks the phase transition from pilot to procurement. A B-stage raise of this size implies the POC gauntlet is cleared. The unanswered metric is net revenue retention. In security software, NRR above 120% is the threshold that justifies 20x multiples. Below it, the multiple reverts to the SaaS mean. The next disclosure settles the debate.

The quiet innovation may be pricing itself. Per-seat security pricing is the industry default — a headcount tax. Agent security allows something different: per-Agent pricing. The customer shifts from paying for human licenses to paying for machine actions. That re-frames security from an overhead line item to a variable cost of autonomy. If Zenity is pioneering that model, the valuation narrative will survive contact with procurement. If it is selling another dashboard, it will not.

The underwrite doesn't move on conviction. It moves on comparables. The reference set for this round is not CrowdStrike's early days. It is the AI-infrastructure wave — companies like Wiz and SentinelOne, priced on category creation before revenue maturity.

The Investor Matrix: Three Non-American Flags

The composition is the message. SoftBank brings a Japan channel effect: its enterprise AI packaging can force-multiply Zenity's adoption curve the way CrowdStrike's alliance with AWS extended its reach. Hitachi lives inside industrial control systems where autonomous agents already govern physical processes. The OT layer needs behavioral baselines because a prompt-injection attack on a factory scheduler is a safety incident, not a data incident. LG deploys agents into smart homes; its ThinQ Agent operates in the most privacy-sensitive deployment environment that exists.

Read the three flags together and the strategy is coherent: Japan, Korea, and the broader Asia-Pacific industrial complex are the growth corridor. The EU AI Act, China's mandatory AI content labeling rules, and NIST's AI Risk Management Framework share one consequence. They push enterprises toward security infrastructure that is demonstrably not controlled by American cloud giants. Digital sovereignty is a procurement criterion now, not a political slogan. Zenity's capital table is a compliance feature.

The revenue geography will follow. A non-American capital table tends to pull customer acquisition toward the investors' home markets. Expect Zenity's go-to-market to weight Japan and Korea more heavily than the United States over the next two quarters. Sovereign procurement follows sovereign capital. The implication for North American enterprises is uncomfortable. The most credible agent security vendor in the Asia-Pacific corridor may be structurally disinclined to prioritize U.S. regulatory accommodations. Data residency, government access requests, and export controls are not abstract concerns in this industry. They are line items.

The absence of American VCs is the loudest signal in the round. No a16z. No Sequoia. No Accel. Two readings. Deliberate selection — strategic capital over financial capital, distribution instead of merely balance sheet. Or wait-and-see — American funds are holding until Zenity proves the U.S. commercial market. The difference matters. One reading says the product is ahead of the market. The other says the market is ahead of the product.

The Competitive Window Has an Expiration Date

The Wiz playbook defined the category-specialist strategy: cloud-native security, four years, zero to a $23B acquisition. A dedicated category expert beats a platform module in an early market because procurement wants a referenceable specialist, not a slide in a mega-suite deck.

The platform counterattack is already on the calendar. CrowdStrike and Palo Alto Networks launched AI security posture management modules in 2024. Microsoft absorbed AI security into Purview and Security Copilot. These vendors hold the enterprise channel and the procurement trust. When Agent security becomes a checkbox in a renewal negotiation, cross-sell pressure becomes existential for independents.

One data point fractures that logic. CrowdStrike's July 2024 global outage — the largest enterprise security failure in history — detonated the single-vendor trust thesis. Multi-vendor strategies are back in fashion. Independent security startups have a window that did not exist before that incident.

The open-source layer is the quieter threat. OpenAI's agent safety evals, LangChain's LangSmith, Meta's Purple Llama — these cover the raw detection surface. When open-source reaches parity there, Zenity's differentiation must move up the stack: compliance reporting, enterprise integration, agent identity, orchestration. That is an entirely different build.

The fight nobody is discussing is taxonomy. If Agent security is reclassified as a feature of network security, the standalone valuation resets to zero. Categories determine multiples. The company that defines the category owns the premium. The $125M is not just working capital. It is the budget for buying the right to define the box.

The Raise as a Sales Artifact

Enterprise security procurement asks two uncomfortable questions. Will you exist in five years? Will your roadmap survive a down quarter? A nine-figure balance sheet answers both in one sentence. This round is a sales tool. The PR emphasis on the investor lineup — rather than product milestones — confirms it.

The exit path is unusual. In a typical security story, the acquirer is an American platform vendor. Here, the likelier strategic acquirer sits inside the Hitachi-SoftBank orbit: a Japanese or pan-Asian security integrator absorbing a global Agent security standard-bearer. The window is two to three years, at a $1.5B to $3B valuation. Wiz's acquisition by Google showed that a security specialist in a defined category commands a platform premium. Zenity's B-round is the first inning of that play.

There is a derivative-market subtext. Cyber insurance became a multi-billion-dollar industry because security tooling created actuarial data. Agent security infrastructure makes Agent liability insurance possible — a policy that prices machine behavior risk. Hitachi and LG may be positioning for that derivative product, not the tool itself. The flags on this round point at a market that does not exist yet.

The term sheet doesn't validate the product. It validates the category.

The Infrastructure Reality Check

This is not an AI-model company. Zenity's capital intensity sits in streaming data pipelines — API call logs, permission changes, data access events — and in multi-region deployment for data residency. Kafka-class message buses. Columnar audit stores. The GPU line item is negligible next to model-layer competitors. That distinction matters for capital allocation: the right spend for a security layer is sales engineering and compliance certifications, not compute.

Re-reading my 2020 DeFi liquidity mapping work — where 60% of apparent organic volume on early yield forks turned out to be insider wash trading — I recognize the same baseline challenge here. Anomaly detection is only as good as the behavioral baseline. Zenity's moat will be built from enterprise baselines: what a normal Agent interaction looks like inside a specific industrial, legal, or healthcare workflow. That data cannot be scraped from the public internet. It must be co-built with customers. Slow, unglamorous, defensible.

Contrarian

Correlation is not causation. The largest risk is not competition. It is the feedback loop this funding creates.

Security tooling does not prevent agent failure. It makes enterprises comfortable deploying agents faster. Every compliance dashboard Zenity sells accelerates deployment of the thing it is supposed to protect. The insurance analogy cuts both ways: insurance does not reduce accidents, it prices them and redistributes the cost. You are not buying safety. You are buying a velocity permit.

The behavioral monitoring problem is political before it is technical. A platform that watches every agent interaction also watches every employee's digital exhaust. That is a procurement risk disguised as a privacy feature. EU data minimization principles and the spread of privacy-preserving modes will determine whether this category grows in the open or under consent decrees.

The uncounted scenario is agent-to-agent delegation. OpenAI's agent protocols are already moving toward machine-to-machine invocation. When one agent calls another, audit trails fragment across ownership boundaries. No vendor in this space has publicly demonstrated working cascade forensics. That is a blind spot the size of the next headline.

And the honest math problem. The addressable market assumes Gartner's 15% agentic workload share materializes. If agents remain in pilot purgatory — and most enterprise AI deployments still are — the security layer becomes a solution searching for an incident. The bear market doesn't spare narrative-driven software. It only delays the reckoning until the next funding round.

The missing American VCs remain the datum I cannot stop re-reading. A billion-dollar narrative with no U.S. institutional validation is a half-corroborated story. In a category still being defined, the absence of the funds that backed Wiz's run is a fact, not a footnote.

Takeaway

Over the next twelve months, three signals carry the information.

Watch whether Zenity ships verifiable agent audit trails — evidence mapping a machine behavior chain to immutable records. Agent-to-agent transactions will require an evidence layer neither party can edit. Autonomous wallets taught me that machine actors generate their own forensic structures. Security tooling must match them.

Watch whether a platform vendor bundles or acquires. That single move defines the category's fate.

Watch the ARR disclosure at the next round. Narrative converts to numbers, or it converts to ash.

The step-up was the thesis. The next term sheet will tell us whether this was a trust bottleneck or a trust bubble.