The OPC UA Moment: How the April 2026 Standards Announcement Changes Industrial AI Forever

On April 20, 2026, the organization that governs the dominant industrial data protocol announced it is converting all 430-plus of its industry information models into formats optimized for agentic AI. This is not a vendor announcement. It is a standards body announcing the native infrastructure for a new category of industrial software — and most of the industrial world has not fully absorbed what it means yet.

OPC UA · Industrial AI Standards · Agentic AI · ICS · SCADA · Sasquatch Labs Industrial Series · July 2026

What Actually Happened on April 20, 2026

Standards body announcements rarely qualify as news. They tend to be incremental refinements to existing specifications, process updates, or administrative changes that matter only to the organizations that spend their professional lives tracking them. Every so often, though, a standards body does something that changes the trajectory of an entire technology category — not because of what it says, but because of who said it and what infrastructure it commits to building.

April 20, 2026 was one of those moments for industrial AI.

Read that last phrase again: the semantic layer between agentic AI and the world of industrial automation. This is the OPC Foundation — not a vendor, not a startup, not an analyst firm — explicitly positioning OPC UA as the interoperability standard that connects AI systems to industrial data. It is a statement of architectural intent from the organization that governs the most widely adopted industrial data protocol in the world.

The industrial community has been waiting for clarity about where AI fits in the existing industrial standards landscape. This announcement provides it, and the implications cascade through every evaluation of industrial AI products, deployments, and roadmaps being conducted in 2026 and beyond.

What OPC UA Is — and Why This Announcement Carries So Much Weight

To appreciate why this announcement is significant, you need to understand what OPC UA actually is and how deeply embedded it is in the global industrial infrastructure.

OPC UA (Open Platform Communications Unified Architecture) is a standard that lets industrial machines, sensors, software, and systems exchange data with each other — securely and regardless of who made them. It is effectively a universal translator for the factory floor. A Siemens PLC, a Rockwell controller, a SCADA system, and a cloud analytics platform all speak different languages. OPC UA gives them one common language so they can share information without custom coding or proprietary converters.

OPC UA’s origins trace to 1994, when a group of software and hardware vendors in the industrial automation sector came together to form what is now the OPC Foundation. The original OPC standard solved an immediate problem — allowing industrial devices from different manufacturers to share data — but was built on Microsoft’s COM/DCOM technology, making it Windows-only, difficult to secure across firewalls, and hard to use across networks. As Industry 4.0 began to take shape in the mid-2000s, manufacturers needed something better. The OPC Foundation responded by designing OPC UA from scratch with a clear mandate: keep everything that worked about OPC Classic, discard the Windows dependency, and make security non-negotiable from the first line of code.

The result is a protocol that has become, over the past fifteen years, the closest thing industrial automation has to a universal data layer. OPC UA is standardized as IEC 62541, a multi-part international standard that defines the complete architecture of the protocol including communication services, security mechanisms, data modeling, and information exchange between industrial systems. It is what PLCs, historians, SCADA systems, and HMIs use to expose their data in a way that other software can query — platform-independently, across vendors, with built-in security, and with a semantic data model that describes not just raw values but the meaning and relationships of what is being measured.

The weight of the April 2026 announcement comes precisely from who OPC UA is. This is not a newcomer proposing a new standard. It is the organization that governs the protocol that the vast majority of modern industrial control systems already use, announcing that it is building native AI interoperability into that protocol’s knowledge base. The infrastructure that makes agentic AI talk to industrial systems is not being invented from scratch — it is being added to something that is already installed in factories, utilities, chemical plants, and water treatment facilities around the world.

What Companion Specifications Are — and Why Converting 430 of Them Matters

The 430-plus companion specifications being converted are the detail that makes this announcement technically significant rather than just symbolically meaningful. Understanding what they are explains why the number matters.

OPC UA gives machines a common language. But a common language is not enough if everyone uses different vocabulary. A packaging machine and a robot arm both speak OPC UA — but how they describe themselves, their data, and their capabilities needs to follow a shared standard too. That is what companion specifications do. They define standardized data models for specific industries and device types, built on top of OPC UA’s core framework. They are the reason a packaging machine from Vendor A describes itself the same way as one from Vendor B — making integration plug-and-play instead of a custom engineering project.

In other words: OPC UA provides the protocol layer — the common language. Companion specifications provide the semantic layer — the shared vocabulary that gives meaning to the data within that language. A tag named “MotorSpeed” on a machine from one vendor means the same thing as the same tag on a machine from a different vendor when both implement the same companion specification. That semantic standardization is what makes it possible for software — including AI software — to understand what it is looking at without needing to be individually configured for every device from every vendor.

The scope of the companion specification library reflects thirty years of industry collaboration. Examples span virtually every sector of industrial automation:

When all 430-plus of these specifications become RAG-optimized and MCP-compatible, an agentic AI system that connects to an OPC UA server does not just see raw tag values. It can understand, through the companion specification vocabulary, what those tags represent, what asset they belong to, how they relate to each other, what their valid ranges are, what state machines govern the equipment behavior, and what industry-standard interpretations apply to their current values. The difference between an AI reading a raw number from a tag and an AI understanding that number in the context of a PackML state machine on a packaging line is the difference between pattern matching and comprehension.

The Technical Formats: RAG, MCP, Vector Embeddings, and Offline LLMs

The OPC Foundation’s announcement is specific about the formats being produced, and those specifics are worth understanding because they reveal the full scope of what is being built.

The prototype repository already demonstrates how OPC UA core specifications have been transformed into AI-ready assets such as Markdown files, image descriptions, token-optimized RAG chunks, vector embeddings, and interfaces for MCP and REST-based querying. Each of these formats serves a distinct use case in the industrial AI ecosystem:

Markdown files and token-optimized RAG chunks

Retrieval-augmented generation is the technique that allows an AI system to retrieve relevant information from a knowledge base at query time and incorporate it into its reasoning, rather than relying solely on what was in its training data. Token-optimized RAG chunks are structured specifically to maximize the amount of relevant information that can be retrieved within the context window constraints of current LLMs. OPC UA specifications converted to this format mean that when an agentic AI system is asked about a specific piece of equipment, it can retrieve the relevant companion specification content in real time and reason about the equipment’s data in the context of its industry-standard definition.

Vector embeddings

Vector embeddings allow semantic similarity search — finding specification content that is conceptually related to a query even when the exact terms do not match. This enables the kind of contextual reasoning that makes an AI useful in a real plant environment, where operators do not always know the exact specification terminology and need the system to bridge between plain-language questions and standardized technical definitions.

MCP interfaces

The Model Context Protocol is an open standard that allows AI systems to connect to external tools, data sources, and services in real time. An OPC UA companion specification exposed through an MCP interface becomes queryable by any AI agent that supports the protocol — including during an active conversation with an operator, without requiring the specification content to have been pre-loaded into the model’s training data.

Offline LLM support

Perhaps the most important format for the industrial audience is the download option for use with offline LLMs. Many OPC UA users are training specialized models and cannot depend on links to the public internet. The new Online Reference provides multiple options including XML for technical specifications, Markdown for the largest amount of information in the fewest tokens, and JSONL as a RAG chunk format supported by local deployments. This is the air-gapped AI use case made explicit — organizations that cannot connect their AI systems to external networks can download the specification content in the appropriate format and run their AI entirely within their own infrastructure.

The SQL Analogy: Why This Is a Category-Defining Infrastructure Moment

The most useful analogy for understanding what the OPC Foundation has set in motion is not from industrial automation — it is from the history of enterprise software.

Before SQL became a universal standard for querying relational databases in the 1970s and 1980s, accessing data in any given database required understanding the proprietary internals of whatever system held the data. Each database had its own query language, its own data model, its own conventions. Software had to be written specifically for each database it needed to talk to. The result was a fragmented landscape where integration was expensive, slow, and brittle.

SQL changed that. Once databases exposed a SQL interface, any tool that could speak SQL could reach any database — without custom integration work for every pairing. The tools that could speak SQL could suddenly serve every organization running any SQL-compliant database. The tools that could not had a shrinking addressable market.

OPC UA has been doing something analogous for industrial data for fifteen years — providing the common language that allows software to reach industrial systems without proprietary integration work for every device from every vendor. What the April 2026 announcement adds is the semantic layer on top of that common language: the vocabulary that allows AI systems to not just read industrial data but understand what it means.

The tools that are built natively on this infrastructure — that speak OPC UA natively, query companion specifications for semantic context, and operate within the standardized vocabulary the OPC Foundation is now formalizing — have a structural advantage over tools that are not. The question for any industrial AI product is: is it OPC UA-native, or is it an application that treats OPC UA as one of several input formats? The answer to that question will increasingly determine fitness for deployment in real industrial environments.

What Is Already Live: The OPC UA Online Reference MCP Server

The April 2026 announcement was not just a promise about future work. By June 2026, the OPC Foundation’s own Online Reference had already been updated to include an MCP server for agentic AI use.

The OPC Foundation noted that AI systems can now access, understand, and utilize OPC UA information more effectively than ever before through the updated Online Reference, which includes an MCP server for agentic AI and the ability to download specifications in multiple formats for use with offline LLMs.

This is not a prototype or a demonstration — it is production infrastructure that any AI system supporting MCP can connect to today. The OPC Foundation has, in effect, already published the API that allows agentic AI systems to query thirty years of accumulated industrial standards knowledge in real time. The Companion Specification conversion announced in April is the broader expansion of that same initiative across all 430-plus domain-specific information models.

The Hannover Messe 2026 Demonstration

Standards announcements and live demonstrations are different things. The Hannover Messe 2026 show provided the demonstration.

At Hannover Messe 2026, AI and MCP-enabled equipment was demonstrated live — gaining live values from an OPC UA-enabled pH meter. The demonstration showed the impact of OPC UA in allowing AI to automatically understand real-world data. The OPC Foundation’s own president highlighted this demonstration in the June 2026 presidential welcome, noting that it illustrated the core concept of what the companion specification conversion makes possible: an AI system that connects to an OPC UA device and, through the companion specification vocabulary, understands not just the raw value the pH meter is reporting but what that value means, what the valid operating range is, what the equipment expects the AI to understand about pH measurement in the relevant industrial context.

That demonstration — an AI agent reading a pH meter via OPC UA and understanding what it is reading — is not conceptually dramatic. A pH meter is not a complex piece of equipment. What is significant is the architecture behind it: no custom integration, no vendor-specific API, no manually written context for the AI to understand the equipment. The OPC UA companion specification provides that context natively, through a standardized interface, for any AI that speaks MCP.

Scale that to the 430-plus companion specifications covering robotics, injection molding, packaging, machine tools, oil and gas, pharmaceuticals, and food and beverage — and the demonstration’s significance becomes apparent. The pH meter at Hannover Messe 2026 is the proof of concept. The 430-plus specifications in conversion are the production implementation.

What This Means for Industrial AI Products and Buyers in Practice

The April 2026 announcement has concrete implications for how industrial organizations should evaluate, select, and deploy AI tools. Several things that were previously matters of architecture preference have become matters of standards alignment.

OPC UA nativity is now a meaningful evaluation criterion

Industrial AI tools that connect to OPC UA as a native data source — rather than requiring operational data to be replicated to a cloud platform first — are now operating on infrastructure that the dominant standards body is actively developing for AI interoperability. Tools that route industrial data through proprietary cloud platforms are working around the standard rather than with it. That is not inherently wrong, but it creates a dependency on the proprietary platform’s own roadmap rather than the standards body’s roadmap.

Semantic understanding is becoming standardized rather than bespoke

One of the most expensive aspects of deploying AI in industrial environments has been the work of teaching the AI system what the data means — which tags correspond to which assets, what the valid operating ranges are, how alarm states relate to equipment conditions. As companion specifications become RAG-optimized and MCP-accessible, that context is increasingly available through a standardized interface rather than requiring plant-specific configuration for every deployment. An AI system that can query the relevant companion specification at runtime has access to accumulated decades of domain expertise without requiring it to be manually encoded for each new plant.

Offline and air-gapped deployment is being designed into the standard

The explicit provision of download formats for offline LLMs in the June 2026 Online Reference update confirms that the OPC Foundation is designing for air-gapped deployment, not just cloud-connected use cases. This is a direct acknowledgment of the reality of the industrial market: a significant proportion of the organizations that use OPC UA cannot connect their AI systems to external networks, and the standards infrastructure needs to serve them as well as it serves cloud-connected organizations.

The integration burden for new industrial AI deployments will decrease

Currently, deploying an AI system in an industrial environment requires significant custom configuration to establish the semantic context the AI needs to be useful. As companion specifications become universally RAG-optimized and MCP-accessible, that configuration burden decreases for deployments in industries with mature companion specifications. An AI system connecting to a packaging line that implements PackML can query the PackML companion specification for semantic context rather than requiring a human to manually map the plant’s tag structure to a contextual knowledge base.

What It Means for Valak Specifically

Valak, built by Sasquatch Labs, is an agentic AI layer for HMI and SCADA systems that connects natively to OPC UA, AVEVA PI, and GE Proficy. The April 2026 OPC Foundation announcement is not about Valak — it is about the standards infrastructure that Valak, and products like it, operate on top of.

What the announcement does is validate the architectural decisions that OPC UA-native industrial AI products like Valak are built around. The OPC Foundation’s explicit positioning of OPC UA as “the semantic layer between agentic AI and the world of industrial automation” is a confirmation that building an agentic AI layer natively on OPC UA is the architecturally correct approach — not a design choice that will need to be revisited when the standards mature, but a foundation that the standards body is now actively strengthening for AI use specifically.

For organizations evaluating Valak or similar products, the OPC Foundation announcement provides an important piece of context: the semantic infrastructure that makes OPC UA-native agentic AI increasingly powerful over time is being built by the standards body itself, not by any individual vendor. A product that is OPC UA-native benefits from that infrastructure investment as it matures, without requiring the vendor to rebuild the semantic layer from scratch for each deployment.

Valak’s air-gapped, on-premises deployment model is also directly consistent with the offline LLM support that the OPC Foundation has built into its updated Online Reference. The explicit acknowledgment that many OPC UA users cannot depend on public internet connectivity is a recognition of the same deployment reality that Valak was designed around from the start.

The Timeline: Where Things Stand and What Comes Next

Frequently Asked Questions