Data Center Engineering Faces an Endless List of Always-Shifting Standards
A platform from Accuris provides intelligence that brings industry standards directly into PLM workflows.
May 12, 2026
At a Glance
- The platform reduces engineering research time from 30% to near-zero.
- AI-powered semantic search connects related standards across multiple regulatory bodies.
- The platform’s digital thread technology preserves decades of engineering knowledge as experts retire.
As AI-driven data centers transform from simple IT infrastructure into complex industrial tools backed by mechanical, electrical, civil, and environmental engineering expertise, the bottleneck isn't capital—it's knowledge. Engineers need instant access to evolving standards covering everything from 100-kilowatt GPU server racks to new refrigerant regulations.
Accuris addresses this challenge by offering contextual engineering intelligence within existing workflows, eliminating the research burden that traditionally consumes 30% of engineering time.
We caught up with Ben Tanner, head of product technology at Accuris to get the details on how the Accuris platform provides engineering knowledge.
Can you tell us about Accuris and what the company provides?
Ben Tanner: We're really about providing engineers the intelligence they need to build better products. We have two main product families: supply chain intelligence and what we call engineering intelligence. Engineering intelligence is about surfacing standards to engineers—whether it's ANSI, IEEE standards, or others—across industries like oil and gas, electrical, aerospace, and defense. But more than that, we give them actual intelligence from those standards as well.
How does the platform fit into existing PLM tools?
Tanner: We already have integrations in our parts space—electronic parts linkage into PLMs—which we've done since well before we were Accuris. But one of the spaces that's really interesting is how we play nice with things like PTC Windchill or the big dog, Teamcenter. That's an area we're working on heavily because engineers invariably spend about 30% of their time researching standards—literally going back and reading, swivel chair activities. The goal is to give engineers the intelligence they need where they're working, minimizing that swivel-chair switch. Those tools are a critical part of the engineering ecosystem. We want to get that information into the PLM platforms so you have it on a digital thread. If there's a standards change, engineers don't have to go hunt around because it's surfaced right in front of them.
Has the demand for data centers put extra pressure on access to standards?
Tanner: We're definitely seeing demand from our customers around this space—more engineers needing access to more standards for different facility types. When you think about a data center and the complexity involved, they're not just IT infrastructure anymore. My career started as a network engineer 20-something years ago—speeds and feeds and all that. Now, the modern data center, if I were to walk around one, there's no resemblance to one I was working in when I used to run BNC cables.
They've become these industrial facilities themselves with serious power and cooling standards, structural and fire safety requirements. That means mechanical, electrical, and civil engineering standards. The demands on that have outstripped what the industry can deliver. If we're going to keep pace, particularly with AI, standards are critical.
For example, look at the American Innovation and Manufacturing Act (AIM Act), which mandates a reduction of something like 85% in hydrofluorocarbons. They're used as coolant in cooling systems for data centers predominantly—they're ozone-damaging and potent greenhouse gases. One of the most common refrigerants in HVAC has been banned from new equipment since January 2025 as a result. Cooling is around 40% of the energy consumption of a data center now, so we've got to start transitioning to alternatives. That's a layer of complexity getting introduced.
Look at power delivery. A single server rack now could consume up to 100 kilowatts of power for a GPU-based AI server rack. We used to suck our teeth at 7 kilowatts with a couple of blade chassis. The standards around how we deliver that power safely, the structural requirements for the floor of the data center with the increased compute density—all of this is shifting. Engineers need standards to be able to keep up with that, and they need to understand how standards are changing and evolving, and it's happening quickly.
I think the bottleneck in data centers and the demand, particularly around AI, isn't so much around capital—there's so much money flowing into this space. They're projected to spend $67 billion building data center facilities this year across the US industry. It's really knowledge. You can break ground in months, but if your engineering team needs to know the right standards to build to. We give the engineers the standards to enable them to leverage their expertise to build these facilities quickly.
How are standards bodies and codes coping with new facility types with new performance requirements?
Tanner: Standards by their very nature change—they have to adapt. Standards exist for several reasons, but one of them is to prevent catastrophic failures. We only have to look at the cost of a failure, whether it's in the nuclear energy space or aerospace—recalls, compensation costs. There's a reason we have those standards.
Change in standards has to be deliberate and methodical because of the criticality of them—what keeps airplanes in the air, and which keeps pacemakers working. Standards are adapting, but the AI-driven demand is unprecedented. NIST has the NIST AI standards that have been evolving now, but trying to keep pace with the industry when there are different competing standards is challenging. You even have organizations like Microsoft trying to play in this space now.
With AI in particular, standards bodies are all consensus-driven organizations. AI and the facility requirements behind it are crossing a lot of traditional boundaries between electrical, mechanical, fire and safety, environmental, and then governance, which is getting mandated now at governmental levels. So it's taking them time to evolve to that.
If you think about building a data center, you're not just consuming standards from IEEE or NIST—you've also got ASHRAE for the environmental standards around cooling and humidity. You've got all these different standards bodies that all have to play in this space and, to some extent, play together. The challenge is when standards don't exist for the evolving space or there's just no connectivity between them.
That's one of the amazing things about what we can do—leverage our software to provide and surface the connectivity and the applicability between different standards. Our THREADS product is exactly that—it's identifying the requirements within, whether it's your own internal standards or some of the other standards bodies that work with us, to actually surface that. We've also got things like the Engineering Workbench with dynamic linking that helps engineers surface this stuff without having to manually chase all these cross-references.
The way I look at it, the bottleneck isn't in writing these standards. It's making the existing body of standards easier to understand, consume, and apply to the project you're working on.
Are standards particularly challenging for multinational companies?
Tanner: When you look at the various regulatory agencies and regulatory compliance, each regulatory agency will have a set of standards they expect you to adhere to. So if you're building a nuclear reactor in Europe versus building one in the United States, there are differences. What we're looking to try to do is really help customers understand that. If you're meeting this standard, can we start to think about surfacing what the common sets of standards are that you can leverage to really help yourself in this space?
If you know you're building a nuclear reactor in Alabama, then you have a combination of what Alabama wants and what that industry wants, and that becomes your world that you work from. You've got to conform to your federal regulations, your state-level regulations. Those regulations will stipulate that you need to follow standards like the boiler pressure vessel code (BPVC).
How are manufacturers capturing and transferring engineering knowledge as teams scale and evolve?
Tanner: Knowledge loss is one of the biggest unspoken risks that we face with both this manufacturing boom and the AI boom as well. Engineers spend about a third of their time searching for information. We even see this in the software space as we're developing things—you need to go and comply to an API standard or you're using a particular library. How do I start to surface the context of using that?
When an experienced engineer retires, you're losing decades of contextual knowledge—not just to the industry, but to your company as well. They'll have with them that knowledge of how to take a particular standard and apply it in a certain way or why we're choosing a certain material. You may even have some quirks or idiosyncrasies that are kind of baked into the knowledge corpus the engineers have. That all walks out the door.
This is where we talk about things like digital threads. How do we connect every engineering decision to its source? Steve's been an engineer here for 30 years and knows why we do things, but if that's not codified in some way, shape, or form through a digital thread, that knowledge leaves with Steve.
What we're trying to do is really start to create mechanisms around why engineers make certain decisions. Using things like our smart annotations, bookmarks, and things like that as we surface the standards, you can provide additional context. How do we build shared knowledge bases? We have our Projects folder concept for that—really just surface that information quicker.
What role does AI play in managing standards?
Tanner: AI has a massive role to play. How we use AI will vary depending on who the standards organization is. Some are supportive, and some are still trying to understand what the implications are. The risk is, when you think about what standards are doing and how critical they are to every industry you're working in, you don't want to have AI that misrepresents standards through hallucination.
I've had this myself when I've been doing some software engineering, where the LLM, whether it's Claude or something, will completely invent an API that just doesn't exist. You've got to have context and constraints when you're dealing with AI, and that idea of the digital thread again, so you can point back and reference every decision.
We do proprietary things with AI—I can't really talk about that because it's some of our secret sauce—but it's in the mix. We also have to reflect the fact that we have versions of standards going back to the '60s where there are scans of microfiches and things like that, because those things are still relevant for certain things that are out there in the world today. You still have to be able to surface some of that older information, whether it's an OCR of a printed document from the 1970s or a more modern one that's defined in modern markup languages.
I think where AI is great is things like semantic searching across the document library. You can have semantic understandings that a sheath is a form of coating. So if I'm looking for what I should sheath a particular material in to insulate it, well, that's a coating—that's the same as this. How do we start to build these graphs of different linguistic terms that apply to standards? That's really important. How do you effectively cross-reference when one standard uses a word like "coating," one uses "sheath," and another one uses "insulator"?
That's really where we can start to apply intelligence—to surface relevant context and requirements. That's how you can actually help engineers surface and understand relevant standards that they didn't even know about, because you understand that A is like B. This is why we call it portfolio engineering intelligence. It's not about just giving engineers access to standards like a library—it's about giving them that actionable intelligence or the context around those standards. But it has to go back to the standard, and that's the crucial thing with AI. You have to have that digital thread back to the standard to say, "This is this clause in the standard." You can't dilute that.
So in some ways, you're operating in the function of what used to be the special librarian at government and corporate organizations?
Tanner: Those roles actually still exist. You will have, for a particular project, a project librarian. I think it's almost like a combination of librarian, research assistant, and note taker. But those roles don't go away. I think what we do is we empower them, but we also put it directly in front of the engineer. It doesn't make a huge amount of sense for one person to have to be the person who maintains that corpus of knowledge. You need to have that more federated.
But the role of the librarian is there. They're the ones that will probably understand, "We need to go and buy a subscription to the BPVC or whatever." They'll understand the broader context, and then how do we take the more detailed information below that and start to surface it to the engineers?
If I look at the software world, which is my predominant area of expertise, we're not very good at surfacing standards. Some standards exist in the software world—we don't tend to surface them very well. Now, with the rise of agentic software development, things like Claude and Cursor are amazing. But if you can't provide your agent with the appropriate standards and constraints and context, it will produce garbage. You have to spend even more time supervising and repairing.
When you think about that and how that applies to all disciplines of engineering, being able to access the standards in a way that is contextually relevant to what you're doing is critical. AI is a crucial part of that, but it also accelerates the problem and the solution—you can't have one without the other.
Building an AI-certified data center, the level of standards and things that apply there vastly outpaced what we would have done 20 years ago. It's interesting, but at the same time, it solves the navigating-the-knowledge-space problem uniquely.