Kao Data’s newly appointed Power and Energy Director, Michael Gordon, has decades of experience in the energy sector and is now applying that expertise to data centres including the ongoing reforms to grid connectivity. 

Faced with seismic change, the UK is overhauling its process for deciding when and how new large demand and generation projects will be assessed for eventual grid connection.  

The National Energy System Operator (NESO) introduced ‘Connections Reform’ in April 2025 which includes significant changes to how projects enter and progress through the connections process. These reforms aim to “remove unviable projects from the queue and prioritise those critical to delivering clean power and wider industrial decarbonisation,” according to NESO. A key part of this reform is the introduction of the new “Gate 2 to Whole Queue (G2TWQ) process” which aims to ensure that only projects meeting specific criteria move ahead. 

Working out what all that means for the data centre industry and its power availability challenges, takes a special blend of commercial and regulatory insight. Fortunately, Kao Data has just that combination of skills in its recently appointed Power and Energy Director, Michael Gordon.

Joining Kao from stints at National Grid and Transmission Investment, Michael is a relative newcomer to the data centre industry but is well placed to use his outsider’s perspective and energy industry know-how to navigate the twisting path ahead.  

 

To kick-off, can you summarise your background in the energy industry and describe the remit of your new role at Kao Data? 

I started out in the power industry back in 1997 after graduating as an electrical power engineer from Manchester University. My early career was spent working directly with power station companies in the UK. A few years into my career, I transitioned over to National Grid, the electricity transmission company, initially working on-site in London before moving down the road to the national control room. I spent two separate stints in the control room, totalling about six or seven years managing the country’s electricity transmission system on a 24/7 shift pattern. It was a technical role – dealing with on-site engineering teams, switching the high-voltage network, analysing stability, and directly responding to major system faults. 

Now at Kao Data, my role is to apply my engineering and power industry background directly to our digital infrastructure pipeline. It’s been a big learning curve interfacing with a different side of the industry, but my core remit is about working out what the current grid limits are, understanding our options, and presenting strategic pathways to progress Kao’s multi-megawatt data centre developments. 

How does the state of the UK’s grid, from an innovation and resiliency perspective, compare to neighbouring countries in your view?  

I don’t think that the wires themselves, or the way the system operator works, is particularly behind in any sense. People say all the assets were built in the 60s and the 70s and they’re old and that’s why we can’t connect data centres. But you know assets that were built then they’ve been added to and maintained and replaced where necessary. I was on shift at National Grid when one of the biggest peak demands hit, which was like about 60 gigawatts on a winter in the early 2000s and since then demand has gone down. So, the need to expand the network wasn’t there. I don’t think NESO are particularly backward in their ways of thinking, or the network has been left to go to rack and ruin, and therefore that’s caused a problem. But I think they are really playing catch up now with something that’s probably caught even the data centre industry by surprise.  

What are the specific challenges with the speed of data centre expansion?   

Data centres get built way quicker than a new overhead line; new overhead lines take like ten years in the planning and construction. These companies aren’t used to working that quickly, so they’re trying to change. Another part of it is the government. Can the government accelerate the way networks can be built out? Can they prioritise planning for these new lines and say, look, this is critical and it’s going to have an accelerated plan. So, I think, does it have to change? Yes, but is there anyone miles ahead in terms of network? I don’t think so. I think the US is ahead because it’s more deregulated. You can put a gas power station up more easily. Enabling new data centres is put above other things, and so I think the regulatory regime over there is very much around accelerating, and we don’t particularly have that here. But I don’t think that grid is way better than ours or anything.  

Let’s drill into the Connections Reform process and your view on whether it has been a positive or negative for adding new data centre capacity?  

Under the old process anyone wanting to connect to the National Grid – generation or large demand – would go to the grid company and have an informal discussion at first. Then they could apply to the grid company, and the grid company was obliged to provide them with an offer to connect within three months. During that process there was a dialogue between the applicant and the grid companies and so at the end of it, you’d end up with a date and certainty. It may be good, it may be bad, but you got your answer quite quickly, and you had the dialogue. 

So why did that system need to change?  

The reason was just the volume of those applications was getting way too much for NESO (National Energy System Operator) to respond in a coordinated way. Secondly, the sort of argument was actually that the UK needs to transition from where we are to an electrified future, and to do that it needs more coordination. So National Grid we’re tasked with actually saying, “Well, where does the country need things, where does it need battery storage, where does it need more wind?”  

Around 18 months ago, they moved to a process where two things happened. First, you can’t apply at any time of the year anymore; there are going to be two application windows a year. Second, if you do apply, they’ll do a quick assessment. If you’re not ready enough, you’ll be put into Gate 1, which won’t really give you anything concrete. But if you’re a more ready project, they’ll put you into Gate 2, and then they’ll take you seriously and assess you against clean power. The start of that in earnest was when they had to take the existing queue and close it to new applicants. Everybody threw their applications in for generation at the start of this, creating a massive queue to assess which projects would progress first and which ones would be put on hold. A lot of what has happened over the last 18 months has been me focusing on that. 

For large demand projects, they also introduced ‘transitional offers’ that just give a rough region and timescale. Many applicants been sitting in limbo and some have been told they are Gate 2 and Phase 2 (post-203) but are still waiting on the actual detailed offers. In the old world, you knew within three months. Now, a lot of projects don’t know their exact timescales or even the specific substation where they will connect because the operator has kept everything generic at the initial stage. 

Since we spoke, demand reform has moved on further. Ofgem has now opened a consultation on proposed data centre connection reforms, including potential for significant commitment fees and new queue management milestones. 

How much have the new reforms elongated the connection process do you think?  

I think what used to take three months has now taken, I don’t know, 18 months or two years to get the answer. But the caveat to that is the enduring process won’t be as bad. I think part of it is there’s a big chunk of dead time because they’re transitioning to the new process, theoretically for next year onwards. What they’ll do is you’ll be in a position whereby they’ll run two windows a year, where you can apply, get your answer. It won’t be 18 months, two years in the future, but it won’t be like the three-month turnaround, apply when you want, but it won’t be as bad as it has been.  

What are the upsides of the reforms?  

Reform theoretically should improve the time to connect for some. If it whittles out the ‘zombie projects’ or people who are blocking other developments from the queue, then it has the potential once done to bring connections forward. Alternatively, if they encourage some of the flexible options, that could also bring connections forward. I think that’s the sort of prize 

Can you explain a little more about these “zombie projects”?  

The term ‘zombie projects’ was coined to describe those projects—starting with generation—that would just sit in the queue. Historically, as their connection date neared and they had to put money down, developers would just ‘mod app’ to push it back. They’d keep doing that every couple of years to keep the project alive on the cheap while blocking new developments. 

Generation reform has worked to reduce that through Gate 1 and Gate 2 filters and clean power plan prioritization, and it has gone some way toward knocking out projects that were holding things up. We haven’t even implemented it for demand yet, though the intention is the same.  

How do you think the reforms could be improved?  

I would hope we go back to a process where you can interact and have discussions along the way to fill the gaps between the prolonged formality. That ability to engage would be the biggest improvement to me—to actually speak to somebody. At the moment, you put your queries on a portal, they may or may not get answered quickly, and it’s all a bit hands-off and removed from being able to have a real conversation. 

What’s your take on the value and viability of demand response for data centres?  

The idea behind it is good. If you think about connecting, say, 10 gigawatts of AI or large demand, that is a hell of a chunk of load. What NESO has to do is secure the system at any given second. If all of a sudden some generators fall off and system frequency drops, having a chunk of that demand respond commercially—by reducing output or running their diesels—can only be a good thing. It means the system can be managed, and ultimately NESO can tolerate the connection of more data centres. Data centres having a degree of flexibility to respond in urgent circumstances, without upsetting their core business, is the right thing to happen. 

In the current demand connection reform, they are considering ‘flex.’ Kao’s view—and my view—is that it shouldn’t be mandated flex, but the option to have it. If you sign up to it and it accelerates your grid connection, then everyone wins. It’s about having the choice. You might have a certain type of customer that wants absolute 24/7 zero interruption, and they should be able to apply for that and possibly wait longer. But if you offer flexibility, you should be able to get a contract that brings benefits, like connecting earlier. 

What do you think about adding capacity in the form of small modular nuclear reactors?  

A few companies working in this area have indicated that the UK regulator is supportive, and asserts that the proposed SMR technology is inherently safe. However public perception is still an important factor, but the technology is pretty mature; it has existed in nuclear submarines for years and it’s eminently doable. 

Finally, there’s a lot of focus right now on the energy consumption of AI but what about the flip side – the use of AI to manage the grid and improve resiliency and efficiency?  

I’m a power system engineer, and analysing the grid is what I’ve done for 30 years. My take on it is that it’s early stage, but it’s got potential. How much it delivers? I haven’t really thought about, but it will undoubtedly get introduced. The simple version of the problem is that to manage the grid and do the study and analysis, you have to make a set of assumptions. They’re probably just two or three assumptions, and they’re probably oversimplified. Therefore, because you follow them, you’re a bit conservative and you don’t maximize the use of the grid. The argument is that AI can look at all sorts of permutations very, very quickly and respond faster, allowing you to utilize the grid in a more efficient way. I think there’s mileage in that. What it looks like and how it works, I haven’t given detailed thought to, but it’s definitely an area that people are talking about.