Posts

Posted on

Latest reporting: Self-Improving AI Could Drive Innovation – But Strain Data Centers

My latest reporting for DataCenter Knowledge.

Recursive self-improvement is touted as AI’s next major milestone. If it’s ever achieved, the impact will be felt across the data center industry and far beyond.

Recursive self-improvement (RSI) describes AI systems that can design, code, and deploy their successors with minimal human input. If achieved, improvements could compound rapidly as smarter models beget even smarter generations – potentially accelerating beyond human comprehension and, worryingly, control. For data center operators, RSI would reverberate across power, cooling, orchestration, governance, and certification.

Although RSI is loosely defined, much like artificial general intelligence (AGI), there are signs it’s edging closer to reality. Achieving RSI depends on fast advances in AI-assisted coding, which labs say are already underway. For example, frontier AI model creator Anthropic, in its report titled “When AI Builds Itself,” says its Claude model now authors a substantial proportion of its own codebase: “As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude. Before Claude Code launched in research preview in February 2025, this number was in the low single digits.” If sustained, this kind of contribution is a step toward systems that can improve themselves.

For more, head to DataCenter Knowledge.

Posted on

After the AI Rush, Can Data Centers Reclaim Sustainability?

My latest reporting for DataCenter Knowledge.

Sustainability’s legacy was sidelined by AI growth and politics, but public scrutiny and hard economics are forcing a course correction.

Data centers are where compute turns into real electrical load, heat, and water demand, and where efficiency and resilience are engineered under grid, thermal, and permitting constraints.

For operators, power, cooling, water stewardship, air quality, and building performance aren’t background considerations but instead define the operating envelope. At the megawatt scale, every kilowatt and liter translates into cost, risk, and compliance. Because data centers are the physical manifestation of the cloud – sited in communities and tied to local grids and water systems – they carry a clearer and more enforceable environmental mandate than the software stack they host. That visibility has brought sharper regulatory oversight and pressure from environmental groups over the years.

For more, head to DataCenter Knowledge.

Posted on

Latest reporting: What Britain’s Next Prime Minister Could Mean for UK Data Centers

My latest reporting for DataCenter Knowledge.

Incoming UK Prime Minister Andy Burnham appears to favor homegrown AI, but planning disputes and data center moratoriums could cause early friction.

Incoming UK Prime Minister Andy Burnham has established an informal dress code that owes more to US West Coast tech execs than the blue-suited standard issue of British politics.

Whether his Silicon-Valley-casual-style – including a black T-shirt reminiscent of Steve Jobs’ trademark black turtleneck – extends to his technology policies has yet to be determined.

Burnham, who has previously held cabinet positions including Secretary of State for Health, most recently served as Mayor of Greater Manchester before standing as a member of Parliament to challenge outgoing UK PM Sir Keir Starmer. He will replace Starmer as leader of the Labour Party this week before being appointed UK prime minister on July 20.

For more, head to DataCenter Knowledge.

Posted on

Latest reporting: The great datacentre backlash: The industry response

My latest reporting for Computer Weekly

In part two of a two-part series looking at attitudes to datacentres, we look at how developers and operators have responded to the wave of anti-datacentre sentiment sweeping through communities and what their plan is to avoid further escalation

When it comes to artificial intelligence (AI), it is all about the tokens, which are the basic units of AI. They’re the way large language models (LLMs) – the core of AI and therefore behind much recent datacentre expansion – measure inputs and their responses, with their cost measured in fractions of a dollar. 

But “token” is also how many datacentre campaigners would characterise the industry’s record stance on community engagement and sustainability. If the industry is to avoid escalation and impacts from protests on build-times and project viability, experts agree there needs to be a shift from a perception of tokenism to a community and socially focused stance. 

In part one of this series, we looked at the various groups, from global to grassroots, that actively campaign against datacentre development. 

In part two, we look at the datacentre industry’s response to this backlash – namely, the operators, LLM labs, technology suppliers and numerous consultants and other constituents of the supply chain that are almost singularly focused on speed, scale and performance.

For more, head to Computer Weekly.

Posted on

Latest reporting: The great datacentre backlash: The campaigners

My latest article for Computer Weekly.

In part one of a series looking at attitudes to datacentres, we look at the organisations that oppose new builds, concerns and motivations, what the industry thinks and what solutions might resolve the various impasses

Just before 1.00am on a Monday in early April 2026, Indianapolis city councillor Ron Gibson awoke as 13 gunshots hit his house. When safe to do so, he found a note on his doorstep that said, “No data centers”. Gibson had apparently been targeted for his support of a proposed multimillion-dollar datacentre project in one of the districts he represents. 

That’s a rare and extreme happening. But anti-datacentre sentiment and campaigning exists and has accelerated in direct proportion to the wave of datacentre expansion and artificial intelligence (AI) ambitions we now live in.

In the first article of a two-part series, we look at opposition to datacentre development, the types of opposition, what the concerns are, how the industry views campaigners and the potential solutions.

For more, head to Computer Weekly.

Posted on

Latest reporting: AI to Transform Data Center Operations – But Not Overnight

My latest reporting for DataCenter Knowledge.

EU agency managing critical IT systems shares its vision for AI in government data center operations.

AI is emerging as a “transformative force for optimizing operations,” according to a report from the organization that oversees much of the European Union’s digital infrastructure.

The Energy-Efficient Data Centers report, from the EU Agency for the Operational Management of Large-Scale IT Systems in the Area of Freedom, Security and Justice (EU-LISA), covers a range of issues from the regulatory landscape to current best practice.

Areas of innovation, including the use of AI to improve the management and operations of data centers, are also examined in the report. Generative AI is a critical workload for data centers, driving key technological shifts, such as the adoption of liquid cooling and the use of graphics processing units (GPUs). But the use of other forms of AI inside the data center has arguably received less attention.

For more, head to DataCenter Knowledge.

Posted on

Latest reporting: AI Boom Threatens Years of Data Center Resiliency Gains

My latest reporting for DataCenter Knowledge.

An industry rooted in resiliency continues to uphold its core tenets, as data center outages decline for the fifth consecutive year, according to Uptime Institute’s latest annual research.

The 2026 Data Center Outage Analysis, published this week by Uptime’s research unit, examines downtime across the industry using a combination of the tier certification organization’s own survey, press reports, company statements, and other sources.

However, a recent AI-driven pivot that appears to prioritize performance over resilience raises questions about how long the overall positive outage trend will continue.

For more, head to DataCenter Knowledge.

Posted on

Why AI being in the trough of disillusionment isn’t bad news

Troughs usually precede inspiring slopes and productive plateaus

Troughs can be good for some

Words matter. But so does context.

Terms like ‘trough of disillusionment” from Gartner’s (in)famous Hype Cycle appear to have obvious negative implications. 

Being in any kind of trough can’t be a good thing right? (farmyard animals and drunken cowboys from 1950s films and TV aside)

But a little bit of analysis, or simply engaging with the detail, reveals a more nuanced story. And let’s face it, amid all of the current hype around Gen AI, a little bit of nuance should be welcome. 

As Gartner’s prognosticator in chief, or to give him his correct title Distinguished VP analyst, chief of research, John Lovelock explained it this week: 

“For AI we are in the trough of disillusionment this year,” he said speaking on a webcast to explain Gartner’s latest quarterly IT spending predictions. “The lowest point for expectations. But let’s be careful to remember that expectation is the vertical axis of the hype cycle. It is not what the technology can do, it is what the users of the technology expect it to do and right now it is at a low point.” 

Gartner Gen AI Hype Cycle

So what Lovelock is explaining is that not that Gen AI is inherently bad or not delivering but rather that customers have adjusted their expectations based on their experience of AI to date. As Lovelock explains: “Even though we are in the trough of disillusionment we are seeing overall AI spending continue to grow. It will hit $4.7 trillion by 2029.” 

“It is not what the technology can do, it is what the users of the technology expect it to do and right now it is at a low point.” 

Gartner’s John Lovelock explaining how to interpret the Hype Cycle

 $4.7 trillion by 2029! That is not a small number. (for context, Gartner predicts that worldwide IT spending is expected to reach $6.31 trillion in 2026, up 13.5% from 2025). 

AI being in the trough, needs to be understood in the broader context of the Hype Cycle. Following the trough, there is fortunately a slope, then a plateau. Everyone loves a plateau – especially a plateau of productivity. 

But while Gartner was positive about the future promise of AI, there are some well documented challenges ahead when it comes to the small matter of return on investment (ROI). Lovelock went on to discuss the thorny issue of how hyperscalers will recoup the hundreds of billions invested into new data center capex over the recent past which is expected to continue into the near future at least. 

Not even including the data center physical infrastructure (power, cooling, building shell etc), there is a significant gap between projected future revenues from selling AI services versus the investment in AI IT infrastructure. 

Gartner AI ROI Gap
Gartner AI ROI Gap

“We notice there is a tremendously large gap,” said Lovelock. “If these hyperscalers are to meet the return on investment of 25% for just the servers and networking equipment they are going to need about $1.5 trillion dollars in revenue.”

“If they stuck to their current revenue models they would be in trouble,”

John Lovelock, Gartner

However, again, the nuance matters. Lovelock explained that if the hyperscalers were approaching AI in the same way as say cloud, they could be facing a write down on some of the capital investment by 2029, 2030. “If they stuck to their current revenue models they would be in trouble,” he said. 

But most hyperscalers appear to be diversifying their revenue streams, using AI servers to buy interests in new AI start-ups for example. The take-away message from Lovelock is that the dynamic and fluid nature of AI investments, adoption and monetization make simplification a dangerous game. “The market is tremendously complicated. Looking at it through a simplified lens could lead you to a simple outcome that is not likely to happen”. 

So there you have it. Words matter. And being in a trough is not necessarily bad if there is a beautiful plateau in your future too. 

The full webcast is available here. There is also some good analysis of the impact of AI on jobs which is also worth chewing over.

Posted on

Understanding AI’s Water Use: Definitions matter

When it comes to debunking nonsense numbers, you can’t do better than the BBC’s More or Less radio programme and podcast.

The team includes the economist and writer Tim Harford. They regularly take politicians and companies to task for crimes against statistics and address issues related to the accurate representation of data.

So it’s not surprising then that the More or Less team recently turned their analytical guns on the misinformation that is flowing (pun intended) around data center water consumption. (Note: the incidents highlighted in the episode actually date back to late 2025 or early 2026. The issue is obviously still very much alive and relevant).

The episode focused around a book called Empire of AI which stated that AI demand could drive up consumption of fresh water to 1.1 trillion to 17 trillion gallons (4 to 6 trillion litres) of fresh water per year by 2027. This amount is apparently equivalent to half the water consumed in the UK.

Consumption vs withdrawal

However, as More or Less explained, when it comes to complex topics such as AI and water usage, definitions matter. It seems that Empire of AI potentially conflated water consumption with withdrawal. The 1.1 to 1.7 trillion gallon figure was sourced from a University of California, Riverside study that was actually for water withdrawal rather than water consumption. There is an important difference:

+ Water withdrawal refers to the amount of water that gets taken out of system or source. Importantly, some of that water will be consumed but some of it will also be returned. (Measures demand)

+ Water consumption is a sub-set of withdrawal and refers to water that is taken out of the water system and importantly not returned. (Measures loss)

The discrepancy was highlighted by independent researcher Andy Masley who published a detailed analysis of the claims and potential discrepancies in his substack which is well worth reading for more depth on this topic. The author of the Empire of AI acknowledged the mistake and issued a correction. The correction also contained links to some other important sources on the topic.

However, according to the More or Less team, the story does not end there. The University of California, Riverside paper also has some potential issues, they claim. The water usage numbers in the paper are apparently based on an estimate for global electricity use in 2027. However, it seems the author of those numbers, Alex de Vries-Gao, was only basing his analysis on servers that could be deployed in 2027 and excluded all existing infrastructure.

So now you have an overestimate intertwined with a potential underestimate!

Some further analysis by Vries-Gao to estimate total AI server capacity, and eventual water consumption, postulated that AI systems at the end of 2025 were consuming water at a rate of up to 750 billion litres of water, according to More or Less.

Is that a big number?

As usual, the programme likes to always ask the question: “Is that a big number?”

The answer seems to be yes – in fact it could exceed the total global consumption of bottled water which is more than 450 billion litres, according to Vries-Gao. However, another important caveat is only about 10 percent of that consumption is happening on-site at data centres, the rest is mostly happening at power stations.

Also the original consumption vs withdrawal issue is important to consider. While they are different, withdrawal is actually a more important consideration than actual consumption.

Power availability: the ultimate bottleneck

The sign-off to the episode unsurprisingly actually focused on power availability rather than water consumption or withdrawal. Accurate estimates on future water use depend on projecting installed AI infrastructure capacity accurately. Growing power availability constraints are making these projections increasingly difficult to predict accurately.

It’s also important to remember that the More or Less programme was created for a generalist audience and the issue of AI and Data Center water use has a huge number of other variables including direct liquid vs air-based cooling, silicon diversification (GPUs vs other chip types and the heat rejection implications), training vs inference (centralised vs distributed AI).

What is clear is that we have entered the third-wave of sustainable IT and this time is very different. The previous two cycles – in the early 2000s and 2010s – were largely driven by the industry responding to governmental scrutiny. This time it is much more grass-roots and bottom-up with community groups often leading the charge. The industry is responding. However, it remains to be seen if the response will match the speed and scale of the actual data center build outs.

Posted on

Data Center Thought Leadership: Vertiv Frontiers

Vertiv Frontiers looks at the macro forces and technology trends shaping the future of the data center industry.

Proud to have finally published the Vertiv Frontiers report which looks at four macro forces and five key technology trends informing the future of the data center.

These range from Extreme densification, and Gigawatt scaling to Powering up for AI and Adaptive resilient liquid cooling:

After two decades of steady evolution – when cloud computing reshaped location and scale but core infrastructure remained largely constant – the next wave of transformation is accelerating at unprecedented speed.

Driven by AI and accelerated compute, this new era is redefining how digital infrastructure is designed, deployed, and scaled. The pace of change is unmatched, creating new possibilities to push the frontiers of innovation.

Vertiv Frontiers offers a lens on the future – an exploration of macro forces and the technology trends reshaping digital infrastructure.

It brings together the expertise of Vertiv specialists across power, thermal, IT systems, prefabricated modular infrastructure, advanced services, and AI infrastructure, reinforcing Vertiv’s position as a leading voice guiding the future of critical digital infrastructure.

The report was published with close cooperation from Vertiv’s chief technology and product officer Scott Armul, as well as other internal thought leaders such as Martin Olsen, Peter Panfil and Steve Madera.

Thanks to all of the other contributors from the Creative, Content and Web teams who helped to get this published.

The full report is available at Vertiv.com