Schneider Electric has published an IDC white paper on maintenance in AI-era data centres, arguing that calendar-based maintenance is no longer fit for purpose in many facilities.

The report says rising rack densities, multivendor estates and shortages of skilled technicians are forcing operators to rethink how they maintain critical equipment. It makes the case for condition-based maintenance, which uses monitoring and analysis of asset behaviour to identify faults earlier and reduce unnecessary service interventions.

Schneider Electric linked the findings to its EcoCare service model, which combines remote monitoring, expert oversight and predictive fault analysis. It said the approach shifts maintenance away from fixed schedules towards interventions based on equipment condition and operating limits.

IDC said the operational backdrop for data centre operators has changed sharply as AI workloads grow. The paper notes that rack power densities have increased from about 15kW per rack in standard data centres to 300kW to 600kW in AI-heavy compute zones, adding pressure on uptime and infrastructure resilience.

That shift is being compounded by the way operators are expanding capacity. According to the research, many are relying on existing installed bases, distributed campuses, on-site generation and brownfield strategies through mergers and acquisitions of local service providers, rather than building entirely new facilities.

Operational strain

The white paper also highlights the complexity of fragmented multivendor environments. Operators that acquire existing facilities can inherit equipment from multiple suppliers without a full operating history, creating challenges when integrating it into asset performance management systems.

“When operators acquire existing facilities rather than build from scratch,” said Luis Fernandes, Senior Research Manager, IDC, “they introduce unknown equipment configurations from multiple vendors, with no operational history, requiring immediate integration with asset performance management systems.”

Labour shortages add to those pressures. The research said the supply gap for skilled technicians has reached unsustainable levels, citing a US example where there is only one qualified person taking up a position for every seven open roles. Operators are struggling to recruit across electrical, mechanical cooling and commissioning roles, including positions that require specialist certification for high-voltage systems.

Against that backdrop, the study argues that fixed maintenance intervals are becoming less suited to the realities of AI-led data centre operations. Rather than carrying out work simply because of a date on a calendar, condition-based maintenance uses equipment data to determine when intervention is actually needed.

Schneider Electric said early adopters of AI-supported condition-based maintenance have reported fewer manual interventions, lower operating expenditure, less unplanned downtime, longer asset lifetimes and better efficiency. It added that its EcoCare offering can deliver up to a 75% reduction in unplanned downtime and a 20% reduction in operating expenditure, while also reducing risk.

Predictive model

Jerome Soltani, Global Head of Services at Schneider Electric, described the model as one focused on identifying abnormal behaviour in equipment and systems earlier. He said combining remote monitoring with AI-assisted orchestration can improve visibility into asset health and reduce disruption from unnecessary maintenance activity.

“By combining remote monitoring capabilities with AI-assisted orchestration, you can gain insights regarding the health of your assets and systems, and get an early identification of abnormal behaviour that might precipitate a failure,” Soltani said.

“This ensures that downtime is minimised, but also that equipment working within specification is not disturbed or needlessly addressed.”

IDC frames the issue as part of a broader shift in how operators manage infrastructure in more complex environments. Instead of treating maintenance as a routine schedule, the paper describes a model in which software-led analysis and human oversight combine to create a more continuous picture of system health.

Fernandes put that argument directly: “Your maintenance schedule doesn’t know when something is failing – your equipment does.”

He added: “Condition-based maintenance is an optimised operating model for AI-era infrastructure that reduces manual interventions, lowers OpEx, and extends asset lifecycle. By scaling predictive analytics to correlate behaviour across every vendor, asset, and failure trajectory, condition-based maintenance enables operators to build machine-driven, human-validated system intelligence.”

KAREN JOY BACUDO

Finance Editor

The UK raised USD $1.8 billion across 181 FinTech deals in the first half of 2026, keeping its position as the world’s second-largest FinTech investment market, according to Innovate Finance.

The UK also led all European markets during the period, even as global FinTech investment fell to USD $28 billion from USD $32.5 billion in the previous half year. That marks a 12% decline worldwide, compared with a 5% fall in the UK.

The data points to a more selective market for FinTech funding, with artificial intelligence attracting a larger share of venture capital. Global venture capital investment in AI reached more than USD $400 billion in the first half of 2026, more than 50% higher than the total invested in AI during all of 2025, the industry body said.

The US remained the largest FinTech investment market, raising USD $17.2 billion in the first half, up from USD $15.6 billion in the previous six months. India ranked third globally with USD $1.5 billion across 122 deals, while France and Singapore completed the top five with USD $1.3 billion and USD $0.6 billion, respectively.

For the UK, the figures suggest a steadier performance than the wider market despite tighter fundraising conditions. The largest UK FinTech deal in the period was Ebury’s USD $203 million raise.

Global rankings

The largest individual FinTech deals of the half-year were concentrated outside the UK. US-based Ramp raised USD $750 million, making it the biggest FinTech funding round globally in the period.

France’s Alan secured USD $554 million, while India’s CRED raised USD $500 million. Mexico-based Plata attracted USD $405 million, and US retirement savings platform Vestwell raised USD $385 million.

Elsewhere, Canada and Mexico each recorded about USD $0.5 billion in FinTech investment. The UAE attracted USD $0.4 billion, and Germany raised USD $0.3 billion.

AI focus

The report also included a first-time analysis of AI investment in the UK alongside FinTech funding. On that measure, the UK ranked third globally, behind the US and China.

The comparison highlights how investors are allocating more capital to AI across the technology sector, even as specialist segments such as FinTech face a slower funding environment. For UK investors and founders, that may help explain why the country’s FinTech sector held its global standing despite a lower total.

Innovate Finance used data primarily from PitchBook, supplemented by Beauhurst and its own analysis. The study covered venture capital equity investment in FinTech and excluded debt capital raises.

“FinTech remains one of the most important applications of AI, and continues to attract significant investor interest. In H1 2026, UK FinTech has once again outperformed the wider market, retaining its position as Europe’s leading FinTech hub, and second globally. That resilience reflects the depth, maturity and international competitiveness of the UK’s outstanding FinTech sector. It is also a testament to the UK’s leadership in technology more broadly that we have claimed third position globally for wider AI investment,” said Janine Hirt, Chief Executive Officer at Innovate Finance.

Marc Lewis has launched SCAFFOLD, an AI platform for creative professionals who want to build and keep a personal AI trained on their own creative process.

Lewis, Dean of the School of Communication Arts in London, created the platform in response to concerns that widely used AI tools are making creative work look and sound more alike.

SCAFFOLD is designed for freelance creatives, in-house teams and agencies. Through a structured conversation with MarcAI, an AI model trained on Lewis’s coaching method, the system maps a user’s thinking, tastes, preferences and working habits into what it calls a Blueprint.

That Blueprint is then turned into an Exoskeleton, a personal AI agent intended to work alongside the user on live briefs. The agent can also work across the AI tools a user already relies on, rather than locking their work into a single platform.

The launch comes amid a wider debate in marketing, advertising and design over whether generative AI is eroding distinction in creative output. As brands increasingly use standard AI tools to produce copy, images and video, the concern is that their content will begin to converge with that of competitors.

Research from Kapwing, cited by the company, found that 59% of videos shown to new TikTok accounts in the platform’s For You feed were classed as “AI slop”. The same research found that rate was roughly three times higher than in a similar analysis of YouTube.

Ownership model

A central part of SCAFFOLD’s approach is ownership. Users keep the Blueprint and Exoskeleton they create even if they stop paying for the service, unlike subscription software models that keep access to user-trained systems and data within the provider’s platform.

The self-paced online version, SCAFFOLD Build, is priced at GBP £28 a month. The company also offers live coaching workshops with Lewis.

Lewis said the decision to let users keep what they build was deliberate.

“Most of what sits on your desktop, you rent, you don’t own it. And the day you stop paying, it locks you out and keeps everything you put inside it,” said Marc Lewis, Founder and Chief Executive Officer, SCAFFOLD.

He added: “The obvious, lazy, deeply profitable move would have been to keep that on our servers and rent it back to you forever, but we couldn’t do it.”

Creative process

Rather than relying on a large archive of past work to tune an AI model, SCAFFOLD is built around a guided two-hour session designed to capture how a person approaches creative decisions. The method draws on Lewis’s 15 years of coaching at the School of Communication Arts, as well as principles from cognitive science, according to the company.

The idea behind the method is that creative identity is shaped not only by outputs but also by judgement, taste, vetoes and habits. In practice, that means the system is intended to reflect how a user thinks through a brief rather than simply imitating finished work.

Lewis framed that as the rationale for the platform.

“AI hasn’t lived. It hasn’t danced. It hasn’t been dumped at 2am and then sat in a kebab shop at closing time trying to make sense of its life. That is where real creative work comes from and no model has it. SCAFFOLD keeps the human in charge of the machine. It learns your taste and your process, then does the grunt work in your voice rather than flattening you into everyone else’s,” said Lewis.

Lewis has worked in advertising education and creative coaching for more than a decade. Earlier in his career, he also founded and sold an internet technology company. His role at the School of Communication Arts has given him visibility across the advertising sector at a time when agencies and brand teams are rapidly testing AI tools for campaign development, ideation and production.

SCAFFOLD enters a growing market of services that promise to personalise AI for professional work. It aims to stand out in two ways: training the system through structured conversations about a user’s decision-making, and letting the resulting AI asset remain with the user rather than the platform.

For creative workers concerned that automation may standardise their output, the proposition addresses a specific fear: that faster production can come at the cost of a recognisable voice. SCAFFOLD’s answer is to make that voice the thing being modelled and retained.