Al is turbocharging demand for digital infrastructure. It is also an essential tool for managing this exponential growth, writes Amy Carroll
The explosion of artificial intelligence is among the most transformative phenomena of our times. AI could contribute up to $15.7 trillion to the global economy in 2030, according to PwC research, more than the current output of China and India combined.
The resulting escalation in demand for data centre capacity and fibre is well documented. AI requires extensive data processing and storage facilities, concnected by fibre optic networks, fuelling a huge and growing opportunity set for digital infrastructure investors.
Meanwhile, AI is also proving to be a powerful asset management tool, allowing digital infrastructure owners to improve efficiency and add value to the data centres, fibre networks and towers in their portfolios.
“The digital infrastructure sector is experiencing a double benefit from AI and other new technologies,” says Stéphane Calas, partner at Cube Infrastructure Managers.
“Not only does it create new revenue streams, driving demand for data centre capacity and high-speed connectivity, but our portfolio companies also leverage these technologies internally to optimise costs and boost profitability.”
AI is being used by digital infrastructure owners for two primary applications: increasing revenues and improving efficiency through automation, according to Alex Kesseler, partner at Antin Infrastructure Partners. A number of Antin’s fibre optics businesses are using AI to identify the optimal areas for network buildouts, both in terms of extending into new regions and densifying already covered areas, for example.
“Advanced data science tools are instrumental in this process, merging information from diverse sources such as data on buildings, including types and usages of buildings, types of tenants, exact locations, corporate company information, and various internal databases including customer data, sales data and technical network data,” says Kesseler.
Meanwhile, when it comes to efficiency deployment, AI is being employed to optimise the activities of the sales force. “It predicts which customers are likely to churn, thereby requiring particular attention from the sales team. This allows for prioritised and focused allocation of resources.”
Ardian’s Spanish telco Adamo is also looking into how it could use AI to help predict churn and aid customer retention. A lot of retail operators are currently gathering customer data for this purpose, including AI analysis of speech-to-text transcripts in order to gauge customer sentiment, according to Ardian’s senior data scientist Guillaume Rigaud.
Energy efficiency
Energy efficiency is another powerful use case for AI in digital infrastructure. Energy-hungry data centres are making headlines for all the wrong reasons. However, the AI which is creating that energy demand also has a key role to play in curbing it.
It is not only data centres where AI can help curtail energy usage, either. “We are using data technologies extensively to support our ESG strategy, particularly in terms of improving the energy consumption of our digital infrastructure assets,” says Ardian director and head of digital innovation Pauline Thomson. “Sustainability has obviously become a key theme for data centres, but 5G antennae also consume a lot more energy than their 2G, 3G or 4G predecessors. AI can help drive energy efficiency.”
Ardian’s Icelandic telco Mila is exploring the use of AI to reduce the energy consumption of radio units on towers. “Those radio units consume energy in two different ways – powering the unit itself to monitor the radio environment and then processing data when there is traffic,” says Rigaud.
“We are investigating the ability to turn the cells off when there is no traffic in order to save energy on unused bandwidth. The first step involves shutting down the cells at night using a fixed schedule. Then we can start to get more sophisticated, creating a dynamic schedule for turning cells off when they are not required, using AI.”
In addition to optimising energy consumption, artificial intelligence can strengthen cybersecurity by detecting and preventing real-time threats, according to Cube managing director Camille Mueller. “AI can also automate processes such as generating purchase orders and handling invoices, freeing up human resources. It can predict equipment failures and schedule maintenance, cutting maintenance costs.
“Overall, AI can improve many ar eas by efficiently allocating resources, automating service provisioning and minimising downtime.”
Next-generation infrastructure
Al is poised to play a transformative role in the global economy
In addition to its revenue-driving and cost-optimisation applications in the traditional digital infrastructure verticals of towers, fibre and data centres, AI is also poised to play transformative role in next-generation networks, cloud infrastructure and the Internet of Things. “The additional complexity that results as more and more devices are connected to the internet is not linear but exponential. The ability of humans to continue to manage that network efficiently is all but impossible and so we need AI tools to do the job,” says Digital Alpha partner Karl Meyer. “In many ways, this is the ideal application for AI. Reduced latency and jitter are measurable outputs, and it is possible to continually refine the configuration of the network in response to real-time data feedback on latency, performance and speed. You can then also add time parameters around season or time of day, for example. This all helps drive the efficiency of digital infrastructure.”
Here, Al is both the trigger and the solution creating huge additional demand for connectivity, as well as the ability to manage that growth. “AI requires a huge amount of compute power and storage, typically in cloud applications. Those cloud applications are constantly communicating with end user devices – be it a laptop, Alexa or your car,” Meyer explains. *AI applications also consume enormous amounts of data and need drastically expand network capacity.” He adds that the internet is optimised for download speed rather than upload. “Fibre deployment is helping to improve symmetricity but, in the US, the vast majority of people are still using cable, particularly in the home. ‘Upload speeds will need to increase dramatically in order to provide a good experience for AI applications. Fibre-to-the-home penetration in the US is currently in the high teens and forecast to reach 30 percent in the next 10 years. As AI proliferates, we need to be able to bring everyone along, including those without access to fibre.”
Data centre demands
In addition to the re-architecture of the Internet with the Internet of Things, and the acceleration of fibre roll-out, the advent of AI is revolutionising data centre requirements, including the proliferation of micro-edge data cen tres, supporting low latency.
However, with the acceleration of edge computing, there is a growing awareness that not all data centre requirements are the same. And in yet another example of AI both creating and solving a problem, it is helping navigate the route forward.
“Historically, data centres have been designed to be online continuously and fully resilient,” says Jon Mauck, senior managing director at DigitalBridge.
“But with the volume of compute that is now required to support the proliferation of AI, as well as cloud adoption and the ongoing digitisation of the economy, it is no longer tenable to have every square metre of data centre capacity aspire to that level of reliability, particularly given supply constraints, long lead times and high costs.
“As a result, end customers are becoming more granular in their assessment of the levels of resilience and reliability needed for different applications, and they are using AI to help them make those decisions.”
For example, AI can help customers analyse the impact of a particular data centre going offline, the risks of the data centre running at 100 degrees rather than 88 degrees, or the impact of porting something in a software layer rather than having a second generator, Mauck explains. “In other words, customers can optimise cost efficiency. That reflects a new level of intelligence going into network and compute architecture.”
Until recently, there hasn’t been the economic mandate to drive this change because those procuring data centre capacity haven’t necessarily taken a holistic view of the challenges.
“Now that demand for data centres is doubling year on year and costs are soaring, decision-making is moving up the ranks as far as CFO level in some cases, and those decision-makers are exploring the trade-offs around things like latency, for example, which means that new data centre locations are becoming viable,” Mauck says.
“Then there are applications that don’t need the same level of reliabili ty, so customers are pulling back on redundancy, for instance. These things can materially change cost profiles.”
With data centres currently costing between $10 million and $15 million per MW of infrastructure, the ability to reduce those costs by as much as 70% is clearly compelling. “That ability is being driven by granular analysis, supported by artificial intelligence,” adds Mauck.
It is clear then that AI is both creating a massive investment opportunity across every aspect of the industry, as well as providing the tools to manage such exponential growth. From revenue optimisation to cost efficiency, artificial intelligence is undoubtedly going to be pivotal to the future of one of the hottest sectors in infrastructure.