BT’s Carbon Network Dashboard now measures AI energy consumption

BT’s Carbon Network Dashboard gives businesses a real-time look at how much electricity their network and data-centre infrastructure is using, and with this update it can tie this consumption to traffic patterns caused by individual applications, such as AI systems.

AI can cause ‘sharp, unpredictable increases’ in bandwidth demand, which in turn can spike power use by infrastructure traditionally designed for predictable workloads, says BT. If these spikes in demand overload individual network devices or servers, this impacts performance and causes them to overheat which wastes electricity, we’re told.

With info as to what’s going on in this regard, businesses can change network design, capacity and management, optimise applications and AI workloads, or develop distributed architectures, which can bring components of AI closer to users, devices and machines, says BT. 

The platform now also incorporates electricity consumption data from a wider range of kit, such as SD-WAN equipment, servers and WAN and LAN devices. Its energy optimisation tools have also been expanded and now include V-App IoT builder integration for energy management of wireless access points, zero-touch automation to enable/disable power over ethernet (PoE) ports, and ‘sustainable device refresh recommendations for end-of-life devices.’

“AI has incredible potential but if not deployed thoughtfully could place unpredictable demands on customers’ digital infrastructure causing surges in electricity use and carbon emissions,” said Sarwar Khan, sustainability director, Business, BT (pictured above). “BT is committed to helping customers innovate to achieve sustainable growth. With our Carbon Network Dashboard, we can help them adopt AI at scale while optimising their infrastructure to achieve their decarbonisation goals. It’s a great example of how BT has their back.”

In the future BT says the Dashboard will be able to identify traffic going to colocation or public cloud services and prioritise what to focus on for optimisation, which we’re told will ensure network devices are ‘running within their design limits efficiently and with flexibility to accommodate growth of AI workloads.’

AI adoption certainly seems to be steaming along at a rate of knots in all sorts of sectors – it now seems almost a necessity that any corporate utterance or product announcement in the tech world includes some mention of new AI functions coming online.

A study earlier in the month claimed telcos are leading the way in generative AI adoption compared to other industries, with 70% of them already using it compared  with an average of 54% in other sectors having fully or partially implemented the technology.

Whether motivated by sustainability concerns or damage to the bottom-line that spiking electricity bills may bring, the energy drain of all these powerful new AI systems could become an increasingly hot topic, as they are widely and eagerly stuffed into any business operation that will take them.