New US coalition targets grid bottlenecks amid rising data centre demand
Google, Tesla, and Carrier have launched a new US lobbying coalition called Utilize, focused on improving how existing grid infrastructure is used. The group argues that much of the US power grid is built to handle short periods of peak demand, meaning significant capacity often sits underused for much of the time. Utilize is therefore pushing for policies that help unlock this hidden capacity more quickly and affordably, rather than relying only on costly and time-consuming new grid buildouts. The coalition is expected to advocate for solutions such as battery storage, demand response, virtual power plants, and grid-enhancing technologies to help meet rising electricity demand from data centres, manufacturing, and wider electrification. It is also backing regulatory reform, including legislation in Virginia that would require utilities to measure how much of the grid is actually being used and factor that into future planning decisions. More broadly, the initiative reflects a growing push in the US to meet new power demand faster, while avoiding unnecessary cost increases for existing customers. A key takeaway for the data centre sector is that power availability and time-to-power have now become the market’s defining constraint. As demand accelerates, the industry is no longer focused only on securing land or capital, but on how quickly usable power can be delivered – making grid access, utilisation, and speed of connection the number one trend shaping data centre growth…Read more
Meta unveils next-generation AI chips to accelerate inference at scale
Meta has announced the next four generations of its in-house Meta Training and Inference Accelerator (MTIA) chips, signalling a major step forward in its strategy to build more of its own AI infrastructure. The new chips – MTIA 300, 400, 450, and 500 – have either already entered production or are due to be deployed over the next 18 months and are primarily designed to support generative AI inference workloads. Across each generation, Meta is increasing compute performance, memory bandwidth, and efficiency, while also aiming to release a new chip roughly every six months. The announcement is significant because it shows Meta moving more aggressively toward custom silicon tailored to its own AI workloads, rather than relying solely on third-party chip suppliers. Later generations of the MTIA platform are designed to handle the rapid growth in generative AI demand, with higher performance, larger memory capacity, and architecture optimised for inference bottlenecks. Meta also noted that the MTIA 400, 450, and 500 can all use the same chassis, rack, and network infrastructure, allowing future chip generations to be integrated into existing data centre environments more easily…Read more
UK to launch £500 million Sovereign AI fund
The UK government is preparing to launch a new £500Mn Sovereign AI fund aimed at supporting the country’s AI development ambitions. The fund, due to launch on 16 April, was first proposed in the government’s 2025 AI Opportunities Action Plan and will be chaired by investor James Wise, a partner at venture capital firm Balderton. Wise has said his role will include supporting strategy around critical infrastructure for AI, suggesting the initiative could have implications beyond software alone. While full details have not yet been disclosed, it remains unclear whether the fund will focus only on AI applications and startups or whether it could also support enabling infrastructure such as semiconductors, compute, or data centre projects. The announcement sits alongside the UK government’s broader push to strengthen domestic AI capability, including its earlier commitment to invest £1Bn to expand the country’s sovereign compute capacity and scale national AI research resources. Taken together, the fund signals that the UK is continuing to take a more active role in shaping its domestic AI ecosystem, with growing attention not just on innovation, but also on the infrastructure and investment frameworks needed to support long-term AI competitiveness…Read more