The new chips designed to solve AI’s energy problem

illuminem summarises for you the essential news of the day. Read the full piece on The Wall Street Journal or enjoy below:

🗞️ Driving the news: Tech firms are racing to design next-generation chips that dramatically reduce the energy consumption of artificial intelligence (AI) systems
• As global demand for AI services skyrockets, projections show energy usage could rise 50% annually through 2030, prompting urgent calls for hardware innovation
• Industry veterans, including Cloudflare’s hardware chief Andrew Wee, are sounding the alarm on the sector’s rapidly growing energy footprint

🔭 The context: AI workloads, particularly large-scale model training and inference, are exceptionally power-intensive—driving surging electricity demand in data centers worldwide
• Traditional CPUs and GPUs, while powerful, are increasingly inefficient for these tasks
• In response, companies are developing custom AI chips and architectures designed to optimize processing and energy use
• From Nvidia and Google to startups and cloud providers, the race is on to balance performance with sustainability

🌍 Why it matters for the planet: AI’s energy intensity risks undermining climate goals if left unchecked
• Data centers already consume over 1% of global electricity, a figure expected to multiply as AI use expands
• Without breakthrough efficiencies, the sector could require vast new energy infrastructure—often fossil-powered — to keep pace
• Sustainable hardware innovation is critical not only to meet climate targets, but to ensure AI’s long-term economic and environmental viability

⏭️ What's next: Leading tech firms are expected to ramp up investments in energy-efficient AI chips, with a focus on specialized architectures, novel materials, and edge computing to reduce server strain
• Policy pressure may also grow, with governments scrutinizing the carbon footprint of AI infrastructure
• By 2030, hardware design could be a key determinant of both AI competitiveness and compliance with corporate net-zero commitments

💬 One quote: “I can’t wrap my head around it … the growth in energy demand for AI is unsustainable.” – Andrew Wee, Head of Hardware, Cloudflare

📈 One stat: AI-related energy use is projected to grow 50% annually through 2030, according to the World Economic Forum

See on illuminem's Data Hub™ the sustainability performance of Nvidia, Google, and their peers Intel, and AWS

Click for more news covering the latest on green tech and power grid