AI Breakthroughs Could Slash Energy Demands, Temasek Says at Bloomberg Summit
Source: The Edge Singapore
Temasek's emerging technologies head says advances in AI architecture, chip design, and materials science could dramatically alter the sector's energy equation, even as the firm acknowledges its own carbon targets face headwinds from AI's rising power consumption.
Every conversation about AI scale eventually hits the same wall: energy. Training a single large model can consume as much electricity as a small town, and data centres are projected to account for 20% of US electricity consumption by 2035. Temasek, Singapore's state-owned investment firm with a portfolio that spans AI infrastructure, chip design, and data centre assets, has a direct stake in whether that trajectory bends.
Russell Tham, head of emerging technologies at Temasek Global Investments, told the Bloomberg Sustainable Business Summit in Singapore on July 22 that advances in AI architecture, materials discovery, and chipmaking could dramatically alter the sector's energy equation. "You may see the energy equation for generating AI tokens change quite drastically," Tham said, describing the current generation of AI as "fundamentally an ineffective AI architecture." Temasek has invested in novel semiconductor materials, chip architectures, and AI model designs that it believes are vastly more energy efficient — though Tham acknowledged that some of these remain unproven at scale.
The timing matters because Temasek's energy problem is also Singapore's energy problem. The city-state has committed to halving its portfolio carbon emissions from 2010 levels by 2030, but admitted earlier this year that AI's rising energy demand makes that target unlikely. Singapore is also one of the world's most data-centre-dense economies, and grid constraints and ballooning electricity prices are pushing operators to explore everything from geothermal to small modular nuclear reactors. If Temasek's thesis on more efficient AI architecture pans out, it doesn't just improve the firm's carbon math — it changes what's possible for Singapore's entire tech economy.
One of Temasek's portfolio companies, UK-based CuspAI (backed jointly by Temasek and Bezos Expeditions), is working on improving semiconductor production by reducing or eliminating the use of rare metals. The broader implication is that the next wave of AI efficiency gains may come not from better algorithms but from better physics —新材料, novel transistor designs, and chip architectures that do more computation per watt. These are bets with long time horizons, but the payoff is structural rather than incremental.
Why it matters for Singapore: Temasek is effectively placing a hedge on behalf of Singapore's economy. If AI energy demand continues its current trajectory, the city-state faces hard trade-offs between data centre growth and carbon targets. If Temasek's investments in more efficient AI architecture succeed, they unlock a path where Singapore can scale its digital economy without breaking its climate commitments. That makes this more than a portfolio play — it's a strategic hedge on the country's energy future.
