Nvidia's AI Boom Traps East Asia in Fossil Fuel Dependency

The global AI frenzy, spearheaded by Nvidia's generative AI breakthroughs, is imposing a hidden environmental and energy security cost on Taiwan and South Korea, the primary manufacturing hubs for the company's advanced semiconductors. Nvidia outsources most of its manufacturing to TSMC in Taiwan and relies on memory chip producers like Samsung and SK Hynix in South Korea. As demand for Nvidia's AI GPUs surges, these nations are increasingly turning to fossil fuels to meet escalating energy needs, rather than prioritizing renewable energy sources. In Taiwan, the construction of new TSMC fabrication plants, such as Fab 25 near Taichung, requires energy and water on a scale comparable to entire cities. Fab 25 alone will need at least 1 GW of power and approximately 100,000 tonnes of water daily. TSMC operates over two dozen fabs and plans ten more advanced facilities like Fab 25. Consequently, TSMC alone is projected to consume 24% of Taiwan's electricity by 2030, a significant increase from over 12% last year. To meet this surge, Taiwan's state-owned utility, Taipower, is planning the construction of 18 gas plants totaling about 18.2 GW, which would represent roughly 40% of Taiwan's current generating capacity. This move further entrenches the island's reliance on fossil fuels, which already account for over 80% of its electricity generation. South Korea faces a similar challenge. Samsung and SK Hynix, key partners in the government's "Yongin Mega-cluster" plan, are pursuing massive expansions. This plan envisions a vast complex of semiconductor fabs and AI factories that will become the world's largest industrial complex. Shockingly, the plan calls for power from six massive LNG units (3 GW) by 2030 rather than renewables. Yongin will ultimately demand about 15 GW of electricity, roughly one-sixth of South Korea's total electricity demand, and will consume about half the water used by Seoul and its 10 million people. Environmental groups like Greenpeace East Asia have criticized Nvidia for what they term "silent AI colonialism," alleging that the company outsources the dirtiest parts of its supply chain to East Asia while actively suppressing transparency. Activists urge Nvidia to disclose its full lifecycle carbon emissions and legally commit, alongside its Asian manufacturing partners, to producing 100% carbon-free AI chips by 2030. However, Nvidia has opposed shareholder proposals for such reporting, labeling them "too prescriptive." This reliance on fossil fuels risks trapping Taiwan and South Korea in a "carbon trap," leading to worsening pollution and runaway carbon emissions. The advancement of AI technology, while promising, appears to be built on a foundation that neglects environmental considerations, posing long-term challenges for regional communities. Jensen Huang's quest for AI supremacy may be built on a "dirty empire" powered by East Asian coal and gas.