AI’s insatiable appetite: How data-centre demand is starving consumer electronics of chips

The boom in AI data-center construction is not merely driving demand for graphics processors and high-bandwidth memory; it is fundamentally reorganising the global semiconductor market, squeezing the very supply chains that feed consumer electronics. As data centres pull an outsized share of precious manufacturing capacity toward accelerator processors and HBM memory chips, makers of smartphones, laptops, and other devices are facing rising costs, component shortages, and delayed product launches.

This supply crunch is not a simple case of competing for the same chip. Data-centre servers and consumer devices use fundamentally different silicon. Phones and PCs rely on energy-efficient system-on-a-chip designs with standard DRAM and NAND memory, while AI servers depend on GPUs paired with high-bandwidth memory. Yet the two markets now compete over the same scarce capacity because chipmakers, dominated by a layered oligopoly of just a few firms, are reluctant to bring new fabrication plants online quickly. The memory-chip industry, in particular, has been shaped by boom-and-bust cycles since 2000, making suppliers cautious about expanding output. Instead, they channel investment into higher-margin, high-value products, leaving consumer electronics manufacturers—Apple, Samsung, and a host of Chinese OEMs—scrambling for supply.

For China, the world’s largest producer of consumer electronics, the implications are acute. The nation’s manufacturing ecosystem, which assembles roughly 60% of the world’s smartphones and a vast share of PCs, faces higher input costs and greater vulnerability to export controls on critical minerals and chip components. At the same time, China’s own semiconductor industry is attempting to close the gap in advanced memory and logic-chip production, making the reallocation of global capacity a strategic challenge. The lesson for Chinese firms is that resilience will demand not only closer integration with domestic foundries but also more aggressive investment in on-device AI capabilities that can run on less exotic hardware, potentially softening their dependence on the most constrained chip segments.

Why it matters:
The AI-driven redistribution of chip capacity is creating a structural tension between the needs of hyperscale computing and the consumer electronics industry. For global investors and manufacturing executives, understanding this dynamic is crucial: it shapes both near-term procurement risks for Chinese OEMs and long-term opportunities for domestic foundries that can capture segments of the high-bandwidth memory and advanced logic market.

Source →