ByteDance's AI Chip Localization Hits Substrate Shortage Bottleneck
nashnova research
ByteDance is reportedly planning to spend roughly RMB 100 billion (~$14.9 billion) per year on chip packaging substrates, but a tight supply of advanced substrates is emerging as a key bottleneck for China's AI compute expansion.
How much is ByteDance spending — and on what?
DIGITIMES reports that ByteDance is planning ~RMB 100 billion (~$14.9 billion) per year in chip substrate procurement.
The target: advanced packaging substrates — the circuit-board base layer that chips are mounted onto before they can function.
This means → ByteDance's AI buildout has moved past "buying chips" into securing the upstream materials needed to assemble them — a spend on par with top-tier semiconductor capex.
Where exactly is the bottleneck?
Surging demand from Chinese AI models is pushing supply-chain pressure upstream from chips to the packaging layer.
Advanced packaging substrate capacity is now the critical chokepoint constraining China's AI compute expansion.
In plain terms = the constraint is not chip design capability — it is that there are not enough of the "base plates" needed to assemble chips into working units.
How much can we actually see?
DIGITIMES flagged the trend, but specific supplier names and capacity details sit behind a paywall — publicly available information is limited.
This means → the true severity of the supply squeeze and ByteDance's specific supplier lineup remain visible only in outline, not in detail.
The key variable to watch: whether substrate capacity expansion can keep pace with the procurement demands of China's AI chip localization push.
市场有风险，内容仅供研究参考，不构成投资建议。