AI Is Putting More Pressure on Chip Supply
Artificial intelligence needs enormous computing power.
But the AI boom isn’t only increasing demand for the GPUs usually associated with companies such as Nvidia.
It is also driving demand for many of the other chips needed inside servers, networking equipment, storage systems and supporting infrastructure.
That demand is now becoming strong enough for China’s largest contract chip manufacturer, Semiconductor Manufacturing International Corporation (SMIC), to increase prices for some of its most sought-after production capacity.
SMIC co-CEO Zhao Haijun said the company had negotiated higher prices with customers and expects to charge more for wafers processed during the third quarter.
More Than US$3 Billion in Quarterly Revenue
The numbers show how quickly demand is growing.
SMIC recorded quarterly revenue of more than US$3 billion for the first time, while profit attributable to shareholders reached US$479.2 million, more than three times the level recorded a year earlier.
Production is also running at a high level.
The company’s utilisation rate reached 93.7%, while monthly production capacity increased to approximately 1.1 million 8-inch-equivalent wafers.
When a chip factory is operating this close to full utilisation, there is less spare capacity available when new orders arrive.
That helps explain why pricing power is shifting towards manufacturers.
It Isn’t Only About CPUs and GPUs
One interesting detail is where the demand is coming from.
SMIC said much of the increase was driven by AI-related demand for chips other than CPUs and GPUs, particularly from customers in China.
An AI server needs far more than its main processors.
It also requires networking chips, power-management components, controllers, storage-related semiconductors and many other supporting devices.
As more AI servers and data centres are built, demand therefore spreads across a much larger part of the semiconductor supply chain.
China’s Domestic Chip Industry Is Growing
China accounted for around 90% of SMIC’s second-quarter revenue, while the United States contributed about 8%.
SMIC is also expanding.
The company added approximately 8,000 wafers of monthly 12-inch production capacity during the quarter and plans to accelerate the ramp-up of new production lines as it works to relieve supply constraints.
For the third quarter, SMIC expects revenue to increase another 2% to 4% sequentially, with wafer shipments continuing to rise.
Why This Matters
The AI infrastructure boom is beginning to affect more than just the companies making high-end AI accelerators.
It is influencing the entire technology supply chain.
More AI computing means more servers.
More servers mean more networking, storage and supporting semiconductor components.
And when factories start approaching their production limits, higher demand can eventually translate into tighter supply and higher prices.
SMIC’s latest results are another sign that the AI infrastructure race is spreading much further through the semiconductor industry than GPUs alone.
Closing Thoughts
Artificial intelligence is often discussed as a software revolution.
But underneath every AI application is a massive hardware ecosystem.
The fact that one of Asia’s largest semiconductor manufacturers is now increasing wafer prices while operating at more than 93% utilisation shows just how much pressure the AI build-out is placing on physical infrastructure.
As companies continue building larger AI platforms and data centres, the next bottleneck may not simply be getting enough GPUs.
It may be getting enough of everything else that has to work alongside them.