SMIC says the AI chip shortage runs through 2027, and its own costs show why

Zhao Haijun, co-CEO of Semiconductor Manufacturing International Corp, China's biggest contract chipmaker, said on Friday that the AI industry is still in its early stages and demand remains strong, when asked whether AI stocks were turning into a bubble 3

3

1

1

. At the same briefing, he forecast that the supply chain for AI-related chips will stay tight through 2027, a window that sets expectations for anyone building AI computing capacity 3

3

2

2

.

The demand he pointed to comes from data centers and private clouds, which have been soaking up chips as technology companies keep investing in AI computing infrastructure 2

2

. Many customers that need computing resources are still facing shortages now, he said 3

3

2

2

.

The cost of keeping up

By his account, keeping up with that demand costs the company money. The company expects full-year revenue to top $1.2 billion as its factories run at higher utilization, meaning a larger share of production capacity is in use 3

3

2

2

.

The bill for building capacity is coming due as well, over that same 2027 window. Zhao said depreciation, the accounting charge for equipment and buildings as they wear out, will rise about 30% year-on-year in 2026 3

3

2

2

.

Expanding is slow, too. Some equipment orders are taking over 24 months to arrive, which the company says will affect its capital spending this year 3

3

2

2

. Lead times that long can make it harder for chipmakers to add capacity quickly enough to meet demand 2

2

.

None of this comes with an end date for AI chips themselves. Zhao did offer one for a related shortage, saying that memory chip supply and demand should come back into balance around the middle to end of 2028 3

3

2

2

.