Foxconn Wants Half the World's AI Server Market and Is Already Running Out of Capacity
The iPhone assembler's chairman says AI infrastructure will make up nearly two-thirds of revenue within five years. The order book is full for the next 18 months.
Key points
- Foxconn's chairman told CNBC his company aims to reach 50% of the global AI server market within a couple of years, up from roughly 40% today.
- Every production line is fully booked for at least the next 18 months.
- AI infrastructure could account for 60% to 65% of Foxconn's total revenue within three to five years, excluding edge devices.
- Foxconn will begin its U.S. Manufacturing push with high-margin products such as AI server racks, where the labour cost premium is easier to absorb.
- Chairman Young Liu warned that "overinvestment will kill the company" even as he signalled spending could exceed the planned $3 billion to $5 billion annual cap.
Foxconn, the Taiwanese giant formally named Hon Hai Precision Industry and best known for assembling Apple iPhones, is remaking itself around a very different product: the racks of specialised computers that power artificial intelligence.
Speaking to CNBC's Christine Tan, chairman Young Liu said Foxconn currently holds about 40% of the global AI server market, expects that share to climb to roughly 45% next year, and aims to reach 50% within a couple of years after that.
The order book backs him up. Every production line is fully committed for at least the next 18 months, a level of advance booking that would be remarkable in almost any manufacturing sector.
Who is actually buying all these AI servers?
Right now it's AI model developers and big cloud providers, the companies that rent computing power over the internet, doing almost all the buying. Governments and large enterprises have barely started, Liu said, which means a second wave of demand is still waiting to arrive.
That context matters, particularly for companies weighing whether to rent AI capacity or invest in their own infrastructure, a question we examined on 29 September. Foxconn's read is more bullish than the cloud giants' own guidance: Liu described demand as "quite strong" and showed no sign of pulling back.
Foxconn's ambitions stretch beyond servers. Liu wants the company to move up the supply chain to cover full rack systems, entire data centres, and computing services sold directly to customers. AI infrastructure, he said, should account for 60% to 65% of total revenue in three to five years, not counting edge AI devices (hardware that runs AI calculations locally, on-site, rather than in a distant data centre).
What does the U.S. Expansion actually look like?
Foxconn is building more manufacturing capacity inside the United States. Higher labour costs there mean the company will begin with products that carry fatter margins, specifically AI server racks and data centre equipment, where the extra cost per unit is easier to absorb. Liu said greater scale over time could bring those costs down.
For enterprises trying to plan their own AI budgets, the implication is straightforward: if the largest contract manufacturer on earth is locking in capacity 18 months out, lead times for custom AI hardware are not getting shorter soon.
Should anyone worry about a spending crash?
Liu is watching for it. "Overinvestment will kill the company," he told Christine Tan, a notable line from the chairman of a business targeting half a global market.
His hedge is geographic diversification across its customer base. If one customer cuts back, others can take up the slack. An industry-wide pullback is a different problem, and Liu acknowledged it would hit Foxconn and every supplier in the chain.
Foxconn's planned spending sits at $3 billion to $5 billion a year, but Liu said the company could go higher depending on how far it pushes vertical integration, the strategy of owning more steps of the production process itself. That tension sits at the heart of this story: the order book screams expansion while the chairman is simultaneously pumping the brakes.
