Foxconn, the Taiwanese electronics maker best known for assembling Apple’s iPhones, aims to win half of the global market for AI servers within a couple of years, its chairman said in an interview published by CNBC on October 30. Young Liu, chairman of Hon Hai Technology Group, the company known as Foxconn, told CNBC’s Christine Tan that the company’s share is currently around 40%, could reach about 45% next year and could grow to 50%. The figures are the company’s own estimates; CNBC’s report does not give the size of the market or cite an independent measure of share.

Liu also said Foxconn’s production capacity is already committed for at least the next 18 months, that AI infrastructure could make up 60% to 65% of its revenue within three to five years, and that the company will control its AI spending tightly because overinvestment carries its own risk.

What Liu told CNBC

On the prospect of a half share, Liu said, “I think we can grow up to 50 percent,” according to CNBC. Liu said the share could reach about 45% next year and 50% in a couple of years. He also said revenue from AI infrastructure could reach 60% to 65% of the company’s total in three to five years, a figure that CNBC said excludes edge AI devices without defining the term.

Liu called AI demand strong. CNBC reported that demand has so far come mostly from AI model developers and cloud-service providers, and that governments and enterprises have not yet fully adopted the technology. CNBC carried a related video segment under a headline saying there is no risk of overcapacity yet because a large part of AI demand remains unrealized, attributed to Foxconn.

The company’s share estimate has been moving up in the public record. Reuters reported in March, when Foxconn published its fourth-quarter 2025 results, that the company expected its AI server market share to reach 40%. In the October interview, Liu described 40% as the current level. The sources reviewed do not explain how Foxconn defines the market or which competitors it counts, so the share figures should be read as a company claim rather than an industry statistic.

The move into the United States

Liu told CNBC that Foxconn plans to move further up the supply chain in the United States so it can respond more quickly to demand. Because manufacturing costs are higher there, the company will start with higher-margin products, namely AI server racks and data-center equipment, and expects that greater scale could eventually bring those costs down. The broader goal, CNBC reported, spans servers, racks, complete data centers and computing services.

Foxconn’s U.S. and North American plans were already public. Reuters reported in March that the company was building AI server plants in Mexico and Texas for Nvidia, and TaiwanPlus reported in August that Foxconn wants to remain Nvidia’s biggest AI server maker and plans factories in Mexico and Texas to build Nvidia servers. Reuters also noted that in August 2025 Foxconn agreed to sell a former car factory in Lordstown, Ohio, for $375 million, including machinery it bought in 2022 to make electric vehicles.

What the financial results show

Foxconn’s own reports show how much the company’s business has shifted toward cloud and networking equipment, the category that includes AI servers. According to Focus Taiwan, reported by ANI, Foxconn’s net profit for April through June was NT$59.97 billion, about $1.86 billion, up 35% from a year earlier and the highest second-quarter profit on record. Revenue was NT$2.53 trillion, about $78.6 billion, up 41%. Cloud and networking products made up 51% of revenue, up from 48% in the first quarter, ahead of smart consumer electronics, which includes iPhone assembly, at 29%.

The margins are thin. Foxconn’s gross margin in the quarter was 6.12%, against 6.33% a year earlier, and its operating margin was 3.75%, up from 3.15%, the same report said. For the first half of 2026, net profit was NT$109.89 billion, up 27%, on sales of NT$4.65 trillion, up 35%. Lens analysis: with a gross margin near 6%, the preference Liu described for starting U.S. production with higher-margin products is consistent with the economics shown in the company’s results.

The trend is visible over a longer period. Reuters reported in November 2025 that cloud and networking supplied 42% of revenue in the September 2025 quarter and outpaced smart devices for a second straight quarter, and that Foxconn’s quarterly profit then was NT$57.7 billion on revenue of NT$2.06 trillion. Foxconn has not always met expectations: Reuters reported that its fourth-quarter 2025 net profit of NT$45.21 billion, down 2%, was about 29% below the LSEG consensus, which the company attributed to a higher tax rate. TaiwanPlus reported that Foxconn shares were up 17% for the year as of August 12, trailing a 57% gain in Taiwan’s broader stock index.

How much Foxconn plans to spend

Liu has raised the spending he describes. In November 2025, he told Reuters at Foxconn’s annual Tech Day that the company planned to spend about $2 billion to $3 billion a year on its AI business, and that AI could make up more than half of its roughly $5 billion annual capital spending. In the October interview, CNBC reported that Foxconn plans to spend $3 billion to $5 billion a year on AI, and that it could spend more depending on how far it takes vertical integration, meaning making more of the components itself. The two reports describe different time periods and may define spending differently, so the figures should not be compared line by line.

Liu stressed the risk of overreach. “Overinvestment will kill the company,” he said, according to CNBC, which reported that Foxconn will control its AI spending tightly. He said that if one major customer reduced its AI spending, growth at other customers could offset it, but that a slowdown across the industry would be a much bigger challenge.

The customers’ spending

Reuters has tied demand for AI servers to the data-center build-out by cloud providers such as Amazon, Microsoft and Google, and several of those companies reported in the same week as Liu’s interview. CNBC reported that Alphabet kept its 2026 capital-spending forecast at $195 billion to $205 billion and said spending will rise significantly in 2027, that Microsoft cut its full-year forecast to $170 billion from $175 billion, and that Meta expects $135 billion to $145 billion. The article added that capital spending by the largest cloud companies is expected to exceed $1 trillion next year, citing a Bloomberg-linked estimate from a September 25 report that put it at about $1.2 trillion, up roughly 50%.

The market’s reaction to that spending was mixed. Alphabet’s shares fell about 13% on October 29, their steepest one-day drop on record, after the company reported negative free cash flow of about $9 billion, CNBC reported. Microsoft’s shares were roughly flat, and Microsoft’s chief financial officer said delays from third-party GPU providers would push some capacity into the first half of 2027. Shares of smaller companies that rent out computing capacity, including CoreWeave, Nebius and IREN, fell after the reports, CNBC said.

Memory supply is another constraint. SK hynix said on October 28 that its high-bandwidth memory is committed through 2029, according to Reuters, and Reuters reported in March that memory shortages and price rises were a concern for Foxconn, though a product mix weighted toward pricier models limited the impact. The sources reviewed do not say how memory supply affects Foxconn’s AI server output.

What to watch

Foxconn reported its September-quarter results on November 12 last year, according to Reuters. This year’s date was not in the sources reviewed. That report will show whether the cloud and networking share of revenue has moved beyond 51% and whether the capacity that Liu described as booked is translating into higher margins.

Lens analysis: the interview contains one claim that can be tested and one that cannot. Foxconn’s revenue mix can be checked each quarter against its filings. The 40% share and 50% goal cannot be checked without a market-size figure and a definition of an AI server, and the company has not published either in the reports reviewed. A second thing to watch is whether the 18-month capacity commitment holds if the largest customers slow their spending, which is the scenario Liu himself described as the harder one for Foxconn to absorb.

Sources

Image: Dingpu Plant, Hon Hai Precision Ind. 20161107.jpg by Solomon203, CC BY-SA 4.0, via Wikimedia Commons. Illustrative photo; it does not show the events described.