Foxconn Chairman: Two of Four AI Customer Segments Still Have Immature Demand
Taylor Wilson
Foxconn chairman Liu Young-way said only model developers and cloud providers have built mature AI compute demand so far — government and enterprise buyers are still in a validation phase, a gap he believes will sustain strong compute-market growth for three to five years.
What are the four customer groups?
Liu divides AI compute buyers into four types: model developers (OpenAI, Anthropic, Google), cloud service providers (AWS, Google Cloud), governments, and enterprise clients.
The first two drove initial demand. Model developers created the need; cloud providers scaled up after confirming that AI infrastructure directly lifts revenue and margins.
This means → today's AI compute spending is essentially carried by just two groups.
Which two are still missing?
Government adoption is clearest in defense — drones and battlefield decision systems — while broader public-sector use remains early-stage.
Enterprise clients, Foxconn included, view AI as strategically necessary, but actual deployment models are still evolving. Large-scale procurement has not materialized.
In plain terms = two of the four buyers are still in a "trial period" — bulk ordering has not started.
What does this mean for the compute market?
Liu's core thesis: because government and enterprise demand has not yet been fully released, the overall AI compute market still has significant room to expand.
His time horizon is three to five years of sustained, strong demand growth.
This reflects a framing focused not on how large current demand is, but on how much demand has yet to arrive — the incremental gap is the point.
Has Foxconn's own investment approach changed?
Liu said Foxconn has shifted from financial investing to strategic investing, targeting opportunities that create synergies with the group's manufacturing, supply chain, and customer base.
The company reviews more than 300 investment opportunities per year, sourced from PE and VC partners, financial advisors, supply-chain relationships, and internal engineering teams.
He evaluates on five criteria: industry, product, timing, team, and customer — stressing that timing and execution often matter more than the technology itself.
What happens to underperformers?
Foxconn has tightened capital discipline: if a portfolio company fails to hit preset milestones within three years, the default is to exit — not to hold on because of sunk costs.
Put simply = you get three years to prove yourself; miss the mark and the money stops.
Liu also disclosed that an IC design-services company, leveraging Foxconn's resources and client relationships, could list on Taiwan's Innovation Board as early as 2026.
Content is for reference only, not financial advice.