Foxconn AI Servers Surpass iPhones: The $3M Rack That Changed Everything

Foxconn AI servers have officially surpassed iPhones as the company’s biggest revenue driver — a historic shift no one predicted this fast. For over a decade, Foxconn was synonymous with one product: the iPhone. The Taiwanese manufacturing giant built its empire assembling hundreds of millions of Apple devices every year. But in 2026, something historic happened — AI servers officially surpassed iPhones as Foxconn’s biggest revenue driver, and it’s not even close anymore.

Foxconn’s Q1 2026 revenue hit $66.6 billion, up 29.7% year-over-year, with AI servers leading the charge. This isn’t a temporary blip. It’s a fundamental restructuring of how the world’s largest electronics manufacturer makes money, and it tells you everything you need to know about where the tech industry is heading.

Foxconn AI Servers: The Numbers That Tell the Story

Let’s start with the revenue breakdown that made analysts sit up straight. Foxconn’s cloud and networking products division — which includes AI servers — now accounts for approximately 40% of total revenue. Consumer electronics, which includes iPhone assembly, has dropped to 38%. That crossover happened in Q2 2025 and the gap has been widening ever since.

In March 2026 alone, Foxconn posted record monthly revenue of NT$803.7 billion (roughly $25 billion), a staggering 45.6% increase year-over-year. Chairman Young Liu confirmed that AI server shipments are on track to double in 2026 compared to 2025, with Foxconn maintaining roughly 40% market share in AI server rack assembly globally.

For the full year, Foxconn’s revenue is projected to exceed $310 billion in 2026, up more than 20% from 2025’s record $258.3 billion. The company was named to TIME’s 100 Most Influential Companies list for 2026, largely because of this AI pivot.

How Foxconn AI Servers Overtook iPhones

The rise of Foxconn AI servers didn’t happen overnight, but it accelerated faster than anyone predicted. Three years ago, Foxconn’s cloud and networking division was a secondary business. The company was known as “the iPhone maker” and its stock price moved based on Apple supply chain rumors.

What changed was the explosive demand for AI training and inference compute. Every major tech company — Microsoft, Meta, Google, Amazon, Oracle — started ordering massive quantities of GPU servers. And someone had to build them. Foxconn, with its unmatched manufacturing scale and precision assembly capabilities, was the natural choice.

The company positioned itself as NVIDIA’s primary manufacturing partner for its most advanced AI server products. When NVIDIA needed someone to assemble the complex, liquid-cooled GB200 NVL72 racks, they turned to Foxconn. That single product line has become perhaps the most lucrative hardware contract in Foxconn’s history.

Foxconn AI Servers and the NVIDIA GB200 Factor

Understanding why Foxconn’s AI server revenue exploded requires understanding the NVIDIA GB200 NVL72. This isn’t your standard server — it’s a $3 million liquid-cooled AI rack containing 36 Grace CPUs and 72 Blackwell GPUs in a single unit. Each rack requires precision liquid cooling integration, high-speed NVLink interconnects operating at 1.8 terabytes per second, and power delivery systems handling up to 120 kilowatts.

Foxconn is the sole assembler of these Blackwell-powered GB200 servers, which carry roughly a 40% price premium over conventional server models. When your product costs $3 million per unit and you’re the only company building it, the revenue math gets very favorable very fast.

Compare that to an iPhone, which generates maybe $20-30 in assembly fees per unit. You’d need to assemble 100,000 iPhones to match the revenue from a single GB200 rack order. The economics of AI server manufacturing are simply in a different league than consumer electronics assembly.

Building AI Factories Worldwide

Foxconn isn’t just riding the current wave — they’re building the infrastructure to dominate AI manufacturing for the next decade. The company is constructing what’s been described as the world’s largest NVIDIA GB200 facility in Guadalajara, Mexico, with capacity for approximately 20,000 AI servers annually.

In Ohio, Foxconn operates a plant producing AI servers for SoftBank’s Stargate project, targeting nearly 2,000 racks per week in 2026. That’s a production rate that dwarfs anything the AI infrastructure industry has seen before.

The geographic diversification is strategic. By building facilities in Mexico and the United States, Foxconn reduces its dependence on Taiwan-based manufacturing — a smart hedge given ongoing geopolitical tensions in the region. It also positions the company to serve North American hyperscale customers with shorter supply chains and potentially lower tariff exposure.

What This Means for Apple

Here’s the uncomfortable question that Apple investors are starting to ask: if Foxconn’s most profitable and fastest-growing business is now AI servers, how does that affect their iPhone manufacturing relationship?

The short answer is that it shouldn’t — for now. Foxconn still assembles the majority of iPhones and the volume is enormous. Apple’s Q2 2026 earnings showed record revenue, and iPhone demand remains strong. Foxconn isn’t abandoning Apple; it’s diversifying beyond Apple.

But the long-term dynamics are worth watching. If AI server margins continue to outpace consumer electronics margins, Foxconn will have every incentive to allocate its best engineering talent, factory capacity, and capital investment toward AI infrastructure. Apple has already been diversifying its own supply chain, bringing in Tata Electronics and other manufacturers in India. The Apple-Foxconn relationship remains important for both companies, but it’s no longer existential for either one.

The Margin Problem Nobody Is Talking About

There’s a catch in Foxconn’s AI transformation that deserves more attention. While revenue is surging, profit margins haven’t kept pace. Foxconn’s Q1 2026 net profit margins were actually lower than the same period last year, despite the massive revenue increase.

The reason is that AI server assembly, while high-revenue, is still fundamentally a contract manufacturing business. Foxconn doesn’t design the chips (that’s NVIDIA), doesn’t supply the memory (that’s Samsung and SK Hynix), and doesn’t make the cooling systems. They assemble and integrate, which means margins are thin relative to the sticker price of the final product.

This is the same challenge Foxconn faced with iPhones — massive revenue, modest margins. The company is essentially swapping one high-revenue, low-margin product line for another. According to IDC research, global AI server revenue will exceed $150 billion by 2027. Whether Foxconn can capture more value in the AI server supply chain through vertical integration or higher-margin services will determine if this pivot truly transforms their profitability, not just their top line.

Foxconn as a Bellwether for the AI Boom

Foxconn’s transformation matters beyond the company itself because it’s a leading indicator for the entire AI infrastructure buildout. When the world’s largest electronics manufacturer fundamentally shifts its revenue mix toward AI in under two years, it tells you the demand is real, massive, and sustained.

The company’s order book is a proxy for how much the hyperscalers are actually spending on AI compute. And right now, that number is going up, not down. Companies like IREN, Microsoft, Meta, and Oracle are all racing to deploy GPU infrastructure at unprecedented scale, and every one of those deployments flows through Foxconn’s assembly lines.

For investors tracking the AI trade, Foxconn’s quarterly results are arguably more informative than the hyperscaler earnings themselves. Foxconn sees the actual hardware orders months before they show up in cloud revenue figures. When Foxconn’s AI server revenue growth starts decelerating, that’s when you should worry about the AI infrastructure cycle peaking.

Foxconn AI Servers: The New Reality

The dominance of Foxconn AI servers in the company’s revenue mix is likely to accelerate further in the coming quarters. With the Guadalajara facility ramping up production and the Ohio plant at full capacity for Stargate, Foxconn AI servers could account for 50% or more of total revenue by Q4 2026.

The rise of Foxconn AI servers reflects the broader industry transformation. For more context, explore our coverage of IREN acquiring Mirantis for AI cloud, Panthalassa floating AI data centers, and the Cerebras IPO. Also check out our tutorial on building AI agents.

Foxconn building more AI servers than iPhones is one of those milestones that sounds surprising until you think about it for 30 seconds. The AI revolution needs physical infrastructure, and that infrastructure needs to be manufactured by someone with the scale, precision, and supply chain relationships to deliver millions of components assembled into the most complex computing systems ever built.

Foxconn has spent 50 years perfecting high-volume precision manufacturing. They just found a customer willing to pay $3 million per unit instead of $999. The revenue math was always going to tip in AI’s favor — the only question was when. In 2026, we got our answer.

As Bloomberg analysts noted, the shift in Foxconn’s revenue mix happened two years faster than anyone predicted. The iPhone didn’t die. It just got overshadowed by something even bigger.