Foxconn's AI Servers Overtake the iPhone: Q2 Earnings Mark a Historic Inflection Point

For more than two decades, Foxconn was synonymous with one product: the iPhone. That era quietly ended on August 12, 2026, when the Taiwanese manufacturing giant reported quarterly earnings in which its cloud and networking segment — driven overwhelmingly by AI servers — crossed the 50% revenue threshold for the first time in the company’s history.

Foxconn’s second-quarter net profit rose 35% year-on-year to NT$59.97 billion ($1.86 billion), edging past the LSEG analyst consensus estimate of NT$58.8 billion. Revenue for the April-to-June period surged roughly 40% year-on-year, a figure the company had already telegraphed in its preliminary July release. But the headline numbers tell only part of the story. The real milestone is structural: a company that built its empire assembling consumer smartphones is now, by revenue mix, primarily an AI infrastructure manufacturer.

The 51% Milestone

In its earnings release, Foxconn disclosed that cloud and networking products — a category encompassing AI servers, networking switches, and data center rack systems — accounted for 51% of second-quarter revenue. Smart consumer electronics, the segment that includes the iPhone, fell to 29%. This is the first time cloud and networking has exceeded the 50% mark.

The shift has been building for years but accelerated dramatically in the last twelve months. In the same quarter a year earlier, cloud and networking products represented approximately 41% of revenue. That segment has now grown by ten percentage points in a single year, cannibalizing share from the consumer electronics business that once defined Foxconn’s identity. Computers and terminals, along with components and other products, made up the remaining roughly 20%.

Foxconn is Nvidia’s largest server manufacturing partner, assembling the GB200 and GB300 AI server platforms that power the world’s biggest cloud providers. The company also manufactures networking infrastructure for hyperscale data centers and is deeply embedded in the supply chains of Amazon, Microsoft, Google, and Meta.

Vera Rubin: The Next Wave

Perhaps the most consequential forward-looking statement came from Foxconn’s rotating CEO Michael Chiang, who addressed the company’s pipeline for Nvidia’s next-generation Vera Rubin server products. Chiang confirmed that AI server racks built on the Vera Rubin platform will enter mass-production preparation in the third quarter of 2026, with initial shipments expected to begin in the fourth quarter.

“We expect production volumes to increase gradually over the next several quarters, and it will become our major product next year,” Chiang said during the earnings call.

This timeline is significant because it signals that Foxconn’s AI server revenue growth — already projected to expand dramatically — has a multi-year runway. Vera Rubin succeeds the current GB300 generation and represents Nvidia’s push toward even higher-performance AI compute density. Foxconn’s explicit commitment to making Vera Rubin its flagship product in 2027 underscores how deeply the manufacturer’s roadmap is now intertwined with Nvidia’s product cycles.

The CoWoS Bottleneck

Chiang struck a note of caution on one critical constraint: CoWoS (Chip on Wafer on Substrate) packaging capacity. CoWoS is TSMC’s advanced packaging technology, essential for producing the complex multi-chip AI accelerators that Nvidia designs. It has been the single most significant supply chain bottleneck in the AI hardware industry.

“The market currently expects CoWoS capacity to grow by more than 50% next year,” Chiang said, “but how much of that can ultimately be translated into shipments of next-generation AI server racks will depend on chip supply.”

The implication is that while demand from Foxconn’s customers is effectively unlimited — Chiang described it as “very strong” — the company’s ability to convert that demand into shipped revenue depends on TSMC’s packaging throughput. CoWoS capacity has roughly doubled in 2026, but the >50% expansion projected for 2027 must materialize for the AI server market to meet its aggressive growth targets.

Capital Expenditure and Global Expansion

Foxconn’s capital expenditure is also scaling aggressively. The company forecast a 30% increase in capex for 2026 compared to the previous year, reflecting investments in new manufacturing capacity for AI servers.

This spending is geographically diversified. Foxconn is currently building new factories in Mexico and Texas dedicated to producing AI servers for Nvidia, part of a broader industry trend to move advanced manufacturing closer to U.S. customers amid geopolitical tensions and tariff risks. The company already produces the majority of iPhones destined for the U.S. market in India, and its EV business continues to expand as a longer-term diversification bet.

The Mexico and Texas facilities are particularly strategic. They insulate Foxconn and Nvidia from potential trade disruptions between the U.S. and China, while positioning manufacturing within North American supply chains that are increasingly favored under U.S. industrial policy.

Market Reaction and Context

Foxconn shares closed 2.7% higher on Wednesday ahead of the earnings release. However, the stock has risen only 17% year-to-date, notably underperforming the broader Taiwan Weighted Index’s 57% gain. This suggests the market has been slower to re-rate Foxconn as an AI infrastructure play than as a consumer electronics assembler.

That discount may narrow. The company reiterated its full-year guidance for “strong” revenue growth, driven by AI demand throughout the remainder of 2026. With cloud and networking now the majority of the business, Foxconn’s earnings power is increasingly correlated with AI capital expenditure cycles rather than smartphone upgrade cycles.

Chiang’s outlook for the third quarter was unequivocal: “AI-related business performance will continue to grow. Combined with ICT products entering their peak season in the second half of the year, we expect significant quarter-on-quarter growth and strong year-on-year growth.”

What This Means for the Industry

Foxconn’s earnings are a real-time barometer of AI infrastructure spending. When the world’s largest contract manufacturer reports that AI servers are now its dominant business — surpassing the iPhone, the most successful consumer electronics product in history — it is a structural signal, not a quarterly blip.

The implications extend across the supply chain:

• For Nvidia: Its primary manufacturing partner is committing capital and capacity to Vera Rubin on a timeline that confirms Nvidia’s product roadmap is on track.
• For TSMC: The CoWoS bottleneck remains the gating factor. Foxconn’s public framing puts additional visibility on TSMC’s packaging expansion as the industry’s pace-setter.
• For cloud providers: The supply of AI servers is scaling, but constrained by advanced packaging capacity. Cloud capex will remain elevated as companies compete for limited server allocations.
• For the broader hardware ecosystem: Foxconn’s pivot validates the thesis that AI infrastructure has become the dominant driver of global electronics manufacturing, not a niche vertical.

The company that once defined itself by the phone in your pocket is now defined by the data center down the road. For the AI industry, that is both confirmation of the boom and a reminder of where the physical bottlenecks still lie.