Foxconn reported a Q3 2026 net profit of about NT$77.5 billion on November 12, up 34% from a year earlier, on revenue that grew 47%. AI server racks are the engine. The headline you may have seen, that Foxconn "lifted its yearly outlook", is the part the sources do not back up: the company repeated that it expects strong growth, and it still gives no numeric guidance.
Update (November 12, 2026): Q4 rack shipments and Vera Rubin assembly
On the same earnings cycle, chairman Young Liu told reporters AI rack shipments should grow by a high double-digit percentage in 2027, per Reuters coverage of the November 12 call. Trade press on the same results also cited high double-digit quarter-on-quarter rack shipment growth in Q4 2026, per coverage of the earnings materials. That is sequential shipment language, not a new numeric annual outlook raise.
On the call Liu said cloud and networking products already exceed 50% of revenue and may pass 60% within three to five years, and that chips, memory and key components remain tight with demand likely still outpacing supply next year, per the same Reuters report. Separately, at Hon Hai Tech Day on October 30, Foxconn showcased cableless Vera Rubin compute-tray designs that trade reporting says cut assembly from more than two hours to about 20 minutes.
CapEx through the first nine months of 2026 reached about NT$161.7 billion, per trade summaries of Foxconn's Q3 presentation slides. We still could not open Foxconn's own Q3 release PDF in this run, so treat shipment percentages as attributed to Liu on the call or in slides, relayed by wires, until a primary transcript posts.
This post separates the numbers that are confirmed from the framing that is not. It also explains why the profit rose slower than revenue, and what the result says about AI capacity heading into 2027. It builds on our earlier coverage of the AI chip supply chain in October and the TSMC Q3 record profit.
TL;DR: Foxconn Q3 2026 at a glance
| Question | Answer | Source status |
|---|---|---|
| Q3 net profit | About NT$77.5B, up 34% year on year | Reuters-based report; Foxconn Q3 release not opened |
| Analyst expectation | NT$69.6B (LSEG consensus), so a beat of about 11% | Reuters via LSEG |
| Q3 revenue | About NT$3.03 trillion (about $95.4B), up 47% | Foxconn monthly sales, reported by Reuters on Oct 5 |
| Biggest month | September: NT$1.16 trillion, up 38%, first month above NT$1 trillion | Reuters, Oct 5 |
| "Raised yearly outlook" | Not supported: outlook described as unchanged and strong | Foxconn Aug 12 release |
| AI rack shipments, 2026 | Expected to more than double | Foxconn Aug 12 release |
| AI production capacity demand, 2027 | Expected to "remain very strong" (no number) | Foxconn Aug 12 release |
| Numeric guidance | None; Foxconn does not publish it | Reuters, Oct 5 |
What did Foxconn actually report?
The profit number comes from Reuters coverage of the November 12 results. Net profit attributable to Foxconn's owners was about NT$77.5 billion, or roughly $2.35 billion. A year earlier it was NT$57.7 billion, which gives the 34% rise. The LSEG consensus Reuters cited was NT$69.6 billion, so the quarter beat by about NT$7.9 billion.
Revenue was known earlier. On October 5 Foxconn published monthly sales that put July to September revenue at about NT$3.03 trillion, up 47% year on year. That beat an LSEG SmartEstimate of NT$2.83 trillion, according to Reuters as republished by The Star. September alone reached NT$1.16 trillion, up 38%, the first month above NT$1 trillion.
One caution on the revenue figure. We could not open the Q3 release, so the NT$3.03 trillion comes from the October monthly sales. Monthly sales and the final consolidated figure can differ slightly. Some November write-ups also quote a NT$7.7 trillion revenue number, which looks like a nine-month total, not the quarter.
Foxconn's own wording in the October statement was careful. It said AI-related operations should keep growing in Q4, and that with the seasonal peak this should "support overall performance in line with current market expectations", while it must "monitor the impact of a volatile global political and economic environment". That is not a raise.
Claimed vs verified: did Foxconn "lift" its outlook?
This is the main correction. Some summaries of the day's results say Foxconn lifted its yearly outlook, and one even says it raised outlook across all four product segments. Here is what we could check.
| Claim in circulation | What we could verify |
|---|---|
| Foxconn lifted its full-year 2026 outlook | Its August 12 release said the full-year outlook for strong growth "remains unchanged". The Q3 coverage we could read does not report a raise. No raise is documented. |
| The outlook was raised for all four product segments | We found no segment-by-segment raise in any readable source. Foxconn's segment commentary was directional (growth), not numeric. |
| Other aggregator claims, such as a future cloud revenue share or specific consumer products | We could not trace these to a Foxconn document, so treat them as unconfirmed. |
A reasonable reading: the quarter beat expectations, and the company sounds confident about Q4 and 2027. The word "raised" is an editorial choice by some headline writers, not a documented company action. That distinction matters because a raise would change how analysts model 2027 AI capacity, while a repeat does not.
Why profit grew slower than revenue
Revenue rose 47% and profit rose 34%. That is not a problem. It is how rack-scale AI servers work for an assembler.
In Q2 2026 Foxconn's net margin was 2.37%, its operating margin 3.75% and its gross margin 6.12%, from the August release. Foxconn passes through very expensive components, mostly Nvidia GPUs, memory and networking, so each rack adds a lot of revenue and only a thin slice of profit. When the mix tilts toward racks, revenue grows faster than margin.
Management gave the same explanation in August. Rotating CEO Michael Chiang said Foxconn's advantage in AI is providing complete end-to-end systems rather than single products. End-to-end integration is one lever that helps protect a thin margin when unit prices are high.
The first-half numbers from the August release show the pattern. First-half revenue was NT$4.65 trillion, up 35%. Operating profit was NT$170.5 billion, up 65%. Net profit was NT$109.9 billion, up 27%. Operating profit is growing faster than net profit because interest costs rose: Foxconn said net interest expense grew by NT$6.6 billion a year as capital spending reached NT$80.9 billion in the half.
What Foxconn said about AI racks, switches and 2027
The August release is the most detailed official statement we could read, and it frames the Q3 result.
- Vera Rubin racks. Foxconn said Vera Rubin (VR) racks were entering mass production in Q3, and that AI rack revenue should rise quarter on quarter, with shipments up by a high double-digit percentage. See our explainer on the Vera Rubin NVL72 claims for what Nvidia says the platform delivers.
- 2026 rack shipments. Full-year AI rack shipments should "more than double".
- Networking. Revenue from 800G and faster switches should double for the year, and co-packaged optics (CPO) switches were on track for mass production and shipment in Q3.
- 2027 demand. In August, rotating CEO Michael Chiang said customer demand for AI production capacity in 2027 should remain very strong. No number came with it.
- Capacity. In August Foxconn said capital spending would rise in Taiwan, the US, Mexico and Vietnam. In the US it named Texas, Wisconsin, Ohio and California.
This lines up with the policy track we covered in Taiwan's plan for a Texas industrial park of up to $4 billion, where the island government said it would back US parks for TSMC and Foxconn suppliers.
Green computer chip in front of three tall server towers, representing AI data center demand behind Foxconn rack orders
How does this fit with the rest of the AI hardware chain?
Foxconn sits downstream of the chip makers and upstream of the cloud providers. When its revenue climbs 47%, it tells you the racks are physically leaving the factory, not only being ordered.
| Layer | Recent signal on explainx.ai |
|---|---|
| Foundry | TSMC Q3 2026 record profit |
| Memory | SK hynix record Q3 operating profit and HBM4 and Samsung's record Q3 |
| Server assembly | Foxconn (this post), and Super Micro's preliminary quarter |
| Constraints | Dell on the memory shortage lasting into 2027 and the PCB and chip-substrate shortage |
Read together, the chain shows demand outrunning parts supply at every step. Foundries, memory makers and assemblers are all reporting records, while memory and substrates stay tight. That mix is why rack prices stay high and why Foxconn's revenue can grow much faster than its profit.
What this means for people building with AI
Foxconn's result does not change an API price tomorrow. It does tell you three practical things.
- Capacity is still being added. Racks are shipping in large numbers and 2027 orders look strong. Expect GPU supply for cloud rentals to keep growing, though memory constraints can still limit how fast.
- Newer platforms are arriving. Vera Rubin entering mass production in Q3 means cloud providers can begin offering that generation. If your workloads depend on the newest Nvidia systems, watch your provider's availability announcements.
- Cost relief is not guaranteed. Records at every layer, plus shortages in memory and substrates, suggest suppliers keep pricing power. Do not assume compute gets cheaper just because more racks ship. Our neocloud shakeout coverage shows the risk on the other side: smaller providers can still struggle if demand shifts.
What is still unclear
- Foxconn's own Q3 release. We relied on Reuters-based coverage and Foxconn's earlier releases. The final consolidated revenue, gross margin and exact segment shares should be read from Foxconn's investor page once you can open it.
- AI-only revenue. Foxconn reports cloud and networking, not AI servers alone. Any percentage labeled "AI server share" is an estimate.
- 2027 numbers. Foxconn gave no 2027 revenue, profit or unit target in the sources we could read.
- The "raised outlook" wording. Unless Foxconn publishes a clear raise, treat it as a repeat of strong growth.
How to check the numbers yourself
- Open Foxconn's press center and read the Q3 2026 results release when it appears; compare consolidated revenue with the NT$3.03 trillion monthly-sales figure.
- Check the outlook paragraph. Look for the words "raised" or "upward"; "unchanged" or "remains" means a repeat.
- Compare the profit with the LSEG consensus quoted by Reuters (NT$69.6 billion).
- Look for the cloud and networking share in the segment table before quoting a percentage.
Related reading
- AI chip supply chain, October 2026: TSMC, Samsung, GlobalFoundries
- TSMC Q3 2026 record profit and what is confirmed
- SK hynix Q3 2026 record and HBM4
- Samsung Q3 2026 record profit and HBM4 contracts
- Nvidia Vera Rubin NVL72 claims, checked
- Taiwan's Texas industrial park plan
- Super Micro preliminary quarter
- Michael Dell on the memory shortage
Figures reflect reporting available on November 12, 2026. Foxconn's final Q3 release may change rounding or add detail.