At a glance. Cloud and networking reached 48% of sales, a step from 40% at year-end. Analysts asked about Google TPU. Management refused the name and restated that ASIC-server revenue would double in 2026. The 800V architecture was not discussed.
What changed this quarter
Operating profit outran revenue. Sales were up 29%; operating profit was up 63%, to a Q1 record. Management pointed to AI-server mix and cost control. Below the line, FX losses rose NT$4.5bn and net interest rose NT$3.8bn, so net income only rose 19%.
A new rotating CEO. Chiang Chi-heng, a business-group president, became the third rotating chief executive.
The US was named as the future largest AI-server plant, with Mexico remaining a volume pair and Taiwan kept for R&D, validation and early production. The board also approved NT$30bn of extra capital for the Kaohsiung Asia Bay supercomputing centre.
The numbers
| Q1 2026 | Value | Change |
|---|---|---|
| Revenue | NT$2.1195tn (~US$67bn) | +29% YoY |
| Gross profit | NT$131.0bn (~US$4.2bn) | +30% YoY |
| Operating profit | NT$75.6bn (~US$2.4bn) | +63% YoY |
| Net income | NT$49.9bn (~US$1.58bn) | +19% YoY |
| EPS | NT$3.56 (~US$0.11) | +17% YoY |
| Gross margin | 6.18% | up YoY |
| Operating margin | 3.57% | up YoY |
| Net margin | 2.36% | — |
| ROE (quarter) | 2.88% | still aiming at 12% for the year |
Q1 capex NT$35.8bn (~US$1.1bn). Full-year capex still guided more than 30% above 2025. Operating cash inflow was NT$3.2bn, a sharp improvement on the year-earlier quarter.
Segment mix: cloud and networking about 48%. Consumer, computing and components were not re-cut as percentages on this call; the qualitative split is in the segments section below.
NVIDIA 800V read-through
Still no 800 V. The rack generation named on this call is Vera Rubin, plus CPO switches and ASIC cabinets — the 2026 build, not Kyber.
What maps onto the 800V stack is the in-house list, repeated with more force:
- Next-generation GPU AI racks are expected to take higher share than the last generation, on broader customer coverage and whole-rack delivery.
- Liquid cooling and high-speed interconnect are the in-house items management cited as the way it keeps that share.
- Consignment is the ASIC default, which keeps high-value chips off Foxconn's working capital. Some GPU customers may follow, including on more customised platforms such as Vera Rubin.
- SiC is still on an EV clock: final validation targeted for 2026Q4. An AI-ASIC chiplet tape-out is guided for the second half.
On TPU, the wording was explicit: no comment on a single customer or product. The substitute sentence was that 2026 ASIC-server revenue would double.
Segments that moved
Cloud and networking is now close to half the group. AI racks are guided to a high-double-digit sequential rise in Q2 and more than a double for the full year. Switches at 800G and above are also guided to double; CPO optical switches start volume in 2026Q3, with full-year shipments guided to tens of thousands of units.
Consumer electronics is in a product transition. Demand is described as stable, with an AI-led hardware upgrade cycle (more memory, more compute) lifting supply-chain value. Q2 and the full year both guided to significant growth.
Computing had a better-than-feared new-product pull-in. Q2 is guided roughly flat sequentially and annually, partly on tight memory.
Components — camera modules, connectors — are growing. Q2 guided to significant growth year on year.
Guidance
- Q2 2026: significant sequential growth, strong annual growth, still led by AI servers.
- Full year 2026: strong growth, with better visibility than in March. AI remains the main engine.
- Operating margin: 3% is the floor; a print above 2025's 3.2% is described as likely.
- Capex: more than 30% above 2025, for regional manufacturing, automation and core capacity.
- North American CSP capex: cited as already above US$700bn and heading toward a trillion, as the demand backdrop rather than a Foxconn order figure.
Management Q&A
Q: Do geopolitics and higher raw-material costs put the 3% operating-margin floor at risk?
A: The buy-and-sell model passes cost swings through, so the effect on absolute profit is limited, and 3% is still treated as the basic operating level. As AI scale and self-made component content rise, this year has a real chance of beating 2025's 3.2%; management called the profit outlook cautiously optimistic.
Q: Where is the growth coming from in GPU versus ASIC chip solutions?
A: Both GPU and ASIC racks are growing hard. Coverage, depth of engagement and whole-rack delivery on the next-generation GPU AI cabinet should all rise, so share should be higher than on the last generation. Foxconn also called itself an important partner on major CSPs' ASIC programmes, and is raising self-made content in liquid cooling, high-speed interconnect and other key parts to lock in that position.
Q: With working-capital needs rising, how is cash flow managed, and can funding costs show up in profit?
A: The focus is the cash-conversion cycle. Q1 operating profit rose NT$29.1bn, well above a NT$3.8bn increase in net interest expense, which management read as evidence that the extra capital is funding core growth. The operating guardrails remain return on capital above 12% and net debt to EBITDA below 1.5x.
Q: What is the mix of buy-and-sell versus consignment, and will consignment rise with more custom work such as Vera Rubin?
A: ASIC server programmes are mainly consignment, which cuts working-capital use and lifts capital efficiency and, on ASIC, value-add. Some GPU customers may expand consignment as GB200/300 mature, on cost, inventory and supply-chain resilience — but the mode still follows the customer and the programme, not the product generation alone.
Q: Does consignment come from general servers or AI servers, and how should AI versus general servers inside cloud revenue be split?
A: Consignment is mainly from ASIC AI servers, because those chips are tailor-made and lack a liquid market price; some GB-series customers may use it too, and it does cut inventory and receivables. In Q1, AI servers already exceeded 50% of server revenue, and general servers have grown at a double-digit rate for two years; that double-digit pace is expected again in 2026, ahead of the industry, as AI data-center build-out lifts CPU boxes as well.
Q: How should the market value Foxconn, and what is the view on Google TPU and other ASIC programmes?
A: Management argued Foxconn is back on a growth-company path and should be valued that way, because operating profit is growing faster than sales; AI-server share is already around 40% and may still rise this year. It would not comment on a single customer or product such as TPU, but said ASIC-server revenue is still guided to double in 2026.
Q: How should Apple's new CEO and second-half memory-price increases be read?
A: Foxconn does not comment on a single customer or on a customer's organisational changes. The second-half market view stays positive and cautiously optimistic: memory and some component costs are up, but the effect on high-end products is more limited than on mid- and low-end, and AI is driving a new hardware-upgrade cycle — higher memory, more compute, better thermal and battery management — that should lift supply-chain value. Consumer smart products are still expected to grow significantly this year.
Q: What is the Computex theme, and will any partnerships be announced?
A: Any announcements were left as a tease for the show floor. This year's theme is AI multi-ecosystem integration, presenting the shift from AI-server supplier to "token factory" supplier, with L1 through L12, Vera Rubin, LPU and CPO on display alongside NVIDIA, Intel and AMD as a full AI data-center solution, plus Nimo Cloud, humanoid robots and a space data-center concept.
Q: How are AI-server capacity investments progressing in Mexico, the United States and Taiwan, and how will that split evolve?
A: The Americas are still the largest production base and the core of global AI-server shipments. The United States is expected to become the largest AI-server plant over time, with Mexico remaining a volume partner under a local-for-local strategy; group capex has been rising 20–30% a year, much of it for that AI capacity. Taiwan stays the R&D, validation and early-production core, with the Asia Bay supercomputing centre and similar projects lifting the high-end server and component mix.
Disclaimer
English notes on Hon Hai Precision Industry's investor conference of 14 May 2026, covering Q1 2026. Figures are as presented by management and have not been independently verified. USD equivalents are approximate at about NT$31.5 to the dollar. For reference only; not investment advice.