AIFEATURE

Taiwan's AI Server ODM 'Big Five': Moats, Product Tiers, and Demand Exposure in Q2 2026

E
EffectStory 編輯部Editorial Team
Published · Updated
Foxconn, Quanta, Wistron, Wiwynn, and Inventec all posted double-digit Q2 2026 revenue growth on AI servers, but their moats diverge sharply: Foxconn scales through diversification, Quanta and Wiwynn concentrate on AI/hyperscale demand, and all five disclose AI exposure only as percentages, not absolute dollar figures, while Wiwynn's new consignment accounting further complicates comparability.

How do the Big Five differ in scale and AI server growth speed?

Quanta (廣達) posted the fastest year-over-year revenue growth among the five in Q2 2026, while Foxconn (鴻海) remained the largest by absolute revenue CITE:E1. Foxconn's consolidated revenue reached NT$2.53 trillion, up 41% year-over-year, with after-tax net profit of NT$59.97 billion and EPS of NT$4.27 CITE:E1. Quanta's consolidated revenue was about NT$1.04 trillion, up 105.6% year-over-year CITE:E2. Wistron (緯創) reported consolidated revenue of NT$895.443 billion, up 62.43% year-over-year and 5.81% quarter-over-quarter — a single-quarter record — with after-tax net profit of NT$14.827 billion, up 54% quarter-over-quarter CITE:E3. Wiwynn (緯穎) posted consolidated revenue of NT$278.2 billion, up 26% year-over-year, with EPS of NT$80.43 CITE:E4. Inventec (英業達) reported consolidated revenue of NT$269.9 billion, up 44.6% year-over-year CITE:E5.

CompanyQ2 2026 RevenueYoY GrowthProfit / EPSAI Server Metric
FoxconnNT$2.53 trillion+41%Net profit NT$59.97B; EPS NT$4.27Cloud & networking 51% of revenue (up from 48% in Q1)
Quanta~NT$1.04 trillion+105.6%Not disclosed in evidenceAI servers 75–80% of server revenue
WistronNT$895.443 billion+62.43% YoY / +5.81% QoQNet profit NT$14.827B (+54% QoQ)AI server business drove profit to record
WiwynnNT$278.2 billion+26%EPS NT$80.43Nearly all cloud/hyperscale white-box servers
InventecNT$269.9 billion+44.6%Not disclosed in evidenceAI server rack shipments drove growth

What different product-tiering strategies do the five pursue — diversified vs. specialized?

Foxconn spreads revenue across three segments, while Wiwynn and Quanta concentrate almost entirely on AI/cloud server demand CITE:E1CITE:E4. Foxconn's cloud & networking products, led by AI servers, made up 51% of consolidated revenue in Q2 — its first time surpassing half, up from 48% in Q1 — alongside consumer smart products at about 29% and computing products at about 15% CITE:E1. Quanta concentrates more narrowly on servers, with AI servers making up about 75–80% of its overall server revenue CITE:E2. Wiwynn's business is almost entirely cloud and hyperscale white-box servers, with large cloud operators including Meta and Microsoft as its main customers CITE:E4. Inventec's Q2 growth was driven mainly by strong shipments of AI server racks, and the company said motherboards (L6) will become its main shipment product in the second half CITE:E5. Wistron attributed its record quarterly profit to continued growth in its AI server business, without a separate product-mix breakdown in the evidence CITE:E3.

How is each company's moat built from scale, specialization, and customer depth?

Foxconn's moat combines the largest absolute revenue among the five with a diversified mix — 51% cloud & networking, 29% consumer smart, 15% computing CITE:E1. Quanta's moat rests on concentration in AI servers, which reached 75–80% of its server revenue and coincided with the fastest growth rate among the five, 105.6% year-over-year CITE:E2. Wiwynn's moat is built on customer depth rather than diversification: its business is almost entirely cloud and hyperscale white-box servers for large cloud operators including Meta and Microsoft CITE:E4. Inventec's position leans on product-cycle timing — integrated-rack shipments drove Q2 growth of 44.6%, with L6 motherboards set to become the main shipment product in the second half CITE:E5. Wistron's moat shows up in execution: AI server growth pushed net profit to NT$14.827 billion, up 54% quarter-over-quarter, and revenue to a single-quarter record of NT$895.443 billion CITE:E3.

How do profitability and margin structures differ, and how does an accounting change affect comparability?

Wistron posted the sharpest profit jump among the disclosed figures, with net profit up 54% quarter-over-quarter to NT$14.827 billion CITE:E3. Foxconn's after-tax net profit was NT$59.97 billion with EPS of NT$4.27 CITE:E1. Wiwynn's EPS was NT$80.43, but comparability with prior periods is complicated: starting in April 2026, Wiwynn shifted some customers' memory purchases to a consignment model, under which the related amounts are excluded from both revenue and cost, changing the year-over-year comparison basis for its revenue and gross margin CITE:E4CITE:E8. Quanta's and Inventec's disclosures in the evidence do not include net profit or EPS figures for the quarter CITE:E2CITE:E5.

What demand exposures do the five face — customer concentration, product-cycle shifts, and disclosure gaps?

Wiwynn carries customer-concentration exposure, since its nearly all-cloud/hyperscale-server business depends on large cloud operators including Meta and Microsoft as its main customers CITE:E4. Its April 2026 shift to a consignment model for some customers' memory adds a comparability risk on top of that concentration, since the related amounts are excluded from both revenue and cost CITE:E8. Inventec faces a product-cycle transition: Q2 growth came from AI server rack shipments, but the company expects motherboards (L6) to become the main shipment product in the second half CITE:E5. Across the group, disclosure is limited to percentage-based breakdowns rather than absolute AI server revenue: Quanta discloses AI servers as 75–80% of overall server revenue and Foxconn discloses cloud & networking as 51% of consolidated revenue, but neither publishes a standalone AI server revenue figure CITE:E7.

What this means

The same evidence set shows both convergence and a shared blind spot. All five companies grew Q2 2026 revenue on AI server demand — from Wiwynn's 26% to Quanta's 105.6% — yet Quanta's fastest growth rate and Foxconn's 51%-of-revenue milestone are both reported as percentages, with no absolute AI server revenue figure disclosed by either company CITE:E1CITE:E2CITE:E7. That gap sits alongside Wiwynn's April 2026 consignment accounting change, which alters the comparison basis for its own revenue and margin even before cross-company comparison is attempted CITE:E8. Wistron's record quarterly profit and Inventec's rack-to-motherboard product shift, meanwhile, show that within-company execution and product-cycle timing — not just AI exposure level — are driving the profit outcomes CITE:E3CITE:E5.

📊 Evidence

📎 Sources

  1. cna.com.tw
  2. chinatimes.com
  3. m.cnyes.com
  4. wiwynn.com
  5. ec.ltn.com.tw
Author's TakeEffectStory 編輯部

The number worth watching in this set isn't the growth rate — it's the disclosure format. Quanta's 75–80% and Foxconn's 51% are both AI-exposure metrics expressed as shares of a larger base, not as standalone AI server revenue, which means the fastest-growing company (Quanta, +105.6%) and the largest company (Foxconn, at NT$2.53 trillion) can't actually be benchmarked against each other in absolute AI dollars. Wiwynn's April 2026 shift to consignment accounting for some customers' memory adds a second layer of noise, since it changes the YoY base for revenue and margin even within a single company. The metric to track next quarter is whether any of the five moves from percentage disclosure to an absolute AI server revenue line — and whether Wistron's 54% quarter-over-quarter profit jump holds up as a repeatable pattern rather than a single-quarter record.

E
EffectStory 編輯部Editorial Team

Related

BRIEF

Cursor's SpaceX-Backed Origin Launches Hours Before a Global GitHub Outage

Cursor began rolling out Origin, a Git hosting platform built for AI agents, to paid users on August 17, 2025, weeks after SpaceX completed a reported $60 billion acquisition of the company. Roughly three and a half hours into the rollout, GitHub suffered a global service disruption that GitHub's own status page and outside accounts measured at between six hours 42 minutes and seven hours 35 minutes, against a backdrop GitHub's CTO has described as under-provisioned for current demand.

Nathan ·
BRIEF

Apple Scraps Per-Install App Store Fees in the EU, Shifts to 5% Transaction Commission

Apple will replace its installation-linked Core Technology Fee with a single 5% Core Technology Commission for EU apps distributed outside the App Store, effective October 1, 2026. The overhaul also revises in-app commission tiers, eases third-party marketplace requirements, and adds child-protection rules, following a €500 million DMA fine and recent dialogue with the European Commission.

EffectStory 編輯部 ·
BRIEF

Hiwin Pushes Into CPO Precision Motion and Humanoid Robot Components at TAIROS 2026

Hiwin Technologies (上銀) occupied a flagship booth of more than 750 square meters at TAIROS 2026, supplying miniature ball screws and single-axis robots for co-packaged optics (CPO) sample testing, a planetary roller screw with 2-3 times the load capacity of conventional ball screws now production-ready, and a 14-actuator dual-arm humanoid system. Its dual-arm and 8-axis logistics robots were both pushed to November 2026 resampling with 2027 mass production targeted, while ball screw and linear guide output rise 20% and 10-15% respectively amid lead times of four to six months. Hiwin set a double-digit percentage revenue growth target for 2026 with capex near NT$2.7 billion.

EffectStory 編輯部 ·