Taiwan ODM/Brands AI server/PC supply chain: AI-server rack ramps and new smartphones are expected to sustain Taiwan ODM revenue growth through late 2026

Overview

This three-month revenue preview assesses 10 Taiwan ODM and brand companies across the AI-server and PC supply chain. Goldman Sachs expects growth to continue into late 2026, led by next-generation rack-level AI servers, ASIC-server penetration and flagship smartphone launches, but sees materially different earnings and valuation outcomes by company.

Core views

Goldman Sachs expects the covered universe to continue expanding after average August revenue rose 9% month-on-month. It forecasts average month-on-month revenue growth of 7% in September, 2% in October and 4% in November 2026; the corresponding year-on-year growth forecasts are 49%, 57% and 52%. The central mechanism is a ramp in next-generation rack-level AI servers, including GPU-, ASIC-, CPU- and LPU-powered configurations, combined with new flagship smartphone models. Higher memory costs also lift selling prices, although they simultaneously constrain PC and broader consumer-electronics demand. The institution is positive on three supply-chain themes: Apple-related suppliers supported by new form factors, including foldable phones in 2026 and iPhone 20 in 2027; AI-server components benefiting from specification upgrades; and rising ASIC AI-server penetration. It expects AI spending to support both AI and general servers, while capacity diversification requirements encourage manufacturers to expand production across regions. Hon Hai is expected to sustain sequential monthly revenue growth through December. Goldman Sachs forecasts 3Q26 and 4Q26 revenue growth of 14% and 23% quarter-on-quarter, respectively, and 3Q26 revenue of NT$2.9 trillion, up 40% year-on-year and 14% quarter-on-quarter. The report attributes this to early delivery of next-generation AI-server racks and CPO switches, with Hon Hai expected to be about two to three months ahead of peers on the latest model, followed by LPU-powered rack shipments from December. Smartphone launches and premium/foldable-phone share gains provide a second driver, despite Goldman Sachs lowering global smartphone volume estimates by 1 million, 8 million and 0.2 million units for 2026-28 to 1.1 billion, 1.2 billion and 1.2 billion because of memory costs. For Quanta, Goldman Sachs expects modest 3Q26 and 4Q26 sequential revenue growth of 2% and 2% because PC demand is slowing after a first-half pull-forward and AI-server platforms are in transition. However, it expects next-generation AI racks to accelerate into December, with December revenue projected to rise 54% month-on-month. Quanta's July and August PC shipments declined 30% and 37% year-on-year. The report nevertheless highlights sustained AI and general-server investment, estimating capex of NT$44 billion and NT$32 billion in 2026 and 2027, versus NT$15 billion in 2025, and forecasting aggregate capex by leading US cloud service providers to grow 76%/35%/8% year-on-year in 2026-28 to US$1,145 billion by 2028. AVC is expected to benefit from rising liquid-cooling adoption as GPU and ASIC AI-server racks ramp. Goldman Sachs forecasts 3Q26 and 4Q26 revenue growth of 18% and 27% quarter-on-quarter and expects equivalent AI-server rack shipments to reach 55,000, 105,000 and 163,000 units in 2026-28. Its thesis also includes expansion with a major ASIC customer, a new ASIC customer in 4Q26, and ultra-thin vapor-chamber adoption in lighter consumer electronics and foldable phones. It forecasts AVC net-income CAGR of 39% in 2026-28. Wiwynn reported August revenue 27% above Goldman Sachs estimates, supported by general-server demand from leading US cloud service providers and new ASIC AI-server ramps. Goldman Sachs expects 3Q26 and 4Q26 revenue to grow 46% and 22% quarter-on-quarter as new air- and liquid-cooled ASIC servers, general servers serving agentic-AI demand, and GPU-powered racks across NVIDIA and AMD platforms enter production. It estimates 2026 capex of NT$44 billion, remaining near that level through 2028, to support site diversification across the US, Mexico and Malaysia and expansion across ASIC and GPU products. Wistron's August revenue rose 166% year-on-year to NT$460 billion and was 48% above Goldman Sachs estimates, reflecting strong rack-level AI-server shipments despite PC weakness and model transitions. The institution expects sequential revenue growth from September through December as GPU, CPU and LPU server-rack shipments ramp. It raised 2026-28 revenue estimates by 11%/7%/6% and net-income estimates by 2%/3%/4%, driven by higher rack shipments and ASPs. Gross-margin forecasts were reduced because rack-level AI servers carry lower gross margins than other product lines. The 12-month target price was raised to NT$295 from NT$281 as the target 2027E P/E increased to 13.2x from 13.1x. Gigabyte is expected to see accelerating 2H26 growth from AI-server racks, higher ASPs, more customer deployments and greater backlog visibility. Goldman Sachs forecasts 3Q26 and 4Q26 revenue growth of 81% and 79% year-on-year, with September and October revenue each expected at about NT$52 billion, up 87% year-on-year. It expects 2026 capex to rise 143% year-on-year to NT$2.8 billion as the company expands and diversifies production capacity. For ASUS, Goldman Sachs expects a September sequential revenue increase but a 7% quarter-on-quarter decline in 4Q26 as PC demand pull-forward and AI-server transitions weigh on the period. It remains positive on the longer-term AI-server expansion and higher-ASP new models, forecasting 3Q26 and 4Q26 revenue growth of 39% and 27% year-on-year. Compal is expected to benefit from AI-server rack ramps and migration from motherboard-level L6 production to L10/L11 systems, producing 3Q26 and 4Q26 revenue growth of 1% and 16% quarter-on-quarter. Its August revenue missed Goldman Sachs estimates by 20%, however, due to lower PC shipments and weak consumer-electronics demand. Inventec's August revenue rose 36% year-on-year but fell 8% month-on-month, broadly in line with estimates. Goldman Sachs attributes the monthly decline to shortages of memory, PCBs and CPUs and expects another sequential decline in September before recovery in October through December as supply improves. It forecasts 3Q26 and 4Q26 revenue growth of 45% and 47% year-on-year, raised 2026-28 revenue estimates by 8%/10%/6%, and lifted net-income estimates by 5%/4%/2%. The 12-month target price rose to NT$61 from NT$56 as the target 2027E P/E increased to 15.3x from 14.6x. Pegatron is expected to grow sequentially from September through December on new smartphone models, higher-ASP AI-server racks, broadband, IoT and game-console seasonality. Goldman Sachs forecasts 3Q26 and 4Q26 revenue growth of 18% and 25% year-on-year. It views expansion into AI servers and automotive electronics positively, but maintains Sell because valuation is stretched; high memory costs are also expected to limit PC and smartphone market growth.

Analysis framework

Goldman Sachs compares reported monthly revenue and notebook shipments with its estimates, then projects monthly and quarterly revenue through December 2026. It links company outcomes to AI-server platform ramps, product mix, ASPs, customer demand, capacity investment, component availability and consumer-device demand, and applies forward P/E benchmarks derived from peer relationships between earnings growth, margins and valuation multiples.

Methodology notes

• Industry AnalysisSupply-demand framework

  Monthly revenue and shipment tracking linked to AI-server, PC, smartphone and component supply-demand conditions.

  The report uses reported revenue, shipment data, customer demand, production capacity and component availability to explain near-term growth differences across the supply chain.

• Industry AnalysisVolume-price decomposition

  Revenue growth is attributed to shipment ramps, product mix and higher average selling prices.

  Goldman Sachs distinguishes the impact of higher rack shipments and specification upgrades from higher ASPs associated with advanced AI-server configurations and memory costs.

• Valuation methodsP/E and PEG Valuation

  Forward P/E targets benchmarked against peer correlations between earnings growth, operating margins and valuation.

  The institution derives company target prices by applying selected 2027E P/E multiples informed by peer earnings-growth and valuation relationships.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Hon Hai (2317.TW)

  Primary beneficiary of early next-generation AI-server rack and CPO-switch deliveries, plus flagship and foldable smartphone launches.

  Strengths

  Expected to lead peer delivery of the latest AI-server model by roughly two to three months.

  Weaknesses

  Global smartphone volume assumptions were reduced because of high memory costs.

  Comparison

  Expected to be ahead of peers on the latest AI-server rack delivery.

  Risks

  Slower AI-server ramp, weaker EV total-solution performance, slower global capacity ramp and stronger EMS competition.

• Quanta (2382.TW)

  AI-server and general-server demand support growth after a platform-transition period.

  Strengths

  High capex and US factory leasing reflect expansion for AI-server demand.

  Weaknesses

  PC shipments declined sharply and AI-server model transition constrains 2H26 sequential growth.

  Comparison

  Valuation is viewed as fair relative to PC/server peers.

  Risks

  PC recovery, AI-server ramp and general-server demand may be stronger or weaker than expected.

• AVC (3017.TW)

  Liquid-cooling supplier benefiting from GPU and ASIC AI-server rack deployment.

  Strengths

  Broad air- and liquid-cooling solution portfolio and exposure to new ASIC customer ramps.

  Weaknesses

  Growth depends on continued cooling penetration and AI-server demand.

  Comparison

  Target valuation is based on technology-hardware peer P/E and EPS-growth relationships.

  Risks

  Slower liquid-cooling adoption, weaker AI-server demand, competition and slower general-server recovery.

• Wiwynn (6669.TW)

  AI-server ODM benefiting from ASIC and GPU rack production and solid general-server demand.

  Strengths

  August revenue beat estimates by 27%; expanding capacity across the US, Mexico and Malaysia.

  Weaknesses

  Execution depends on sustained customer AI-server and general-server demand.

  Comparison

  Target P/E is derived from PC/server peer valuation and EPS-growth correlation.

  Risks

  Weaker AI-server demand, slower general-server recovery and more severe server-ODM competition.

• Wistron (3231.TW)

  Rack-level AI-server shipments drive revenue and earnings upgrades.

  Strengths

  August revenue was 48% above Goldman Sachs estimates; expanding AI-server capacity in the US and Vietnam.

  Weaknesses

  A greater mix of rack-level AI servers reduces gross-margin forecasts.

  Comparison

  The higher 13.2x 2027E target P/E remains between historical average and average-plus-one-standard-deviation trading P/E.

  Risks

  Weaker PC recovery, slower AI-server ramp and weaker general-server demand.

• Gigabyte (2376.TW)

  AI-server rack ramps, customer expansion and backlog visibility support 2H26 growth.

  Strengths

  Expected 2026 capex growth of 143% year-on-year to capture demand and diversify production.

  Weaknesses

  Growth remains exposed to PC, gaming-PC and AI-server execution.

  Comparison

  Valuation uses a PC/server peer P/E and EPS-growth relationship.

  Risks

  PC recovery, gaming-PC growth and AI-server ramp may differ from expectations.

• ASUS (2357.TW)

  AI-server expansion and higher-end PC mix support year-on-year growth.

  Strengths

  New higher-ASP AI-server models and funding for capacity expansion and diversification.

  Weaknesses

  4Q26 sequential decline is expected after PC demand pull-forward and during AI-server transitions.

  Comparison

  Target multiple is based on brand-maker relationships among P/E, earnings growth and operating margin.

  Risks

  PC, AI/gaming-PC and AI-server demand could vary from expectations.

• Compal (2324.TW)

  AI-server ramp and progression from L6 motherboards to L10/L11 systems support growth.

  Strengths

  Global capacity expansion and business diversification.

  Weaknesses

  Consumer electronics remain the main revenue source and are exposed to high raw-material costs; August revenue missed estimates.

  Comparison

  Target valuation is based on PC/server peer P/E and EPS-growth correlation.

  Risks

  PC recovery, AI-server ramp and tablet demand may differ from expectations.

• Inventec (2356.TW)

  Higher AI-server ASP supports earnings revisions despite short-term component shortages.

  Strengths

  2026-28 revenue and earnings estimates were raised; AI-server expansion remains a positive structural driver.

  Weaknesses

  Memory, PCB and CPU shortages and weak PC demand weigh on near-term revenue.

  Comparison

  Target P/E was increased to 15.3x 2027E based on peer earnings-growth and valuation correlation.

  Risks

  PC recovery, AI-server ramp and general-server demand may differ from expectations.

• Pegatron (4938.TW)

  New smartphone models, AI-server racks and seasonal broadband, IoT and console demand support late-2026 revenue.

  Strengths

  Expansion from consumer electronics into AI servers and automotive electronics.

  Weaknesses

  High memory costs constrain PC and smartphone demand; August revenue was 12% below estimate.

  Comparison

  The report considers valuation stretched despite operational growth drivers.

  Risks

  Stronger smartphone recovery, faster AI-server ramp and stronger consumer-electronics demand.

Key data

• Universe August revenue growth+9% MoMAverage across the 10 covered companies
• Universe monthly revenue growth forecast+7% / +2% / +4% MoMSeptember/October/November 2026
• Universe year-on-year revenue growth forecast+49% / +57% / +52% YoYSeptember/October/November 2026
• Wistron August revenueNT$460 billion, +166% YoY48% above Goldman Sachs estimate
• Wistron target priceNT$295Raised from NT$281; based on 13.2x 2027E P/E
• Inventec target priceNT$61Raised from NT$56; based on 15.3x 2027E P/E
• Wiwynn 2026E capexNT$44 billionExpected to remain at a similar high level through 2028
• Leading US CSP aggregate capexUS$1,145 billion by 2028EGoldman Sachs forecasts +76%/+35%/+8% YoY in 2026-28

Impact & implications

The report argues that AI-server rack deployment is shifting growth toward companies with advanced rack integration, ASIC/GPU exposure, cooling capability and capacity expansion. At the same time, weaker PC demand after first-half pull-forward, high memory costs, component constraints and platform transitions make the near-term trajectory uneven across ODMs and brands.

Risks

• AI-server ramp-up or customer deployment could be slower than expected.
• High memory costs may continue to cap PC, smartphone and consumer-electronics demand.
• Shortages of memory, PCBs and CPUs may constrain production and near-term revenue.
• Competition could intensify in server ODM, cooling and consumer-electronics manufacturing.
• PC, general-server, gaming-PC and tablet recovery may diverge from company expectations.

What to watch

• Monthly September-to-December revenue trends against the forecast 7%/2%/4% average month-on-month growth path.
• The timing and scale of next-generation GPU, ASIC, CPU and LPU AI-server rack ramps.
• AI-server capacity expansion and capex commitments by ODMs and cloud-service-provider customers.
• PC shipment trends following first-half demand pull-forward and elevated memory costs.
• Component availability for memory, PCBs and CPUs, particularly for Inventec.
• Smartphone model launches, foldable-phone adoption and consumer-electronics demand.