April AI Rack Shipments Steady, Full Year Outlook Doubles, Prefer Wistron

Overview

This research report is a monthly tracking report by Morgan Stanley focusing on the Greater China and Asia Pacific tech hardware industry, analyzing shipment performance of major ODM vendors in the GB200/300 NVL72 AI server rack sector for April 2026. Core conclusion: Industry momentum remains strong, forecasting 2026 full year rack shipment volume to reach 70,000-80,000 units, achieving over 100% YoY growth. Although overall April shipments declined slightly MoM, major vendors including Hon Hai, Quanta, and Wistron maintained strong momentum, with Wistron receiving higher evaluation due to exceedance expectations in compute tray shipments.

Core views

Industry Total & Trends: The report estimates April 2026 GB200/300 total rack output to be approx 8.3k units, down 3% MoM. Despite slight fluctuations in single months, the firm maintains 2026 full year shipment forecast of 70,000-80,000 units, compared to approx 29,000 units last year, anticipating YoY growth exceeding 100%. This indicates downstream rack assembly business will continue to maintain strong growth momentum this year. Major Vendor Performance Differentiation: Hon Hai (Hon Hai) GB rack shipments in April were flat MoM, approx 3.7k units, and projected Q2 AI server rack shipments to continue growing MoM, reaching approx 100k units (MoM +18%). Quanta (Quanta) April revenue was NT$34 billion (MoM -6%, YoY +121%), revenue MoM decline mainly affected by reduced notebook shipments, but its GB200/300 rack shipments were approx 2.1k units, slightly lower than March's 2.1-2.2k units, projected to complete over 7k unit rack shipment targets in Q2. Wistron (Wistron) April revenue was NT$28.3 billion (MoM -15%, YoY +112%), its GB200/300 server compute tray shipments were approx 1.3-1.4k units equivalent racks, though slightly declining MoM but exceeding firm expectations, suggesting upside space in Q2, therefore revising Q2 rack shipment forecast from previous flat to approx 4.2k units (MoM +4%). Investment Preferences & Logic: Among major GPU AI server ODM vendors, Morgan Stanley's preference order is Wistron > Hon Hai > Quanta. This ranking is based on comprehensive assessment of each company's progress in AI server business, margin improvement potential, and extent of non-AI business (such as notebooks, consumer electronics) drag. Notably, the report indicates actual rack quantities delivered to end customers may be lower than statistical figures, as statistics include Wistron compute trays (L10) equivalent racks, without fully deducting L11 level rack assembly and testing time.

Analysis framework

The firm adopts a Bottom-up supply chain tracking method, inferring overall industry momentum by aggregating monthly shipment data from major ODM vendors (Hon Hai, Quanta, Wistron, etc.). The analysis logic first focuses on physical shipment volume of core products GB200/300 NVL72 racks, using them as the most direct indicator to measure AI infrastructure demand. Subsequently, combining each company's overall revenue structure (such as distinguishing AI servers from traditional notebooks, consumer electronics businesses), analyze how fluctuations in a single business segment affect overall performance, thereby more accurately assessing each company's true benefit from the AI trend. Finally, based on quarterly shipment rhythm deduction, dynamically adjust forecasts for upcoming quarters, and based on this give relative preference sorting for individual stocks.

Methodology notes

• Industry/Industrial Analysis FrameworkVolume-price decomposition

  Verify industry demand through tracking physical shipment volume (Volume) of core hardware (e.g., GB200/300 racks), rather than relying solely on financial revenue data.

  In the AI server industry, revenue may be affected by product mix or price fluctuations, while physical shipment volume reflects downstream demand strength more truly. This report judges whether the industry is in a high prosperity cycle by breaking down each vendor's rack shipment quantity.

• Valuation MethodRIM Residual Income Model

  Use Residual Income Model to perform fundamental valuation for Hon Hai, Quanta and Wistron.

  This is a valuation method based on book value and future excess returns. The report sets specific equity cost (e.g. 8.5%-9.0%), mid-term growth rate and terminal value growth rate to calculate the company's intrinsic value, supporting its overweight rating.

• Industry/Industrial Analysis FrameworkSupply Chain Upstream-Midstream-Downstream Transmission

  Focus on time difference impact of assembly and testing links from components (compute tray L10) to full rack (L11) on shipment statistics.

  The report specifically points out that since statistical caliber includes partial uncompleted final assembly/testing component equivalents, resulting statistical shipments higher than actual terminal delivery. Understanding this supply chain transmission lag helps investors view short-term shipment data volatility more rationally.

Asset mapping & comparison

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

• Wistron (3231.TW)

  Primary target, benefiting from exceedance in AI server compute tray shipments and revision up of Q2 shipment forecast.

  Strengths

  GB200/300 server compute tray shipments exceeded expectations, suggesting upside space in Q2; potential product mix optimization bringing margin expansion.

  Weaknesses

  Speed of consumer electronics business spin-off may be drag if slower than expected; Notebook demand weakness.

  Comparison

  Top ranked preference among ODM vendors, superior to Hon Hai and Quanta.

  Risks

  Consumer electronics business spin-off slower than expected; Notebook demand weaker than expected; Sales shortfall and fierce competition leading to margin contraction; AI server penetration rate improvement slower than expected.

• Hon Hai (2317.TW)

  Core beneficiary target, holding largest rack shipment share, Q2 expectation continues to grow MoM.

  Strengths

  April shipments flat MoM at high level (3.7k units), Q2 AI rack shipment expectation grows MoM 18% to 100k units.

  Weaknesses

  iPhone sales situation still affects overall sentiment; Electric vehicle business development progress slowing down may affect valuation.

  Comparison

  Preference ranking second, obvious scale advantage.

  Risks

  iPhone sales below expectation; AI business profit contribution small; EV business development slow; Geopolitical developments impact foreign investment inflow.

• Quanta (2382.TW)

  Main participant, rack shipments stable, but dragged by traditional notebook business decline in revenue.

  Strengths

  GB200/300 rack shipments stable around 2.1k units, expect to complete target above 7k units in Q2.

  Weaknesses

  April revenue MoM decline 6%, mainly affected by significantly reduced notebook shipments (MoM -34%).

  Comparison

  Preference ranking third, need to watch traditional business decline risk.

  Risks

  Notebook demand weaker than expected; Apple Watch demand weak; Labor cost increases and sales shortfall leading to weak margin performance; Large data center segment price competition fierce; AI server penetration rate improvement slower than expected.

Key data

• April GB200/300 Total Shipments~8.3k unitsDown 3% MoM
• 2026 Full Year Shipment Forecast70-80k unitsExpected YoY growth over 100% (last year approx 29k units)
• Hon Hai April Rack Shipments~3.7k unitsFlat MoM
• Quanta April Rack Shipments~2.1k unitsSlightly lower than March's 2.1-2.2k units
• Wistron April Equivalent Rack Shipments1.3-1.4k unitsExceeded expectations, suggesting upside space in Q2
• Wistron Q2 Shipment Forecast Revision Up~4.2k unitsGrowth MoM 4%, prior forecast was flat MoM

Impact & implications

The report believes 2026 will be a strong year for downstream rack assembly business, major ODM vendors will continue to benefit from increased capital expenditure on AI infrastructure construction. For investors, although overall industry is favorable, individual stock performance will depend on their ability to gain market share in AI server field and extent of drag from non-AI businesses. Wistron is assigned highest preference due to exceedance performance in AI server compute tray shipments and potential margin improvement; Hon Hai follows with huge shipment scale and stable growth; Quanta needs to focus on traditional notebook business decline offsetting overall revenue.

Risks

• AI server penetration rate improvement speed slower than expected
• Notebook and other traditional businesses demand weak dragging overall revenue
• Large data center field price competition fierce leading to margin contraction
• Geopolitical developments may have negative impact on foreign investment
• Consumer electronics business spin-off or EV business development progress not meeting expectations

What to watch

• Achievement rate of actual shipments of major ODM vendors' GB200/300 racks in Q2
• Progress of Wistron consumer electronics business spin-off
• Hon Hai iPhone sales situation and EV business progress
• Quanta notebook and Apple Watch demand changes
• Profitability performance and price competition status of AI server business