GB200/GB300 NVL72-equivalent rack output by major server ODMs Report Interpretation

Overview

This monthly Greater China technology-hardware tracker examines GB200/GB300 NVL72-equivalent rack output among major server ODMs. Morgan Stanley finds that August production accelerated and argues that AI-server demand is creating upside to near-term and, for Quanta and Wistron, full-year shipment expectations.

Core views

Morgan Stanley estimates industry-wide GB200/GB300 NVL72-equivalent rack output at about 8.4K in August, up 27% month on month. The report forecasts 80-85K racks for full-year 2026, compared with about 26.5K in the prior year, implying more than 100% year-on-year growth. It characterizes 2026 as a strong year for downstream rack assembly. The firm cautions that actual deliveries to end customers are likely below its reported output figures because its Wistron estimate includes L10 computing-tray rack equivalents without allowing for L11 rack assembly and testing time. Quanta shipped an estimated 2.5-2.55K GB200/GB300 racks in August, up 22% month on month. Its August revenue was NT$424.0B, up 16% month on month and 177% year on year; Morgan Stanley attributes the monthly increase mainly to higher AI-rack shipments, partly offset by notebook shipments of 2.6M units, down 7% month on month and 37% year on year. Combined July-August AI-rack shipments of about 4.6K already represent a substantial portion of Morgan Stanley's approximately 6.4K 3Q forecast. July-August revenue of about NT$790B equals 81% of Morgan Stanley's and 71% of consensus 3Q revenue estimates, which the report says points to upside to both 3Q expectations and its full-year estimate of about 18.7K racks. Wistron shipped an estimated 2.3-2.4K GB200/GB300 rack equivalents in August, roughly doubling month on month. August revenue reached NT$460.1B, up 49% month on month and 166% year on year. Morgan Stanley attributes the increase to stronger AI-server computing-tray shipments and Wiwynn revenue of about NT$144.3B, up 23% month on month and 50% year on year. These gains were partly offset by weaker notebook, desktop and monitor shipments: notebooks were 1.4M units, down 13% month on month and 36% year on year; desktops were 500K, down 29% and 38%; and monitors were 750K, down 17% and 6%, respectively. July-August revenue of about NT$768.3B represented 74% of Morgan Stanley's and 75% of consensus 3Q revenue estimates, supporting the firm's view of 3Q and full-year upside relative to its approximately 17.9K-rack full-year forecast. Hon Hai's August revenue was NT$921.8B, down 3% month on month but up 52% year on year. Morgan Stanley reports that server-related revenue delivered the strongest monthly growth, followed by components and other products, while computing products and the consumer segment, primarily iPhone assembly, declined month on month. The firm estimates GB rack shipments at about 2.6K, up 6% month on month, and leaves its approximately 31.7K full-year rack forecast unchanged. Across the major server ODMs, Morgan Stanley's preference is Wistron over Hon Hai over Quanta, based on upside to price targets. The report's positive operating interpretation rests on the acceleration in AI-rack output and the extent to which early-quarter revenue and shipment progress already covers the firms' respective 3Q expectations.

Analysis framework

Morgan Stanley tracks monthly and quarterly GB200/GB300 NVL72-equivalent rack output by major server ODMs, then compares estimated shipment volumes and reported monthly revenue with its own and consensus quarterly forecasts. It uses the resulting shipment and revenue run rates to assess potential forecast upside. The report also references residual-income valuation assumptions for the individual covered companies.

Methodology notes

• Industry AnalysisSupply-demand framework

  Monthly rack-output and shipment tracking across major server ODMs

  The report estimates AI-rack production and compares ODM shipment run rates with revenue and quarterly forecasts to judge the strength of downstream rack-assembly demand.

• Valuation methodsRIM (Residual Income Model)

  Residual-income valuation

  Morgan Stanley cites base-case residual-income valuations using company-specific costs of equity, medium-term growth assumptions and a 3% terminal growth rate.

Asset mapping & comparison

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

• Wistron Corporation (3231.TW)

  Preferred major server ODM based on upside to price target and AI-rack shipment momentum.

  Strengths

  Estimated August GB200/300 rack-equivalent shipments of 2.3-2.4K, roughly doubling month on month; August revenue rose 49% month on month and 166% year on year.

  Weaknesses

  Notebook, desktop and monitor shipments declined month on month.

  Comparison

  Ranked ahead of Hon Hai and Quanta in Morgan Stanley's stated order of preference.

  Risks

  Weaker notebook or Apple Watch demand, weak margins from labor costs and sales shortfalls, intense mega-data-center price competition, and slower AI-server penetration.

• Hon Hai Precision (2317.TW)

  Major server ODM ranked second in Morgan Stanley's stated preference order.

  Strengths

  Server-related revenue showed the strongest monthly growth; estimated GB rack shipments rose 6% month on month to about 2.6K.

  Weaknesses

  Computing products and the consumer segment, mainly iPhone assembly, declined month on month; August total revenue fell 3% month on month.

  Comparison

  Ranked behind Wistron but ahead of Quanta based on upside to price targets.

  Risks

  Lower iPhone sell-through, smaller AI-business profit contribution, slower EV-business progress, and geopolitical developments that could negatively affect foreign investment.

• Quanta Computer Inc. (2382.TW)

  Major server ODM with potential upside to quarterly and full-year rack forecasts.

  Strengths

  July-August rack shipments of about 4.6K and revenue of about NT$790B point to potential upside to 3Q and full-year forecasts.

  Weaknesses

  Notebook shipments fell 7% month on month and 37% year on year in August.

  Comparison

  Ranked behind Wistron and Hon Hai in Morgan Stanley's stated preference order.

  Risks

  Weaker notebook demand, softer Apple Watch demand, weaker server demand, and slower-than-expected AI-server penetration.

Key data

• Industry GB200/300 rack output, August 2026~8.4K racks+27% month on month
• Full-year 2026 industry rack forecast80-85K racksMore than 100% year-on-year growth versus ~26.5K in the prior year
• Quanta August revenueNT$424.0B+16% month on month, +177% year on year
• Quanta August GB200/300 shipments2.5-2.55K racks+22% month on month
• Wistron August revenueNT$460.1B+49% month on month, +166% year on year
• Wistron August GB200/300 rack equivalents2.3-2.4KApproximately +100% month on month
• Hon Hai August revenueNT$921.8B-3% month on month, +52% year on year
• Hon Hai August GB rack shipments~2.6K racks+6% month on month

Impact & implications

The report interprets the August ramp in GB200/GB300 output as evidence of robust AI-server and rack-assembly demand. It sees the strongest near-term forecast-upside signals in Quanta and Wistron because their July-August shipment and revenue progress already accounts for a large share of 3Q expectations, while Hon Hai's full-year rack forecast is unchanged.

Risks

• Actual end-customer rack deliveries may be lower than reported output because Wistron's L10 computing-tray equivalents do not account for L11 rack assembly and testing time.
• For Hon Hai, lower iPhone sell-through, weaker AI profit contribution, slower EV development and adverse geopolitical developments could weigh on results.
• For Quanta, weaker notebook, Apple Watch or server demand and slower AI-server penetration are identified downside risks.
• For Wistron, delayed consumer-electronics divestiture and margin contraction from sales shortfalls or fierce competition are identified downside risks.

What to watch

• Monthly GB200/GB300 NVL72-equivalent output and shipment progress at Quanta, Wistron and Hon Hai.
• Whether Quanta's and Wistron's July-August revenue run rates translate into upside to 3Q forecasts.
• The pace of AI-server penetration, server demand and rack assembly and testing progress.
• Notebook, desktop, monitor and consumer-electronics demand trends that may offset AI-server growth.