AI server rack shipments declined slightly in April, but full-year growth expectations remain robust.
AI summary card
AI server rack shipments declined slightly in April, but full-year growth expectations remain robust.
Morgan Stanley forecasts that GB200/300 rack shipments will exceed 70,000–80,000 units by 2026, doubling year over year; the preferred order is Wistron > Foxconn > Quanta.
- In April, total industry shipments amounted to approximately 8,300 units, down slightly by 3% month-over-month.
- Full-year shipments in 2026 are expected to reach 70,000–80,000 units, representing a year-on-year increase of over 100%.
- Institutional Preferences: Wistron > Foxconn > Quanta
- Wistron’s Q2 shipment forecast has been revised upward to 4,200 units.
- Hon Hai’s Q2 AI rack shipments are expected to reach 10,000 units.
Report interpretation
Overview
This report tracks the April 2026 rack shipments of GB200/300 NVL72 AI servers from leading original design manufacturers (ODMs). Although overall shipments declined slightly month-over-month, we maintain our view of robust full-year growth in 2026, with annual shipments expected to reach 70,000 to 80,000 units—more than doubling the approximately 29,000 units shipped in 2025. Among the major GPU‑based AI server ODMs, our preference order is Wistron > Hon Hai > Quanta.
Core views
April Shipment Performance and Structural Differentiation: In April 2026, total industry shipments of GB200/300 racks are estimated at approximately 8,300 units, down 3% month-over-month. Specifically, Hon Hai’s shipments remained flat at around 3,700 units; Quanta shipped roughly 2,100 units, slightly below March’s 2,100–2,200; and Wistron reported 1,300–1,400 rack-equivalent units, marginally lower than its March range of 1,500–1,600. Industry analysts note that the actual number of racks delivered to end customers may fall short of these figures, as the statistics include Wistron’s L10‑equivalent rack counts but do not fully account for the assembly and testing time required at the L11 stage. Recent Developments and Outlook for Major OEMs: Quanta (2382.TW): April revenue totaled NT$340 billion, down 6% month-over-month but up 121% year-over-year. The month-over-month decline was primarily driven by a 34% drop in notebook shipments to 3.5 million units. However, on the GB200/300 rack front, April shipments reached approximately 2,100 units, with analysts projecting that Quanta could still achieve over 7,000 rack shipments in Q2 2026. Wistron (3231.TW): April revenue stood at NT$283 billion, down 15% month-over-month but up 112% year-over-year. The decline stemmed from reduced revenues at Wiwynn, weaker notebook and desktop/display shipments, and a slight pullback in GB200/300 server rack‑equivalent shipments. Nevertheless, April’s rack‑equivalent shipments exceeded expectations, suggesting potential upside in the second quarter. Accordingly, the firm has revised its Q2 rack shipment forecast for Wistron upward to roughly 4,200 units, representing a 4% MoM increase, compared with the earlier expectation of flat MoM growth. Hon Hai (2317.TW): April GB rack shipments were essentially unchanged month-over-month, at around 3,700 units. Analysts anticipate that Hon Hai will continue ramping up AI server rack shipments in Q2 2026, with quarterly volumes potentially reaching approximately 10,000 units—up 18% sequentially, slightly above the previously projected 15% growth rate. Full-Year Growth Dynamics: The firm reaffirms that 2026 will be a robust year for downstream rack assembly. Based on current trends, annual GB200/300 rack shipments are expected to range between 70,000 and 80,000 units, marking more than a 100% year-over-year increase from roughly 29,000 units in 2025. This strong growth outlook underpins a favorable outlook for the relevant ODM players.
Analysis framework
Institutional investors employ a bottom-up supply-chain tracking approach, leveraging monthly revenue data and channel checks to estimate AI server rack shipments across ODM vendors. The analytical framework focuses on disaggregating the revenue contributions of different business segments—such as traditional laptops versus AI servers—to isolate the true growth momentum of the AI segment. At the same time, by comparing month-over-month changes in shipment volumes and factoring in seasonality and new-product ramp-up trajectories, analysts refine quarterly and annual shipment forecasts. This methodology enables timely identification of short-term supply-chain fluctuations while validating the plausibility of long-term growth assumptions.
Methodology notes
AI Server Supply Chain Tracking
By tracking the monthly shipment volumes and revenue mix of upstream ODM vendors, we can gauge the underlying demand dynamics for downstream AI infrastructure deployment. This high-frequency data monitoring helps validate whether sector prosperity is sustainable.
Revenue Breakdown and Business Contribution Analysis
Break down the company’s total revenue into its various product lines—such as laptops, servers, and other components—and, by analyzing the quarter-over-quarter and year‑over‑year changes in each segment, determine whether key growth drivers (e.g., AI servers) are offsetting the decline in traditional businesses (e.g., PCs).
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Wistron (3231.TW)Benefit: AI server compute tray shipments exceeded expectations, and Q2 guidance has been raised.
- Strengths
- Strong execution and optimization of the product mix are poised to drive an improvement in underlying profit margins.
- Weaknesses
- If the divestiture of the traditional consumer electronics business proceeds more slowly than expected, it could weigh on overall performance.
- Comparison
- Institutional favorites, with preference ratings higher than those for Foxconn and Quanta.
- Risks
- NB demand weaker than expected; intensifying competition has led to margin compression.
- Hon Hai (2317.TW)Benefit: AI server rack shipments are growing steadily, with strong Q2 expectations.
- Strengths
- Strong economies of scale; stronger-than-expected iPhone sales could provide additional support.
- Weaknesses
- If the AI business’s profit contribution falls short of expectations, it could weigh on the company’s valuation.
- Comparison
- Institutional second choice, with the largest shipment base.
- Risks
- iPhone sales fell short of expectations; geopolitical risks
- Quanta Computer (2382.TW)Benefit: AI server shipments remain steady, but are weighed down by the decline in the PC business.
- Strengths
- It has deep expertise in the large-scale data center sector.
- Weaknesses
- The sharp decline in the notebook business has offset some of the growth benefits from AI.
- Comparison
- The third choice for institutions faces short-term pressure from the PC cycle.
- Risks
- Weak demand for NB products; rising labor costs are putting pressure on profit margins.
Key data
- Total industry shipments in April 2026Approximately 8,300 unitsDown 3% month-on-month
- 2026 Full-Year Shipment Forecast70,000–80,000 unitsYear-on-year growth exceeds 100% compared to 2025 (~29k).
- Quanta’s April rack shipments~2,100 unitsSlightly below the March level.
- Wistron’s Q2 Rack Shipment Outlook~4,200 unitsA 4% month-over-month increase, compared with the prior expectation of flat performance.
- Hon Hai’s Q2 AI rack shipment outlook~10,000 unitsA month-over-month increase of 18%
Impact & implications
For investors, despite a slight month-over-month pullback in April’s data, this reflects more of a short-term fluctuation than a trend reversal. The full-year forecast of doubling growth underscores the continued robust demand for AI servers. Wistron, benefiting from stronger-than-expected shipments of compute trays and an upgraded Q2 guidance, demonstrates strong execution capabilities and the potential for upside surprises; Foxconn, supported by its sizable base and steady sequential growth, continues to reinforce its leading position in AI server manufacturing; while Quanta, weighed down by its PC business, has maintained stable AI rack shipments in line with expectations. Institutions maintain their preference order of “Wistron > Foxconn > Quanta” and recommend closely monitoring each vendor’s actual delivery performance in the second quarter.
Risks
- AI server penetration has been slower than expected.
- Demand for notebook computers (NBs) remains persistently weak.
- Price competition in the large data center sector is intense.
- Geopolitical developments have had a negative impact on foreign investment.
- Rising labor costs have resulted in weak profit margins.
What to watch
- The actual AI rack shipment performance of each ODM vendor in the second quarter of 2026
- The yield rate and delivery schedule of the new GB200/300 products
- Signs of a recovery in demand for the traditional PC business (notebooks/desktops)