Wiwynn Corp (6669): AWS T3 is expected to offset general-server moderation and extend Wiwynn’s AI-server growth into 2027.
JPMorgan keeps Wiwynn Overweight with a Jun-27 target price of NT$2,520. Stronger general-server demand lifts 3Q26 estimates, while AWS T3, Vera Rubin and AMD ramps are expected to sustain the inference-driven growth cycle after 4Q.
Summary
JPMorgan keeps Wiwynn Overweight with a Jun-27 target price of NT$2,520. Stronger general-server demand lifts 3Q26 estimates, while AWS T3, Vera Rubin and AMD ramps are expected to sustain the inference-driven growth cycle after 4Q.
- 3Q26 revenue and earnings estimates rise 10% on stronger general-server momentum.
- AWS T3 is projected to represent about 60% of 4Q revenue and offset a modeled 12% sequential decline in general-server shipments.
- JPMorgan raises 2027E/2028E revenue by 7%/6% and EPS by 5%/2%.
- AI-related revenue mix could approach about 70% in 2027, according to the report.
Report Interpretation
Overview
This earnings-review update argues that Wiwynn’s growth driver is shifting from unusually strong general-server demand to AWS T3 in 4Q26, without undermining the broader inference-led server upcycle. JPMorgan maintains Overweight and raises longer-term estimates as ASIC and GPU opportunities build into 2027.
Core views
JPMorgan raises its 3Q26 earnings estimate by 10% because general-server momentum has been stronger than expected, but modestly reduces its 4Q26 expectation as the report anticipates normalization in general servers and a slower VR200 mass-production schedule. The institution does not interpret the expected moderation as a demand inflection: management attributed year-to-date strength to resilient general-server demand, which represented more than 50% of revenue. JPMorgan models 3Q26 general-server shipments to Meta and Microsoft rising 23% and 22% quarter on quarter, respectively, followed by an approximately 12% sequential decline in 4Q26 from a high base as platforms transition. The report’s central argument is that AWS T3 will take over as the principal 4Q growth driver. T3 is entering mass production in 4Q, with initial shipments of the liquid-cooled rack already under way and tracking ahead of schedule. JPMorgan expects it to contribute roughly 60% of Wiwynn revenue in 4Q and to more than offset softer general-server demand; management expects ASIC revenue to exceed 50% of sales as early as September. The report sees no evidence of Trainium-related order cuts or shipment constraints and forecasts AWS ASIC revenue growth of 62% in 2026 and 48% in 2027. Further adoption of liquid-cooled T3 racks is expected to sustain this momentum into 2027. The broader demand thesis rests on inference and agentic-AI workloads. JPMorgan’s research indicates that CPU demand excluding headnodes could grow 37% in 2027 and 27% in 2028, while Wiwynn’s server shipments to Meta and Microsoft are forecast to rise about 30% year on year in 2027. Meta’s recently launched Muse AI agent could add to general-server demand: JPMorgan estimates that each 100 million Muse monthly active users could generate incremental demand for 0.6 million CPUs and 0.3 million general servers, producing an incremental revenue TAM of US$4.395 billion for Wiwynn. This equals 38% of Wiwynn’s estimated 2026 Meta revenue, or 10% of total revenue, based on assumptions including three hours of daily usage, a 30% vCPU usage ratio, a 2:1 virtual-to-physical-core ratio, and 80% of vCPU use allocated to agentic-AI workloads. GPU servers provide another 2027 growth leg, although timing is partly deferred. JPMorgan expects an initial Vera Rubin ramp in 4Q26 but lowers second-half shipment assumptions because some orders may move into 2027. It forecasts approximately 1,700 Vera Rubin rack units and 2,400 AMD MI400 racks in 2027. The AMD forecast is deliberately below indicated orders of 8,000–10,000 racks because the report assumes cloud service providers will prioritize constrained memory for Nvidia platforms. GPU servers are projected to account for about 24% of 2027 revenue; better-than-feared AMD component availability would be upside to the 2027 forecast. Wiwynn is also reportedly bidding for Google TPU v9 projects, which JPMorgan identifies as a longer-term upside driver. The estimate revision reflects this changing mix. JPMorgan raises 2026E/2027E/2028E revenue by 2%/7%/6% to NT$1,377,915 million, NT$2,474,996 million and NT$3,325,328 million, respectively. Adjusted EPS is revised to NT$118.1, NT$153.6 and NT$188.5, changes of 2%, 5% and 2%; 3Q26 EPS is raised 10% to NT$32.6, while 4Q26 EPS is lowered 1% to NT$33.0. The firm expects a more favorable 4Q26 gross-margin and operating-margin mix due to a lower contribution from NVL72 racks. Its Jun-27 target price of NT$2,520 is based on approximately 15x 12-month forward EPS, above mid-cycle valuation multiples because of expected strong earnings growth. The report notes the shares had underperformed the TAIEX by 15% since September amid supply-chain noise and dilution concerns.
Analysis framework
JPMorgan combines management commentary, shipment and product-ramp forecasts, customer-demand analysis, scenario analysis for Meta’s Muse workload, earnings revisions and forward P/E valuation. It assesses how general-server, ASIC and GPU volumes and product mix affect revenue, margins and earnings through 2028.
Methodology notes
Server demand-and-supply analysis
The report assesses inference workload demand, component constraints, customer shipments and platform transitions to determine whether general-server moderation signals a cycle downturn or a mix shift.
AI workload to server-demand transmission
The Muse scenario translates user activity and virtual CPU assumptions into physical CPUs, general servers and potential Wiwynn revenue.
Forward P/E valuation
JPMorgan sets the Jun-27 target price using approximately 15x 12-month forward EPS, citing strong earnings growth.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Wiwynn Corp (6669.TW)Primary covered company; expected to benefit from AWS T3, general-server demand and GPU-server ramps.
- Strengths
- AWS T3 mass production, liquid-cooled rack adoption, resilient general-server demand and potential Meta Muse-related demand.
- Weaknesses
- General-server demand is expected to normalize in 4Q26 and some Vera Rubin orders may slip into 2027.
- Comparison
- AMD MI400 assumptions of 2,400 racks in 2027 are below indicated orders of 8,000–10,000 because cloud providers may prioritize memory for Nvidia platforms.
- Risks
- Supply constraints, weaker-than-expected server demand, and potential delays in platform ramps.
Key data
- 3Q26 earnings estimate revision+10%Raised on stronger-than-expected general-server momentum.
- 4Q26 general-server shipment outlook~12% QoQ declineFor Meta and Microsoft shipments off a high base as platforms transition.
- AWS T3 4Q revenue contribution~60%Expected to become the primary sequential growth driver.
- AWS ASIC revenue growth62% in 2026; 48% in 2027JPMorgan forecast.
- 2027 GPU-server revenue contribution~24%JPMorgan forecast.
- 2027E revenueNT$2,474,996 millionRaised 7% versus prior estimate.
- 2027E adjusted EPSNT$153.6Raised 5% versus prior estimate.
- Target priceNT$2,520.00Jun-27 target based on ~15x 12-month forward EPS.
Impact & implications
The report expects Wiwynn’s revenue growth to broaden from general servers to AWS ASIC and GPU platforms. It argues that this mix transition preserves the multi-year inference-driven growth case despite near-term normalization in general-server shipments, while component availability and execution on new ramps remain decisive.
Risks
- Supply constraints or component constraints could limit shipments.
- Server demand could be weaker than expected.
- A slower VR200 mass-production schedule or further order slippage could delay GPU-server revenue.
What to watch
- AWS T3 mass-production execution, liquid-cooled rack adoption and its 4Q revenue contribution.
- General-server shipment trends at Meta and Microsoft after the expected 4Q normalization.
- Vera Rubin and AMD MI400 ramp timing and component availability.
- Meta Muse user adoption and potential TPU v9 design wins.