Quanta’s AI Server Trajectory Holds, But Rubin Standardization Poses a Threat

Core Conclusion

Quanta’s AI server growth remains on a powerful trajectory, with 2Q26 revenue guided to rise double-digits sequentially and full-year CY26 set for triple-digit year-on-year expansion. Notebook headwinds are contained and shrinking as a share of the mix. The defining medium-term risk, however, is NVIDIA’s drive to standardize computing trays for the Rubin platform. If hyperscalers resist and turn to proprietary ASIC designs, Quanta’s design-integration value could erode, introducing a potential earnings headwind visible only from CY27.

What the Market May Be Underestimating

The market may not fully price the earnings leverage from scale. Quanta plans to double its AI server production capacity by end‑2026, backed by NT$30bn in capex, split primarily between Thailand and the US. Some customers have agreed to a consignment model, shifting working-capital and capex pressure away from the ODM. This financial framework allows Quanta to accelerate investment cycles while less-funded competitors face constraints, potentially capturing above-industry volume growth — even if total server unit demand moderates. Capacity build-out, not just order flow, becomes the moat.

Evidence Chain

AI server momentum: Following a doubling of AI server revenue in CY25, guidance calls for another triple-digit y/y jump in CY26. 2Q26 is expected to deliver double-digit sequential growth, with GB300 volumes plateauing in 3Q without an abrupt drop. General server sales are guided to expand double-digits y/y for the full year.
Notebook drag contained: NB shipments are guided flat to slightly up q/q in 2Q26, and the full-year outlook remains a high-single to double-digit y/y decline. NB contributed under 15% of revenue in 1Q and will stay below 20% for the full year, blunting its profit impact.
Platform transition visibility: The GB300‑to‑VR200 migration begins in 4Q26, but VR200 volumes are mostly a CY27 event, with sampling shipments starting in 3Q26. The GB300‑heavy year provides near‑term revenue certainty.
Capacity and capex: AI server capacity will double by end‑2026. Consignment model adoption by certain customers relieves ODM funding burdens, easing the risk of supply bottlenecks and investment delays.
Standardization tension: NVIDIA aims to standardize computing trays for Rubin, but Quanta is uncertain about CSP acceptance. A forced standard could drive hyperscalers toward ASIC-based alternatives, jeopardizing the high‑value integration work that defines Quanta’s AI server income.

Key Debate and Risks

The principal debate now centers on tray standardization. If NVIDIA imposes a common tray architecture, large CSPs that prize proprietary rack designs may accelerate their push toward in-house ASIC solutions. For Quanta, whose AI growth depends on deep customization for each hyperscaler, such a migration would shrink design-in content per server and slow new project intake. The risk is not imminent — Rubin volumes won’t ramp until CY27 — but early indicators from 3Q26 sampling will be closely watched.

A secondary risk: a sharper-than-guided NB downturn could offset AI server gains, though the revenue mix has already shifted decisively below the 20% threshold. Demand elasticity in enterprise general servers also bears monitoring, but current double-digit growth guidance provides a baseline buffer.

Valuation and Trading Implications

No target price or rating change is attached to this note, but the takeaways point to a favourable near‑term setup. The AI server expansion and aggressive capacity build-out are visible catalysts that support estimates for 2H26. The stock’s valuation will increasingly hinge on the Rubin tray standardisation debate. A successful CSP adoption of a common tray could commoditize part of Quanta’s value-add, compressing multiples; conversely, if Quanta preserves architectural flexibility and customer stickiness, its AI premium could extend. The initial sampling feedback in 3Q26 will be the first real test of this narrative. Accumulate on weakness driven by the standardization overhang, but size positions to reflect the asymmetric CY27 risk.