Quanta Raises 2026 Capex to NT$40 Billion as AI Order Visibility Extends to 2028

Nashnova编辑部
Published todayAbout 12 min read

Quanta Computer lifted its 2026 capex target from NT$30 billion to NT$40 billion (~US$1.28 billion), driven by customer AI roadmaps now stretching to 2028 — a shift from build-to-order to locking in capacity three years ahead.

01

NT$40 billion in capex — where is the money going?

First-half capex reached NT$21.8 billion, with over 80% directed at plants, facilities, and equipment in California and Thailand.
Quanta says expansion is on track; total capacity will double versus end-2025 by year-end.
This means → the buildout is no longer a plan on paper — money is already deployed, and the capacity ramp can be measured quarter by quarter.
02

Why spend NT$19.7 billion on AUO's old factory?

The board approved acquiring AUO's Hwa Ya factory complex in Taoyuan for NT$19.7 billion; contract details are still being negotiated.
Two key advantages: the site sits next to Quanta's R&D headquarters and existing server plants, shortening the design-to-production path; and the buildings were originally built for panel manufacturing, so they come with a high power-capacity quota — critical for high-end server lines.
In plain terms = high-end AI servers are extremely power-hungry. A greenfield site would need a lengthy power-allocation application; buying a factory that was "born with ample power" trades money for time.
Quanta stated explicitly that existing capacity cannot cover demand already visible through 2028. The site handover is expected in Q4 2027; once integrated, Taiwan back-end AI server assembly capacity will double again.
03

US$1 billion in overseas injections — which two bets?

The board also approved injecting roughly US$1 billion into two wholly owned overseas subsidiaries, to be disbursed in phases.
US$973 million goes to U.S. subsidiary QMN for additional AI server assembly capacity in California; ~US$22.5 million goes to German subsidiary QCG to expand the autonomous-driving vehicle product line.
This reflects a clear capital-allocation pecking order: AI servers take the overwhelming share; automotive is a second curve, but the spending gap is stark.
04

What is happening to the notebook business?

First-half shipments beat expectations for two consecutive quarters, but Quanta itself flagged the reason: customers pulled orders forward to dodge rising costs, effectively borrowing from second-half demand.
Q3 notebook shipments are expected to fall more than 20% quarter-on-quarter; full-year volumes will see a double-digit decline versus 2025.
Chromebooks are worse off — component price increases are slowing, but supply conditions have turned unfavorable, forcing brands to raise retail prices. Q2 Chromebook shipments were less than 10% of total notebook volumes.
05

Can margins keep up?

Quanta expects the margin uplift from a richer AI product mix in the second half to be offset by the loss of scale from a shrinking notebook business; full-year margins are projected to be flat versus the first half.
The company flagged a specific volatility risk: high-end AI servers carry high unit prices but concentrated shipment volumes, so delivery timing and cost-recognition shifts can swing any single quarter's results.
In plain terms = one AI server can match the revenue of hundreds of notebooks, but if a large order's delivery slides by a week or two, a given quarter's P&L can look dramatically better or worse — which is why Quanta urges investors to evaluate performance over longer periods using absolute profit growth.
06

Is 2028 the make-or-break checkpoint?

Quanta emphasized that all spending is underpinned by confirmed customer demand and is "neither preemptive positioning nor speculative investment."
The other side of the coin: whether Taiwan capacity actually doubles on schedule by 2028 will determine if this NT$40 billion-plus capex cycle converts into real earnings growth.
This means → the key metric for the next two years is not quarterly margin fluctuation but capacity ramp progress and actual order-conversion rates.

Content is for reference only, not financial advice.