Auras Technology
Designs and manufactures thermal modules (heat sinks, heat pipes/vapor chambers, fans) and, increasingly, liquid-cooling components (cold plates, manifolds, CDU parts, quick disconnects) sold to Nvidia, server ODMs/OEMs, and GPU/AIB partners; component + module supplier margins, scaling with AI-server unit growth.
Earnings, margins, COGS & capex
Revenue grew ~47.5% in 2025 to NT$23.28B as liquid cooling jumped from ~12% of sales in 2024 to >40% in 2025; net profit NT$2.70B (+39.7%), EPS NT$28.26. Momentum accelerated into Q1 2026 (revenue NT$8.55B, +93.7% YoY / +12.6% QoQ; net profit NT$1.16B, EPS NT$12.61, +127.7% YoY off a low base), and management raised the FY2026 revenue-growth target from 50% to 70%. Gross margin ~27-30% reflects a component/module mix; the mix shift toward higher-value liquid-cooling assemblies is the swing factor for margins vs. the capex/pricing drag.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~73¢ is cost of goods and ~13¢ operating expense, leaving ~14¢ of operating profit (~12¢ net).
Revenue trend
Margins
improving with liquid-cooling mix
up with scale
improving with operating leverage
COGS structure
Dominated by metals (copper, aluminum), heat pipes/vapor chambers, fans/pumps, and increasingly precision cold-plate and manifold machining; commodity-metal and precision-machining costs are the main COGS drivers, partly offset by richer liquid-cooling content per rack.
Capex
Rising — capacity build-out for cold plates, manifolds and CDU-related parts to serve GB200/GB300 volume; a gating factor for near-term FCF. Exact capex not disclosed at 7 July 2026.
Latest earnings
Strong sequential and YoY growth; management raised FY2026 revenue-growth guide from 50% to 70% — a positive revision rather than a formal beat/miss vs. a single consensus
FY2026 revenue growth target ~70%; liquid-cooling demand seen sustaining growth in thermal management through ~2028
- Liquid cooling % of revenue
- >40% (2025) vs. ~12% (2024)
- FY2025 net profit
- NT$2.70B (+39.7% YoY), EPS NT$28.26
- Q1 2026 net profit
- NT$1.16B (EPS NT$12.61, +127.7% YoY)
- FY2025 dividend / share
- ~NT$11.90 (payout ~42%)
Growth drivers
- AI-server liquid cooling: penetration rising sharply as GB200/GB300 NVL72 racks exceed air-cooling limits
- Content-per-rack expansion — cold plates + manifolds + quick disconnects raise dollar content vs. legacy air modules
- Subsidiary Fositek in universal quick disconnects (UQD), in mass production for Nvidia GB200/GB300
- GPU/AIB win: cold plates and manifolds on Nvidia's recommended supplier list for RTX 50-series (Blackwell)
- Geographic diversification of manufacturing to serve US/global hyperscaler demand
Bull & bear
A designed-in, capacity-constrained supplier of the exact liquid-cooling parts that gate AI-server shipments, growing 70%+ with rising margins and rising content per rack.
- Liquid cooling went from ~12% to >40% of revenue in one year and is still early in penetration
- Management raised FY2026 growth to ~70% and sees the cycle running through ~2028
- Qualified on Nvidia's recommended list for GB200/GB300 and RTX 50-series — a durable design-in
- Fositek UQD ramp adds a scarce, high-value component into the same Nvidia platforms
- Content-per-rack expansion (cold plate → manifold → CDU → quick disconnect) grows dollars faster than units, with Q1 2026 gross margin expanding to ~29.7%
A commodity-margin component maker levered to one hyperscaler capex cycle, priced at ~29x on cycle-elevated earnings, facing bigger rivals and multi-sourcing.
- ~29x TTM P/E prices in sustained hypergrowth; any AI-capex digestion de-rates the multiple and the estimates together
- ~27-30% gross margin is metal-and-machining economics — Nvidia multi-sourcing every part caps pricing power
- AVC has greater scale and Cooler Master led first-batch GB200 rack share (50%+ vs AVC ~30-40%); Auras fights for share, not owns it
- Capex-heavy ramp depresses free cash flow; a demand miss leaves stranded capacity
- Concentration and FX/commodity exposure make earnings more cyclical than the growth rate suggests
What it is worth
P/E and reverse-DCF sanity check vs. Taiwanese thermal peers (AVC 3017 TT, Jentech 4968 TT) and US data-center cooling comps (Vertiv VRT). Market cap ~US$3.0B (NT$94B) on NT$1,010 at ~29x TTM EPS (TTM net ~NT$3.2B) and ~20x forward FY2026 if the ~70% growth guide holds.
AI-capex digestion + margin compression from multi-sourcing cuts growth to sub-20% and de-rates the multiple to ~12-15x → meaningful downside as estimate cuts and multiple compression compound.
Growth decelerates from ~70% toward 30-40% as comps harden and pricing normalizes; ~20-25x forward → roughly range-bound to modestly higher, tracking earnings.
Sustained ~60-70% growth into 2027 with content-per-rack expansion and Fositek UQD ramp holds a ~29x multiple → material upside from NT$1,010; the liquid-cooling cycle runs through Rubin (2028+).
~29x trailing / ~20x forward is a growth multiple that implicitly requires sustained 40%+ revenue growth and stable-to-rising ~27-30% gross margin. That is defensible while AI-server liquid-cooling penetration climbs, but leaves little cushion if hyperscaler capex digests or Nvidia multi-sourcing compresses price. The multiple is the thesis: it is priced for continued share-of-wallet gains, not just unit growth.
SWOT
Strengths
- Established #2 Taiwanese thermal-module maker with a designed-in position across notebooks, GPUs, and now AI servers
- Full-stack thermal breadth: air + liquid, plus cold plates, manifolds and (via Fositek) quick disconnects
- Qualified into the Nvidia GB200/GB300 and RTX 50-series supply chain — high switching cost once designed in
- Rapid, profitable growth: revenue +47.5% in 2025 with double-digit operating margin, accelerating in Q1 2026
Weaknesses
- Concentrated demand exposure to the Nvidia AI-accelerator cycle and a handful of server ODMs
- Gross margin ~27-30% is component-grade, not software-like; vulnerable to metal-cost and pricing pressure
- Heavy capex to add capacity compresses near-term free cash flow
- Smaller scale than AVC and less first-batch GB200-rack share than leader Cooler Master (50%+) and AVC (~30-40%)
Opportunities
- Liquid-cooling penetration still early — headroom as air cooling hits thermal limits at Rubin/Rubin Ultra
- Higher dollar content per rack as cooling migrates from cold plate to full loop (manifolds, CDUs, UQDs)
- US/overseas manufacturing footprint to win hyperscaler and OEM programs seeking supply-chain diversification
- Adjacent markets: automotive/EV thermal and energy-storage cooling
Threats
- AI-capex air-pocket: any hyperscaler digestion pause hits order flow directly
- Intense competition from AVC, Cooler Master, Jentech, Delta and Boyd on cold plates/manifolds compressing price
- Nvidia dual/triple-sourcing every cooling component caps share and pricing power
- Commodity-metal inflation and FX (NT$/US$) swings on an export model
Moats, dependencies & bottlenecks
Moats
Being on Nvidia's recommended supplier list and qualified into GB200/GB300 raises switching costs for the design cycle, but Nvidia deliberately multi-sources.
Full-stack thermal capability plus Fositek quick disconnects lets it bid whole-loop content, not just one part.
Capacity ahead of demand is a real advantage in a supply-tight market, but smaller than AVC.
Microchannel cold-plate (MCCP) R&D shared across the top Taiwanese players — a lead, not a lock.
Dependencies
GB200/GB300 and RTX 50-series drive the growth; Nvidia sets the cooling spec and the supplier list.
Quanta, Wiwynn, Supermicro, Dell, HPE) Modules ship through rack integrators; program wins gate volume.
Commodity inflation pressures gross margin.
Microsoft/Meta/Amazon/Google capex is the ultimate demand source; a pause hits orders fast.
Export revenue vs. NT$ reporting; currency swings move reported margin.
Advantages
- Early, profitable pivot into liquid cooling (>40% of revenue) ahead of most peers
- Whole-loop content (cold plate + manifold + UQD via Fositek) rather than a single part
- Qualified across both AI-server (GB200/GB300) and consumer GPU (RTX 50) platforms
- Capacity expansion positioned for the Rubin/Rubin Ultra transition through ~2028
Weaknesses
- Component-grade gross margins (~27-30%) with limited pricing power under multi-sourcing
- Sub-scale vs. AVC and behind Cooler Master in first-batch GB200 rack share
- Free cash flow pressured by the capex ramp
- Revenue concentration on one accelerator cycle and a few ODM customers
Bottlenecks
- Capacity for cold plates, manifolds and quick disconnects — supply reported tight, gating revenue conversion
- Precision-machining throughput and yield on microchannel cold plates
- Nvidia qualification slots per component (multi-sourcing limits share)
- Skilled process engineering to scale liquid-cooling assembly without margin loss
Top signals & trends
Top signals
12% (2024) → >40% (2025); the mix shift is the core thesis.
Management confidence in order visibility.
Scarce high-value component entering volume.
Caps share and pricing power per component.
The swing risk — any digestion pause de-rates the group.
Trends
GB200/GB300 NVL72 thermal density forces direct-to-chip liquid cooling — Auras' core tailwind.
Manifolds, CDUs and quick disconnects expand the addressable BOM.
Hyperscalers/OEMs favor diversified thermal suppliers.
Even higher power/heat extends the cooling upgrade cycle.
Efficiency race among AVC, Auras, Cooler Master.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Universal quick disconnects (UQD) in mass production for GB200/GB300 — captive high-value component supply.
Copper/aluminum & precision-machining vendors Raw metals and machined parts for cold plates, heat pipes, vapor chambers.
Active-cooling subcomponents for modules and loops.
Sets the cooling spec; Auras cold plates/manifolds on recommended supplier list for GB200/GB300 and RTX 50-series.
AI-server OEM integrating liquid-cooled NVL72 systems.
Enterprise/AI-server OEM in the GB200/GB300 ecosystem.
AI/HPC server OEM adopting direct-liquid-cooled racks.
Taiwanese rack ODMs (2317 TT / 2382 TT / 6669 TT) that integrate Auras cooling into Nvidia racks.
Larger Taiwanese thermal peer; ~30-40% share of first-batch GB200 racks and strong in cold plates.
Private; led first-batch GB200 rack cooling with 50%+ share; strong in cold plates and manifolds.
Taiwanese thermal-module maker expanding into server/liquid cooling.
Power + thermal + CDU scale; broader data-center franchise.
US-listed data-center thermal/CDU leader; supplies coolant distribution units, more system-level than component.
Private; validated on Nvidia GB200 NVL72 vendor list for cooling.