AlphaLens
Research
研报Overweight5月5日 · Morgan Stanley

Aspeed Technology: Riding the AI GPU and CPU Wave to Revenue Growth; OW

中文EN⚠ quality lint: see notes

Aspeed Technology: Riding the AI GPU and CPU Wave to Revenue Growth; OW

Core Conclusion

Aspeed Technology’s Q1 2026 earnings beat, margin expansion, and accelerating monthly revenue confirm that demand from global CSPs for AI and general server solutions is outstripping supply, with visibility extending through 2027. The stock’s current P/E of ~58x the 2027 EPS estimate only partially reflects the revenue upside from TPU forecast upgrades and broadening ASIC/GPGPU adoption. An Overweight rating is maintained with a NT$20,000 price target, offering ~12% upside from NT$17,790.

What the Market Likely Underestimates

The market underweights Aspeed’s revenue growth runway driven by two parallel forces: rising TPU-related demand and the migration of AI workloads from general-purpose GPU servers to custom ASIC/GPGPU solutions.
Evidence: The report explicitly notes “the continued increase in TPU forecasts also helps drive Aspeed’s revenue in 2027,” and Aspeed is identified as a key beneficiary across CPU servers, ASIC, and GPGPU. This dual exposure creates a longer and more resilient growth trajectory than what a single-server-cycle thesis would imply.
Investment implication: If TPU upgrades continue and ASIC adoption accelerates, consensus 2027 revenue estimates are likely to be revised upward, potentially justifying a valuation multiple above the current 58x.

Evidence Chain

Earnings beat confirms operational momentum.
Q1 2026 EPS of NT$37.42 (+18% Q/Q, +60% Y/Y) exceeded consensus by 14% and the sell-side estimate by 8%, driven primarily by stronger non-operating gains.
Gross margin expanded 0.7ppt Q/Q to 69.2%, exceeding expectations by 4.9ppt vs. the sell-side and 1.3ppt vs. consensus, attributed to a favorable product mix with higher AI content.
Implication: Operating leverage is improving even as revenue scales, implying that margin expansion could persist as AI-related product mix shifts upward.

Revenue momentum is accelerating.
April 2026 revenue of NT$1,279mn (+4% M/M, +82% Y/Y) beat the sell-side estimate of NT$1,269mn. The analyst models a 19% Q/Q sales increase for Q2 2026, described as “quite achievable.”
Implication: Sequential revenue growth of ~19% in Q2 would confirm that demand is not decelerating, supporting the view that the ongoing revenue ramp is durable.

Demand is supply-constrained through 2027.
The report states “demand surpassing supply and likely to continue into 2027,” with Aspeed positioned across CPU servers, ASIC, and GPGPU. This is not a cyclical spike but a structural shortage driven by CSP capex.
Implication: Any supply-side improvement (capacity additions, yield gains) would directly translate into incremental revenue, creating potential upside to current forecasts.

Key Risks and Divergences

Upside risks

  • Stronger cloud demand from hyperscalers could accelerate orders.
  • Faster-than-expected spec migration would expand content per server.
  • Mild competition allows Aspeed to retain market share and pricing power.

Downside risks

  • A softening in cloud capex or general server demand would directly pressure revenue.
  • Slower spec migration could delay the AI content uplift.
  • Intensified competition (e.g., from new BMC entrants) could compress margins.
  • Further policy tightening in China could restrict end-market access or supply chain.

Investment implication: The most likely risk to the thesis is a demand slowdown, but the current supply-constrained environment and long-term AI buildout make a sharp near-term deceleration unlikely. The downside scenario would require a macro-driven cut in CSP spending, which is not reflected in current guidance.

Valuation or Trading Implication

At 58x the 2027 EPS estimate of NT$307.49, the stock already prices in strong growth, but consistent earnings beats and an expanding margin profile suggest the multiple could be supported or even re-rate higher as 2027 visibility improves.
The residual income model used for the NT$20,000 price target assumes a 9.8% cost of equity, 21.4% medium-term growth, and 5.5% terminal growth. At the current price of NT$17,790, the implied upside is ~12%, which is modest but reasonable given the low-P/E multiple relative to historic tech AI plays.
Investment implication: If revenue estimates continue to rise (e.g., from TPU forecast upgrades), the forward P/E will compress, making the stock appear cheaper and potentially catalyzing further multiple expansion. The Overweight rating is therefore justified on a risk-reward basis, with the key catalyst being positive revenue revisions over the next 12 months.

Appendix Data Summary

Table 1: Key Financials (ModelWare estimates)

Metric2025A2026E2027E2028E
Revenue (NT$mn)9,08516,12924,05530,060
EPS (NT$)103.94202.73307.49393.53
EBITDA (NT$mn)4,9849,24214,10018,091
P/E (x)69.887.857.945.2
EV/EBITDA (x)53.471.346.435.9

Table 2: Valuation Methodology and Risk Summary (Residual Income Model)

AssumptionValue
Cost of equity9.8% (2.0% Rf, 6% ERP, 1.3 beta)
Medium-term growth21.4%
Terminal growth5.5%
Cash payout ratio85%
Upside catalystsStronger cloud demand, faster spec migration, mild competition
Downside risksSoftening cloud demand, slower spec migration, intensified competition, China policy tightening

Related (同 ticker)