CPO Long-Term Logic Unchanged, but 2027 Shipment Expectations Need Downward Revision

Overview

This research report addresses recent market concerns regarding delays in the introduction of Co-Packaged Optics (CPO). Morgan Stanley believes that while the long-term growth story for CPO remains unchanged, short-term market expectations need recalibration. The firm's model shows that actual CPO optical engine shipments in 2027 will be far below investors' current optimistic expectations, primarily constrained by yield bottlenecks in advanced packaging and assembly processes. The report advises investors to distinguish between short-term sentiment resets and long-term industry inflection points, pointing out that 2028 and beyond marks the starting point for true CPO volume ramp-up, and adjusts allocation recommendations for related supply chain companies accordingly.

Core views

Significant Deviation in Shipment Expectations: The report estimates 2027 CPO optical engine (including Scale-up and Scale-out solutions) shipments at approximately 6-7 million units, whereas current investor consensus is as high as 20-30 million units. This massive expectation gap implies that CPO-related stocks may face sentiment cooling and valuation correction pressures in the short term. Manufacturing Yield is the Core Constraint: Although TSMC plans to expand Photonics Integrated Circuit (PIC) capacity to 10,000 wafers per month in Q1 2027, final output remains constrained by two key yield metrics. First, SoIC packaging yield is currently only 50%-60%, and second, downstream assembly yield is as low as 20%-50%. These two manufacturing bottlenecks jointly determine that actual delivery volumes cannot achieve explosive growth in the short term. Technology Roadmap in Transitional Coexistence Period: The report views 2026 to 2028 as a phase where multiple interconnect technologies coexist. Since mainstream solutions remain at 1.6T/3.2T speeds and CPO mass production ramp-up takes time, traditional pluggable optical modules and copper cable connections will coexist in the market with new optical interconnect solutions like CPO/NPO. The true prosperity period for CPO is expected to begin in 2028, providing a more sustained catalyst path for long-term investment. Clear Stock Differentiation: Based on the unchanged long-term logic, the report maintains an Overweight (OW) rating on key CPO enablers, including TSMC, ASE Technology Holding, HuaStar Optoelectronics, Quanta Computer, Winbond Electronics, Hon Hai Precision, and Winbond Electronics. Among them, HuaStar Optoelectronics' CPO revenue is expected to start ramping up at the end of Q3, with previously flat performance being a normal preparation phase before mass production. In contrast, for A-share listed Suzhou Tfc Optical Communication, while acknowledging its opportunities in the CPO component sector, considering its significant exposure to pluggable transceiver component businesses and a notable valuation premium compared to A-share peers, it is given an Equal-weight (EW) rating.

Analysis framework

The report adopts a bottom-up deduction model of 'Capacity-Yield-Shipment'. The analysis does not stop at upstream foundry expansion plans (such as TSMC's PIC capacity) but further introduces yield parameters from two critical stages—advanced packaging (SoIC) and backend assembly—as adjustment factors, thereby deriving terminal shipment forecasts that are more prudent than market consensus. Simultaneously, the report employs a technological generational evolution analysis framework. By judging the match between current mainstream speeds (1.6T/3.2T) and new technology maturity, it defines the coexistence window for old and new technologies. This method helps investors identify the temporal mismatch between the 'concept introduction phase' and the 'performance explosion phase', avoiding the pitfall of linearly extrapolating long-term visions to short-term performance.

Methodology notes

• Industry/Sector Analysis FrameworkVolume-Price Breakdown

  Effective Output = Theoretical Capacity × Comprehensive Yield

  In the field of semiconductor advanced packaging, nominal capacity does not equal actual supply. The report restores the true deliverable quantity by multiplying foundry expansion data by the low yield coefficients of SoIC and assembly stages. This method reveals why, even with significant upstream capacity expansion, terminal products may still face supply shortages or fail to meet volume expectations.

• Industry/Sector Analysis FrameworkPenetration S-curve

  Multi-solution Coexistence Characteristics During New Technology Introduction Phase

  The report points out that 2026-2028 is a coexistence period for CPO and traditional solutions, which aligns with typical characteristics of the early stage of the technology S-curve: before new technology costs and yields reach a critical point, old technologies do not exit immediately but run parallel with new solutions based on cost-performance ratios in different scenarios. Understanding this pattern helps avoid excessive linear extrapolation of new technology substitution speeds.

• Event Gaming and Behavioral FinanceExpectation Gap/Expectation Management

  Analysis of Deviation Between Institutional Forecasts and Market Consensus

  The report explicitly quantifies the huge gap between its own forecast (6-7 million units) and market consensus (20-30 million units). This expectation gap analysis is a key tool for judging short-term stock price volatility: when real data regresses toward conservative institutional forecasts, high-valuation growth stocks often face severe valuation correction risks.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• TSMC (2330.TW)

  Core CPO Enabler, providing PIC chips and SoIC packaging capacity

  Strengths

  Possesses irreplaceable photonics integration and advanced packaging capabilities, with clear capacity expansion

  Weaknesses

  SoIC yield remains an internal bottleneck constraining the volume ramp-up of its CPO business

  Comparison

  Compared to pure optical device manufacturers, TSMC holds a more upstream core position in the supply chain

  Risks

  Advanced packaging yield improvement slower than expected

• ASE Technology Holding (3711.TW)

  Key CPO OSAT Partner, benefiting from downstream assembly demand

  Strengths

  Solidified leadership in OSAT, directly undertaking CPO assembly orders

  Weaknesses

  Downstream assembly yield fluctuates significantly (20-50%), affecting revenue recognition

  Comparison

  Forms front-end and back-end synergy with TSMC, an indispensable link in the CPO manufacturing process

  Risks

  Margin pressure if assembly yield improvement falls short of expectations

• HuaStar Optoelectronics (3363.TWO)

  CPO Optical Engine Supplier, revenue entering ramp-up phase soon

  Strengths

  CPO revenue expected to start scaling at end of Q3, strategically positioned in core optical engine segment

  Weaknesses

  Flat performance before mass production, lacking short-term earnings support

  Comparison

  More directly benefits from overall optical engine shipments compared to pure component factories

  Risks

  Mass production delays or extended customer validation cycles

• Suzhou Tfc Optical Communication (300394.SZ)

  Potential Beneficiary of CPO Components, but existing business dominated by pluggables

  Strengths

  Good entry opportunity in CPO component sector, solid technical reserves

  Weaknesses

  Pluggable transceiver business still accounts for a high proportion, valuation significantly premium to A-share peers

  Comparison

  Lower business purity compared to pure CPO targets; more expensive valuation compared to peers

  Risks

  Valuation premium requires high earnings growth to digest, CPO transformation speed slower than expected

Key data

• 2027 Optical Engine Shipment Forecast6-7 Million UnitsIncludes Scale-up and Scale-out solutions, significantly lower than market expectations
• Market Consensus Shipment Volume20-30 Million UnitsCurrent general investor expectation for 2027, carrying risk of overestimation
• SoIC Packaging Yield50%-60%Key front-end packaging bottleneck constraining final CPO shipment volumes
• Downstream Assembly Yield20%-50%Major fluctuation factor in backend manufacturing, directly impacting finished product delivery
• TSMC PIC Capacity Target10,000 Wafers/MonthExpected to be achieved in Q1 2027, but still limited by subsequent yields

Impact & implications

For the CPO supply chain, short-term stock price corrections reflect the market's reasonable response to downward revisions in shipment expectations, but this does not change the long-term direction of technological evolution. The report prompts investors to focus on 2028 as the true industry inflection point, rather than overly fixating on short-term fluctuations in 2027. For stock selection, leading companies with core manufacturing capabilities and strategic positioning advantages (such as TSMC and ASE Technology) retain long-term allocation value after sentiment resets; however, for companies whose valuations already fully reflect optimistic expectations and whose business structures are still in transition (such as Tfc Optical Communication), caution is advised, waiting for more certain performance validation or valuation digestion.

Risks

• SoIC packaging yield remains stuck at low levels of 50-60%, leading to insufficient effective output
• Downstream assembly yield fails to break through 50%, severely constraining final delivery volumes
• Delayed CPO technology introduction, pushing the 2028 surge node further back
• Excessively high market expectations for 2027 shipments triggering sharp corrections in related stocks
• Extended lifecycle of traditional 1.6T/3.2T solutions squeezing early CPO market share

What to watch

• Quarterly trend changes in TSMC's SoIC packaging yield
• Specific progress and timeline for yield improvements at downstream assembly plants
• Whether HuaStar Optoelectronics' CPO revenue starts ramping up as scheduled at the end of Q3
• Whether actual 2027 optical engine shipments approach the 6-7 million unit forecast range
• Adjustments in procurement ratios between 1.6T/3.2T and CPO solutions by major cloud providers