Three Major Investment Banks Turn Bullish on Zhongji Innolight on the Same Day, but Target Prices Differ by 100%
nashnova research
Citi, Goldman Sachs and Morgan Stanley all rated Innolight (中际旭创) buy/overweight on September 7, yet A-share targets range from ¥1,205 to ¥2,645 — the highest more than double the lowest. The three banks are not buying the same story.
All three are bullish — what do they actually agree on?
The shared thesis: Innolight has evolved from a traditional optical-module supplier into an AI optical-interconnect platform leader — no longer just a component vendor.
All three confirm the product roadmap along 800G → 1.6T → 3.2T, with AI cluster networking expanding across scale-out (more racks), scale-up (more compute per rack), and scale-across (inter-cluster links).
This means → There is no disagreement on the direction — only on how far and how fast this road runs.
Why does Goldman set ¥2,645 — nearly double Morgan Stanley?
Goldman's A-share target is ¥2,645, implying a 2027E P/E of roughly 35.6× — the most aggressive of the three.
The bull case rests on two pillars: silicon-photonics penetration (share of shipments rising from ~50% in 2025 to 86% by 2028), and network demand broadening from pure scale-out into multi-dimensional expansion.
Goldman is the most sanguine on CPO — co-packaged optics, embedding optical modules directly into switch chips. Even if CPO reaches 30% penetration in scale-out, Goldman argues the pluggable-module addressable market could still expand roughly tenfold as rack counts and specs rise.
In plain terms = Goldman is pricing "sustained high growth recognized by the market" — betting the pie itself is still growing fast enough to outrun any substitution threat.
Citi is the most optimistic on volume — so why only ¥1,325?
Citi's A-share target is ¥1,325, based on a 2027E P/E of just 16.2× — one standard deviation below the company's three-year average multiple.
On volume, Citi leads the pack: 2028E shipments of ~96.4 million units, revenue of ¥456 billion, and net profit of ¥167 billion — all the highest among the three.
Yet the valuation carries a steep discount, driven by two risks: roughly 90% of revenue comes from outside China, and the top five customers account for ~76% of sales. High internationalization and high customer concentration are both a scale advantage and a geopolitical-risk amplifier.
This means → Citi is pricing "volume expansion delivered, but with a valuation safety cushion" — the most generous on numbers, the most cautious on price.
Why is Morgan Stanley the most conservative?
Morgan Stanley's A-share target is ¥1,205, the lowest of the three. It uses a residual-income model weighted across three scenarios: bull ~¥1,609 (30%), base ~¥1,201 (50%), bear ~¥609 (20%).
Its 2028E shipment forecast of ~49.5 million units is roughly half of Citi's — the most cautious on long-term volume ramp.
Morgan Stanley treats CPO as a medium-to-long-term variable that must be priced in: large-scale CPO adoption is more likely after 2027–2028, gaining stronger economics at 3.2T and above. It also highlights NPO — near-package optics, placing modules close to the chip without embedding them — as a potential "defensive growth" path against CPO risk.
In plain terms = Morgan Stanley is pricing "growth is real, but the long-term endgame still needs continuous proof" — not disbelieving, just demanding evidence along the way.
With a 2× price gap, what should investors watch?
Four axes of disagreement: long-term shipment scale (2028E from 49.5M to 96.4M units), earnings delivery speed (2026E net profit from ¥34.3B to ¥40.4B), how CPO disrupts (optimistic vs. scenario-discounted), and what multiple to pay for growth (P/E from ~16× to ~40×).
This means → The 2× target-price gap is fundamentally a divergence in how much to discount growth durability and technology-route risk.
The key checkpoints ahead: whether quarterly shipment data tracks the high-end forecasts from Goldman and Citi, and the actual pace of CPO penetration at the 3.2T stage — these two data points will determine which pricing framework lands closest to reality.
市场有风险，内容仅供研究参考，不构成投资建议。