Innolight's Hong Kong IPO Receives Dual Approval, Fundraising Scale Raised to Up to $8 Billion
Alina Collins
Innolight Technology (300308.SZ) has secured approval from both the CSRC and the HKEX listing committee for its Hong Kong IPO, with bookbuilding set to begin as early as next week and the fundraising ceiling now raised to $8 billion — potentially Hong Kong's largest listing since Alibaba's $12.9 billion debut in 2019, and a real-time test of whether the market can absorb AI-hardware valuations at this scale.
How big is $8 billion?
The earlier market expectation was around $7 billion. The new ceiling of $8 billion is roughly 14% higher.
This means → if fully subscribed, it would surpass Luxshare's $3.1 billion share placement on July 6 and become Hong Kong's largest IPO since Alibaba's $12.9 billion listing in 2019.
Bloomberg data shows Hong Kong IPOs have raised a cumulative ~$35 billion this year, approaching the full-year 2025 total. In plain terms = this single deal could account for nearly a quarter of Hong Kong's 2026 IPO volume so far.
How fast is the business actually growing?
Revenue tripled in three steps: RMB 10.7 billion in 2023 → RMB 23.9 billion in 2024 → RMB 38.2 billion in 2025, a 60.3% year-on-year increase in 2025.
Q1 2026 was even more dramatic: revenue hit RMB 19.5 billion, up 192% year-on-year; net profit reached RMB 5.74 billion, up roughly 262% — one quarter's earnings already exceeding all of 2024.
Gross margin rose from 36.1% to 45.5% in the same period. This means → Innolight is not just shipping more — it is earning more per module, a clear sign that AI demand is giving it real pricing power.
Where is the technology moat?
According to Frost & Sullivan, Innolight has been the world's largest optical interconnect provider by revenue for five consecutive years since 2021, holding a 21.2% global share in 2025 and 28.1% in high-speed data-communication optical interconnects.
The company was the first to mass-produce 400G (2018), 800G (2020), and 1.6T optical transceivers — the core devices in data centers that convert electrical signals into light, enabling faster data movement across AI computing clusters — leading the nearest competitor by roughly six months each time.
As of Q1 2026, products using silicon photonics — a technology that replaces some electrical wiring with light, boosting speed and cutting power — accounted for about 70% of high-speed product revenue. This reflects a shift from "selling modules" to "selling a technology platform."
Nvidia, Google, and Meta are all customers — what does that tell us?
Innolight's core products are optical transceivers used in AI computing clusters and data centers. Nvidia, Alphabet (Google's parent), and Meta are among its major customers.
In plain terms = the three companies spending the most on AI infrastructure globally are all ordering from the same supplier. That is both a powerful demand endorsement and a reminder that customer concentration is a double-edged sword.
Driven by AI infrastructure demand, Innolight's Shenzhen-listed A-shares surged 454% over the past year, but have pulled back roughly 29% from their June peak. This reflects a market that has already begun digesting valuations at the top.
Three optical-transceiver firms heading to Hong Kong at once — is that too crowded?
Innolight is not alone. Competitor Eoptolink (新易盛, 300502.SZ) has reportedly filed confidentially with the HKEX, targeting $4–5 billion; Eoptolink Technology (易飞扬) is preparing a raise of up to $5 billion.
Combined, the three firms' potential fundraising reaches $17–18 billion. This means → the supply of AI-hardware equity hitting Hong Kong will expand sharply in a short window, and whether the market can absorb all of it at once is a very real question.
Bloomberg notes that the sharp rally in AI-related stocks is already showing cracks, with growing skepticism about whether current valuations can hold.
What will the money be used for?
Proceeds are earmarked for four areas: optical interconnect R&D, global capacity expansion, strategic M&A, and working capital.
The priority is delivering 1.6T and faster products while advancing R&D and mass-production readiness for 3.2T transceivers. In plain terms = the current product is 1.6T; the capital raise is to get 3.2T onto the production line — a bet that AI computing demand will scale up by another order of magnitude.
Goldman Sachs, CICC, Morgan Stanley, and GF Securities are joint sponsors. Whether the final offering size can attract full subscription in the current market mood will be the IPO's defining test.
Content is for reference only, not financial advice.