Silicon Photonics Concept Stocks Ride the Windfall: The Industry Chain Restructuring and Investment Path Behind AI Computing Power’s “Optics Advancing While Copper Retreats”

Silicon photonics technology is moving from laboratory experiments toward large-scale deployment.
2026 is widely regarded by the industry as the year of commercial transformation for CPO (co-packaged optics),
TSMC announced the start of mass production of its COUPE photonic interconnect platform,
and Nvidia invested over $6.5 billion in optical interconnects within three months.
These signals indicate that: the interconnection bottleneck for AI compute clusters has shifted from the chips themselves
to the data transmission channels between chips.
Traditional copper cable electrical interconnects have reached physical limits in power consumption, bandwidth, and latency,
making it impossible to support the expansion of training clusters with hundreds of thousands of GPUs,
while silicon photonics and CPO technology offer a feasible path for “optical replacing copper.”

The core judgment of this article is: silicon photonics is not just a short-term speculative theme,
but a fundamental reconfiguration of communication architecture driven by AI compute demands.
This reconfiguration is changing the value distribution logic of the optical communication industry chain—
design capabilities and advanced packaging abilities are replacing discrete component assembly,
becoming new high-value areas.
For investors in the crypto market, participating in US stock trading of silicon photonics concept stocks via Gate has become a feasible way
to track this industry trend.

Why has silicon photonics upgraded from an alternative technology to a core bottleneck for AI compute?

The scale of AI large model training compute is evolving from clusters of ten thousand GPUs
to hundred thousand and even million-GPU clusters.
At this level, the physical limits of traditional electrical interconnects are fully exposed:
signal attenuation and latency accumulate to unacceptable levels with each meter of copper cable transmission;
power consumption share rises from less than 10% in thousand-GPU clusters
to over 30% in million-GPU clusters.
An industry consensus is: once cluster size exceeds 50k GPUs,
the marginal cost of electrical interconnects surpasses that of the chips themselves.

CPO technology integrates high-speed optical engines with exchange or AI compute chips
on the same substrate through advanced packaging,
limiting high-speed electrical signals to millimeter-range transmission,
while medium- and long-distance transmission is handled by optical fibers.
Compared to traditional pluggable optical modules, CPO reduces power consumption by over 40%,
triples bandwidth, and shortens latency by 50%.
The fundamental change here is:
optical interconnects are no longer optional “add-on” components,
but become infrastructure at the same level of importance as the compute chips.

At TSMC’s technology forum in May 2026,
its photonic interconnect platform COUPE was highlighted as the most important layer in future AI platform architecture,
and the company announced mass production of the world’s first 200 Gbps micro-ring modulators.
ASE Technology Holding’s CEO Wu Tianyu also clearly stated that replacing some electronic communication with optical communication is an inevitable direction,
and mass production of CPO in 2026 is just a matter of time.
Signals from the industry’s upstream chain indicate that:
silicon photonics is shifting from a technical alternative to an essential component of AI infrastructure.

How is the value distribution in the optical communication industry chain undergoing structural shifts?

In traditional optical module industry chains, value is dispersed among upstream discrete components such as optical chips and electronic chips.
The core competitiveness of optical module manufacturers lies in procurement and integration, not in underlying technology barriers.
Silicon photonics has moved the industry focus upstream,
using semiconductor manufacturing logic to reshape the previously component-assembly-dependent optical module industry chain.

The key impact of this change is:
companies with PIC (photonic integrated circuit) design capabilities and advanced packaging abilities
significantly enhance their industry chain influence.
In the traditional pattern, optical module manufacturers are mainly assemblers of discrete components;
under the silicon photonics route, design and integration capabilities become the high-value assets.
For example, leading companies like InnoLight and Innolux benefit continuously from upgrades to 800 G / 1.6 T generations,
but industry profits are increasingly tilting toward firms with autonomous PIC design capabilities.

A deeper structural shift involves the deep involvement of semiconductor manufacturing and packaging giants.
TSMC leverages its COUPE platform to incorporate photonic integration into its advanced process services;
ASE, KYEC, and other packaging firms are entering the optical engine market through SiP (System-in-Package) technology.
This means that profit centers in the silicon photonics industry are shifting from traditional optical module manufacturers
to those with semiconductor fabrication and packaging capabilities.
This trend aligns with the overall direction of the AI chip market—
whoever masters advanced packaging will hold the pricing power of the next-generation compute infrastructure.

What industry trends are reflected by Nvidia’s $6.5 billion intensive investments?

In spring 2026, Nvidia made a highly strategic series of investments in the optical interconnect field:
in March, investing $2 billion each in Coherent and Lumentum;
three weeks later, investing another $2 billion in Marvell;
in early May, investing $500 million in Corning;
and participating in Ayer Labs’ Series E funding.
The total investment over three months exceeded $6.5 billion, covering the entire technology chain from optical devices and architectures to optical I/O chips.

This concentrated investment is not purely financial speculation, but a supply chain security strategy.
Nvidia’s GPU sales heavily depend on upstream optical interconnect component supply capacity.
According to LightCounting, due to supply constraints of indium phosphide materials,
the supply gap for traditional EML optical chips in 2026 is significant,
and this gap will mainly be filled by silicon photonics solutions.
Nvidia’s capital moves essentially lock in silicon photonics capacity for the next three years,
ensuring its AI server shipments are not limited by interconnect bottlenecks.

From an industry impact perspective, Nvidia’s investments directly alter the market positions of related companies.
After Lumentum released unexpectedly strong earnings in February 2026, its stock surged over 7% in a single day,
and it was subsequently included in the S&P 500 and Nasdaq 100 indices, with optical component capacity sold out through 2028.
Coherent also benefited from the investments, reaching a historic high in stock price.
This phenomenon indicates that in the AI compute supply chain, upstream firms with scarce technologies are gaining unprecedented pricing power and capital premiums.

How do silicon photonics concept stocks in the US and Taiwan differ in their market presentation?

US silicon photonics concept stocks focus on defining standards and system integration.
Marvell, through strategic cooperation with Nvidia, is expected to deeply participate in the NVLink Fusion ecosystem, playing a key role in custom XPU and optical interconnect architectures.
Broadcom’s Tomahawk series CPO switches have become the standard choice for AI data centers.
Credo’s acquisition of Israeli silicon photonics company DustPhotonics for $750 million aims to build a complete tech stack from electrical to silicon photonics.

In Taiwan, silicon photonics concept stocks are mainly “implementers” supporting the supply chain.
In April 2026, driven by Credo’s acquisition, stocks like Lianya, Walsin, Zundar-KY, Shangquan, Guangzhao, Chuangwei, QianDing, and WIMAX all hit the daily limit.
Among them, BoroWai controls passive optical components; Lianya provides laser sources; Shangquan collaborates deeply with TSMC on fiber array connection tech; and Pan-Chuan’s optical path positioning tech can improve silicon photonics packaging yield.
Display panel giants AUO and Innolux are also actively investing in MicroLED as short-distance transmission sources for CPO.

In A-shares, Zhongji Xuchuang and Innolux benefit directly from the 800 G / 1.6 T optical module wave.
Yuanjie Technology, Shijia Photonics, and Changguang Huaxin have domestic substitution capabilities in CW (continuous wave) light sources.
The core logic of the two polarizations is: US stocks focus on technological standards and system integration, while Taiwan and A-shares focus on manufacturing and packaging support.
For investors, these two types of targets have different risk-return profiles—
the former benefits from technological barriers and gross margin advantages,
the latter benefits from capacity expansion and order certainty.

How does Gate stock trading support crypto users’ participation in silicon photonics industry trends?

Gate has officially launched stock trading services, which differ fundamentally from market tokenized stocks or derivatives:
Gate stocks are not assets mapped on-chain or tokenized derivatives,
but are provided through compliance broker-dealer connections, offering users direct stock and ETF trading.
In terms of asset coverage, Gate stocks currently support over 10,000 stocks and ETFs,
covering major US exchanges such as NYSE, NASDAQ, NYSE Arca, NYSE American, and BATS.

For the silicon photonics theme, the US stocks covered include the core targets mentioned above.
Users can directly use USDT within the platform to participate in US stock trading,
with account transfers and trade execution all within the same Gate App account system.

From a business model perspective, Gate stocks operate as spot trading,
without funding rates or overnight holding fees,
making it more suitable for long-term tracking of the silicon photonics industry trend.
The launch of Gate stock trading reflects a typical extension of crypto trading platforms into traditional financial assets,
indicating a trend of platform competition expanding from pure crypto assets to cross-asset categories.

Industry data shows that the global silicon photonics market size was about $2.81 billion in 2025,
expected to reach $3.51 billion in 2026, and potentially surpass $31.9 billion by 2035,
with a compound annual growth rate over 27.5%.
In China, the silicon photonics chip market is projected to reach between $3.27 billion and $50k in 2026,
with data centers and AI accelerators accounting for over 55%.
800 G and 1.6 T optical module shipments will double in 2026,
with silicon photonics solutions accounting for over 50% in 800 G and up to 70-80% in 1.6 T.

The scaling inflection point for the silicon photonics industry is forming.
This trend is driven not by hype but by the physical limits imposed by expanding AI compute clusters.
For crypto investors, participating in US stock trading of silicon photonics concept stocks via Gate has become a direct way to follow this industry trend.
Risks to watch include:
CPO packaging yield not meeting expectations,
competition and diversion from LPO and other tech routes,
and fluctuations in AI investment cycles.
The true value realization of the silicon photonics industry will likely await large-scale mass production data in 2027–2028.

FAQ

What are the main differences between silicon photonics technology and traditional optical module technology?

Silicon photonics deeply integrates optical components with silicon-based semiconductor manufacturing,
using CPO packaging to directly integrate optical engines with AI chips,
whereas traditional optical modules rely on discrete optical components and electronic chips, resulting in larger size, higher power consumption, and lower integration.

Why did Nvidia invest over $6.5 billion in optical interconnects within three months?

Nvidia’s intensive investments aim to lock in silicon photonics capacity for the next three years,
ensuring its AI server shipments are not constrained by interconnect bottlenecks,
and addressing supply gaps in traditional EML optical chips caused by indium phosphide material constraints.

What is the projected market size for silicon photonics in 2026?

The global silicon photonics market is expected to reach about $3.51 billion in 2026,
with China’s market between $3.27 billion and $4.215 billion,
mainly driven by data centers and AI accelerators, which account for over 55%.

Which are the representative stocks of silicon photonics concept in the US and Taiwan markets?

In the US: Lumentum, Coherent, Marvell, Broadcom, Credo.
In Taiwan: Lianya, Walsin, Zundar-KY, Shangquan, Guangzhao, Chuangwei, QianDing, WIMAX.

How does Gate stock trading differ from stock tokenization products?

Gate stocks are not on-chain mapped assets or derivatives; they are provided through compliant broker-dealer connections,
allowing users to directly trade stocks and ETFs using USDT within the platform.

Which silicon photonics concept stocks are supported by Gate?

Core US stocks include those mentioned above, available for direct USDT trading after KYC and eligibility checks,
covering major exchanges like NYSE and NASDAQ.

What are the main risks facing the silicon photonics industry?

Risks include:
CPO packaging yield not meeting expectations,
competition from LPO and other tech routes,
and demand fluctuations due to AI investment cycle volatility.

When is the scaling inflection point for the silicon photonics industry expected?

While 2026 is seen as the year of commercial transformation,
large-scale mass production and performance realization are expected to become evident around 2027–2028,
requiring ongoing monitoring of yield data and order backlogs in the industry chain.