One of the most significant trends shaping the global capital markets in 2026 is quietly unfolding deep within the optical communications supply chain. As AI training clusters scale from tens of thousands to hundreds of thousands, and even millions of GPUs, the main bottleneck limiting computational power is shifting—from GPU supply shortages to the efficiency of data transmission between chips. Against this backdrop, US optical component giant Coherent (NYSE: COHR) is attracting sustained capital attention in the secondary market thanks to its irreplaceable technological positioning. Meanwhile, NVIDIA is backing the commercialization of photonic interconnects with substantial strategic investments, signaling strong confidence in this direction.
For investors focused on AI infrastructure and the global computational power supply chain, understanding the industrial logic behind this wave of optical interconnect upgrades—and how to efficiently allocate US equity assets within a crypto account framework—has become a topic of enduring value.
Why Data Transmission Is Overtaking Compute Power as the Core Bottleneck for AI Clusters
As the parameter count of AI models grows exponentially, the bottleneck for modern AI clusters is undergoing a fundamental shift. Greyhound Research Chief Analyst Sanchit Vir Gogia explains, "Every accelerator in an AI cluster relies on dozens of high-speed links to communicate with its neighbors. Multiply this across the entire rack, and you end up with thousands of interconnected links running continuously. Each link consumes power, introduces latency and signal integrity challenges, and carries a risk of failure." As clusters expand from thousands to tens of thousands of GPUs and beyond, chip-to-chip connectivity consumes an increasingly large share of total system power, becoming the main constraint on overall performance.
The industry has reached a consensus on "electrical compute, optical transmission"—using optics for data transfer between compute units. Co-Packaged Optics (CPO) technology has emerged as a leading solution: it integrates high-speed optical engines and switch chips or AI compute chips on the same substrate via advanced packaging, confining high-speed electrical signal transmission to millimeter-scale distances while relying on fiber optics for medium- and long-range transmission. Compared to traditional pluggable optical modules, CPO can reduce power consumption by over 40%, triple bandwidth, and cut latency by 50%.
According to LightCounting, global shipments of high-speed optical modules (400G and above) will exceed 60 million units in 2025 and are expected to approach 100 million units in 2026. These figures are direct evidence that demand is moving from theoretical forecasts to large-scale orders.
NVIDIA’s Multi-Billion Dollar Bet: A Systematic "Optical Upgrade" Strategy
In spring 2026, NVIDIA’s investment pace in optical interconnects drew widespread market attention. On March 2, NVIDIA simultaneously announced $2 billion investments in US optical technology companies Lumentum and Coherent, totaling $4 billion. The Coherent agreement includes multi-billion dollar purchase commitments and future access to advanced laser component production capacity.
Just three weeks later, NVIDIA invested another $2 billion in Marvell Technology to jointly develop silicon photonics and custom XPU solutions. In early May, NVIDIA announced a $500 million investment in fiber optics giant Corning, securing priority access to its advanced optical connectivity capacity. NVIDIA also participated in Ayar Labs’ $500 million Series E funding round.
In three months, five companies, and at least $6.5 billion invested—NVIDIA has built a highly systematic supply chain network: Coherent and Lumentum cover the optical device layer, Marvell covers the optical interconnect architecture layer, Ayar Labs covers the optical I/O chip layer, forming a complete chain from photonic devices to system interconnects.
It’s worth noting NVIDIA’s investment logic: opting for strategic investments rather than outright acquisitions. This approach locks in future production capacity for key components without disrupting supplier independence, while simultaneously factoring in geopolitical risks—by supporting US domestic manufacturing expansion, "reducing geopolitical risk and aligning with domestic industrial priorities."
Coherent: From Supply Chain Node to AI Compute "Valve"
Coherent has not only recently come under the spotlight in capital markets. Even before NVIDIA’s investment, the company had already delivered better-than-expected results driven by strong demand for high-speed optical modules in AI data centers. In Q1 of fiscal 2026 (ending September 30, 2025), Coherent reported $1.58 billion in revenue, up 17% year over year, with a non-GAAP gross margin of 38.7% and an operating margin rising to 19.5%. Data center and communications revenue grew 26% year over year, with accelerated adoption of 800G and 1.6T transceivers as the main driver. Notably, order volumes hit "record levels," and some customer forecasts now extend through 2028.
The company expects Q4 fiscal 2026 revenue to range between $1.91 billion and $2.05 billion, with non-GAAP gross margins projected at 39% to 41%. Raised guidance further validates the positive outlook for optical interconnect demand.
Coherent currently holds about 20%–30% market share in the transceiver segment. Bank of America Securities predicts it stands to benefit most from shipment growth in 800G and 1.6T transceivers, raising its target price to $400. Since the start of 2026, Coherent’s stock has surged 96.62%, with a market cap around $70.7 billion. Looking further ahead, Bank of America forecasts the total addressable market for AI data centers could reach $1.7 trillion by 2030; Bernstein estimates the global AI connectivity market will grow from $14 billion in 2025 to $73 billion by 2030, with a compound annual growth rate of 39%—the highest among all segments of the AI supply chain.
Accelerating Market Structure Upgrades: Coherent’s CPO Engines Enter Mass Production
On the technical front, Coherent is executing a clear roadmap for mass production of CPO and high-speed transceivers. The company has shipped its first batch of transceivers from its 6-inch indium phosphide wafer fab, with initial yields already surpassing those of its mature 3-inch production lines. Its optical switch platform has been delivered to seven customers, with revenue expected to accelerate in 2026.
Globally, upstream supply chain upgrades are happening in tandem. On June 2, 2026, NVIDIA announced that its Spectrum-X Ethernet silicon photonic switch entered full-scale production, delivering a 5x improvement in energy efficiency, a 5x boost in AI uptime, and over 30% increase in deployment efficiency. Credo completed its acquisition of Israeli silicon photonics startup DustPhotonics, with optical business revenue expected to surpass $500 million in fiscal 2027. TSMC achieved a breakthrough in its 3nm process, reducing optical signal loss to below 0.1dB, paving the way for large-scale commercial adoption of CPO.
These cross-supply-chain signals point to a clear conclusion: CPO is no longer a distant laboratory concept. It is now transitioning from technical validation to large-scale deployment.
Gate Stock Trading: Allocate COHR and Other AI Assets with USDT on a Crypto Platform
For investors tracking this industry trend, allocating to COHR and related assets is becoming increasingly convenient. On June 1, 2026, Gate officially launched real stock trading services, allowing users to trade US-listed stocks and ETFs directly on the platform using USDT. As of June 3, 2026, Gate supports over 10,000 stocks and ETF assets, covering the NYSE, Nasdaq, NYSE Arca, and other major US securities exchanges and liquidity networks.
Gate’s stock trading differs fundamentally from tokenized stocks or RWA-mapped products on other platforms: users are buying real underlying assets traded in sync with Nasdaq and NYSE, holding genuine ownership certificates—not derivatives or on-chain representations. Stocks are custodied by Alpaca, a compliant broker-dealer licensed and qualified for clearing in the US. The partner broker is also a member of the Securities Investor Protection Corporation, providing asset protection under applicable conditions. Trading uses an independent account system, with funds managed separately; stock account assets are not subject to forced transfer due to contract liquidation or spot trading losses. In terms of holding costs, Gate spot stocks do not incur perpetual contract funding rates or CFD swap/overnight fees, resulting in zero holding costs.
Operationally, investors simply need to: ① update the Gate App to the latest version (iOS users must upgrade to version 8.21.5); ② enter the TradFi stock section; ③ transfer USDT and then view and trade COHR and other stocks.
It’s important to note that Gate stocks and tokenized stocks on the platform offer complementary asset access paths: tokenized stocks support 24/7 trading, ideal for flexible, short-term exposure; spot stocks provide compliant holdings and long-term allocation. Both operate seamlessly within the same account framework.
Conclusion
From a technology perspective, the transformation of AI data centers to optical interconnects is not a short-term narrative—it’s a long-term path confirmed by both industry capital and supply chain players. NVIDIA’s $6.5 billion bet on photonic technology in just three months, and Coherent’s dual validation through its core position in high-speed transceivers and CPO, underscore persistent supply-demand gaps and robust market growth expectations—from $14 billion in 2025 to $73 billion in 2030—offering clear long-term growth prospects for the entire supply chain.
For investors focused on AI infrastructure, understanding the industrial logic behind the shift "from electrical to optical" helps build a forward-looking investment framework amid the long-term narrative of explosive AI compute growth. With Gate’s convenient stock trading tools, allocating USDT from crypto accounts directly to COHR and other core AI photonic interconnect assets is becoming a practical option. This product model also reflects a broader trend in 2026—the systematic blurring of boundaries between crypto assets and traditional financial markets.
