كيف تحول رهان شركة صينية على الضوء منذ 10 سنوات إلى أكبر فرصة في البنية التحتية للذكاء الاصطناعي

For most of the last decade, betting the future of computing on light beams instead of copper wires looked like a niche academic obsession. Now, with AI data centers buckling under the physical limits of electrical signaling, that bet has become a multi-hundred-billion-dollar reality. At the center of it is CAS Star, a Chinese venture capital firm that spent ten years quietly funding the photonics revolution before the rest of the world caught on.

Why Photonics Suddenly Matters to AI

The explosive growth of generative AI has exposed a fundamental bottleneck inside data centers: the I/O wall. Modern AI training clusters connect tens of thousands of GPUs that must exchange enormous volumes of data at high speed. Copper interconnects—the traditional wiring between chips—suffer from escalating signal loss, heat, and power consumption as data rates climb, creating a hard physical ceiling on scalability .

The solution is to replace electrons with photons. Photonic links use light to move data, offering far higher bandwidth, lower power per bit, and dramatically better scalability across massive GPU clusters . This is not a theoretical upgrade; it has become the central infrastructure priority for every major hyperscaler.

• The five largest tech companies—Amazon, Google, Microsoft, Meta, and Oracle—have collectively committed an estimated $660–$690 billion to data-center infrastructure in 2026, roughly double the prior year’s spend, with a large portion flowing into optical interconnects and silicon photonics .
• Nomura has characterized the 1.6T optical module market as an “unprecedented supercycle,” with shipments expected to surge from 2.5 million units in 2025 to 20 million in 2026 as light officially dethrones copper for chip-to-chip communication .
• Google Trends data shows that the search term “photonics” began spiking sharply in summer 2025, driven by this sudden infrastructure pivot rather than any long-haul telecom cycle .

Stock movements underscore the scale of the revaluation. Applied Optoelectronics (AAOI) surged more than 1,100% over the past year, fueled by back-to-back hyperscale orders for 800G and 1.6T optical transceivers . The market is no longer rewarding generic photonics exposure; it is rewarding direct exposure to the AI data-center buildout .

Mi Lei’s Contrarian Conviction

For Mi Lei, the photonics boom is not a surprise but a delayed validation of a thesis he has been funding since before most investors knew the term “I/O wall.”

Mi Lei holds a doctorate in optics from the Xi’an Institute of Optics and Precision Mechanics under the Chinese Academy of Sciences (CAS) and joined the institute in 2001, where he worked on commercializing self-focusing lenses . Around 2010, he formed the conviction that optics would follow the trajectory of integrated circuits—shifting from discrete components to large-scale photonic integration—and began building an investment strategy around silicon photonics .

That conviction led to the founding of CAS Star in 2013 as a “hard-tech” venture capital firm incubated within CAS. Today the firm manages approximately ¥18 billion ($2.5 billion) in assets and has backed over 600 portfolio companies . More than 200 of those companies sit in the broader photonics sector, spanning sensing, communications, computing, storage, and display—precisely the applications now being revalued by AI infrastructure spending .

Mi Lei describes his firm’s core mission as “pushing scientific research towards commercialization” to achieve industrial upgrading through new technologies . For most of the past decade, that mission looked like a patient, deeply science-led contrarian play. Today it reads like a roadmap that global capital markets have only just begun to follow.

From Hidden Portfolio to IPO Windfall

CAS Star’s photonics-heavy portfolio has shifted into value-realization mode at striking speed.

• In March 2026, the firm closed the first tranche of its “Pioneer” fund at ¥2.617 billion ($360 million), allocating 70% of capital to early-stage hard-tech projects, with artificial intelligence positioned as a key theme intersecting photonics .
• In May 2026, CAS Star achieved three initial public offerings within 34 days, triggering what Chinese media described as the value-realization phase for its “light + AI” pipeline .
• One of its portfolio companies, Yuanjie Technology, saw its share price increase elevenfold over the preceding year, mirroring the broader photonics supply-chain rally .

The lesson is not that photonics suddenly became important. It is that the physical limits of copper made photonics indispensable, and the firms that had spent a decade building in that direction—before the market priced it in—are now reaping the returns.

What This Means for AI Infrastructure

The photonics boom is not a short-term component cycle. LightCounting projects that AI-related optical connectivity demand will drive market expansion through 2030, as the industry moves from 800G modules to 1.6T and beyond . Silicon photonics (SiPh) penetration is projected to hit between 35% and 50% inside data-center networks by the end of 2026, up from a single-digit share just two years earlier .

In other words, the copper-to-optics transition is just beginning. The firms that understood the physics early—and had the institutional patience to fund it for a decade—are no longer waiting for validation. They are setting the pace.