KraneShares is doubling down on one of the least-discussed bottlenecks in AI infrastructure: getting chips to talk to one another. The firm's flagship AI ETF, AGIX, has added a private stake in Ayar Labs, a co-packaged optics pioneer now weighted at 1.04% of net assets (US$4.875 million) as of July 27, 2026. Days earlier, on July 15, 2026, KraneShares launched a companion fund, the Photonic and Optical ETF (LUMA), a pure-play ETF targeting the optical interconnect supply chain. Together, the moves underscore KraneShares' broader investment thesis: as AI clusters scale, the connections between chips are becoming as important as the chips themselves. Goldman Sachs estimates the AI optical networking opportunity could reach US$154 billion, including approximately US$91 billion in co-packaged optics (CPO).
AGIX Broadens Its AI Infrastructure Portfolio
Rather than focusing solely on AI model developers, AGIX seeks exposure across the broader AI ecosystem, including compute, networking, infrastructure, data services, and increasingly the optical connectivity technologies needed to scale next-generation AI clusters. The actively managed ETF invests at least 80% of its net assets in securities tied to companies represented by the Solactive Etna Artificial General Intelligence Index, while retaining the flexibility to add emerging private companies such as Ayar Labs alongside established public AI leaders.
Unlike traditional AI ETFs that invest only in listed companies, AGIX combines public equities with selected private holdings, giving investors exposure to emerging AI infrastructure businesses before they reach public markets.
Ayar Labs, a Silicon Valley-based semiconductor company, builds co-packaged optics that sit directly beside the chip rather than in a separate pluggable module. Its TeraPHY optical I/O chiplet and SuperNova external light source are designed to cut energy use by roughly 85% versus traditional pluggable modules while enabling up to 10× higher bandwidth at the package edge and up to 100× higher bandwidth within the same footprint. Because Ayar Labs is designed to support multiple communication protocols and semiconductor packaging platforms, it could serve AI chips developed by multiple cloud providers rather than a single proprietary ecosystem, making it one of the few independent optical platforms positioned to work across competing AI chip ecosystems. NVIDIA, AMD, and Intel Capital have all backed the company, and it has potential manufacturing pathways through GlobalFoundries and the TSMC ecosystem (via Alchip and Global Unichip).
AGIX Private Holdings (as of July 27, 2026)
AGIX Fund Snapshot (as of July 27, 2026)
AGIX has meaningfully outpaced its benchmark, the Solactive Etna Artificial General Intelligence Index, which tracks companies with significant exposure to artificial intelligence technologies, returning 26.53% YTD and 40.63% over one year through the same period. The outperformance coincides with the fund's active management approach, which includes exposure to private companies such as Anthropic and Ayar Labs.
What is Photonics? Why it is a Critical Layer of AI Infrastructure
Photonics is the science and technology of generating, transmitting, and processing information using light instead of electrical signals. The field emerged following the invention of the laser in 1960 and advanced rapidly with the adoption of fibre-optic communications, which revolutionised global telecommunications and the internet. Today, the same technology is becoming a critical building block for artificial intelligence, enabling AI systems to move vast amounts of data faster and more efficiently than traditional electrical connections.
As AI models become larger and more complex, the challenge is no longer just building faster chips it is ensuring those chips can communicate with one another efficiently. NVIDIA's NVLink 6, one of the industry's most advanced copper interconnect technologies, delivers exceptional performance within a server but becomes less efficient over longer distances. As AI clusters expand from dozens to thousands of accelerators spread across racks and data centres, traditional copper connections consume more power, generate more heat, and limit data transfer speeds.
Photonics addresses this bottleneck by transmitting data using light through optical interconnects instead of electrical signals. This enables significantly higher bandwidth, lower latency, and improved energy efficiency, making optical networking an increasingly essential component of next-generation AI infrastructure.
For investors, photonics is emerging as one of the fastest-growing themes within AI infrastructure. Goldman Sachs estimates the AI optical networking market could reach US$154 billion, including around US$91 billion in co-packaged optics (CPO), as hyperscalers accelerate spending on next-generation AI data centres. Industry investment is already gathering pace, with companies across the optical networking supply chain announcing more than US$11.3 billion in acquisitions, strategic investments, customer agreements, and prepayments over the past year.
That wave of investment helps explain why KraneShares has expanded beyond traditional AI software and semiconductor companies. Through AGIX's investment in Ayar Labs and the launch of LUMA, the firm is positioning investors to benefit from what it sees as the next critical layer of AI infrastructure: the technologies that connect AI chips at scale.
LUMA Brings Dedicated Exposure to AI Photonics
So why launch another fund if AGIX already owns Ayar Labs? The two products serve different roles in a portfolio:
While AGIX provides diversified exposure across the AI value chain, LUMA offers a more concentrated way to invest specifically in photonics and optical networking. Launched on July 15, 2026, LUMA seeks exposure to both public and private companies developing optical interconnects, transceivers, fiber-optic cables, and other light-based infrastructure that could benefit from the transition from electrical to optical data transmission, mirroring AGIX's public-plus-private approach. As a brand-new fund, performance history is not yet meaningful, but the initial portfolio provides a clear indication of where KraneShares sees value concentrated.
LUMA Top Holdings (as of July 27, 2026)
LUMA Fund Snapshot (as of July 27, 2026)
Notably, LITE and COHR, the two optical component suppliers benefiting from NVIDIA's reported $4 billion commitment, collectively account for just over 12% of LUMA's portfolio, reflecting the fund's concentration in companies already positioned as laser and optical-component suppliers to hyperscalers.
Optical Connectivity Emerges as the Next AI Investment Theme
KraneShares frames this as more than a single product launch. First, optical connectivity is migrating from a peripheral networking component to a core determinant of AI system performance, a shift visible in NVIDIA's own roadmap (Spectrum-X Photonics in H2 2026, Vera Rubin Ultra in 2027, Feynman in 2028). Second, the AI accelerator market is fragmenting: Google's TPU, Amazon's Trainium, Microsoft's Maia, and Meta's MTIA are all custom silicon competing alongside NVIDIA. Because hyperscalers are increasingly developing their own AI accelerators, optical interconnects could become an industry-wide requirement rather than a technology tied to a single chipmaker; each needs flexible, protocol-agnostic optical connections rather than NVIDIA-proprietary NVLink.
Implications for GCC Investors
For investors, the story is shifting from AI compute to AI connectivity. As models become larger and clusters expand, the companies enabling faster, lower-power data movement could become an increasingly important part of the AI investment universe. For GCC investors following global AI infrastructure trends, AGIX's Ayar Labs stake and the LUMA launch mark a maturing thesis: the AI trade is no longer just compute and models; it is increasingly about the infrastructure connecting them. With Goldman Sachs projecting a US$154 billion optical networking opportunity and more than US$11.3 billion already committed through acquisitions, investments, and customer agreements, photonics is emerging as a distinct, investable layer of the AI value chain.
Bottom Line
KraneShares' addition of Ayar Labs to AGIX and the launch of LUMA signal growing conviction that optical connectivity will play a critical role in the next phase of AI infrastructure. As hyperscalers build larger AI clusters and develop their own custom chips, demand for faster, more energy-efficient interconnect technologies is expected to rise. While the commercial adoption of co-packaged optics is still in its early stages, these product launches provide investors with differentiated ways to gain exposure to a structural trend extending beyond AI compute alone.
