On July 27, 2026, the Chinese AI startup Moonshot AI officially open-sourced the weights for its flagship AI model, Kimi K3. Boasting a total parameter count of 2.8 trillion (2.8T), Kimi K3 currently stands as the world’s largest open-source model, establishing a new benchmark for frontier AI models in China. Shortly after, DeepSeek released DeepSeek V4 Flash on July 31, and Alibaba Cloud announced Qwen 3.8 Max on August 3. These rapid developments have fueled market discussions regarding a potential “second DeepSeek moment” and the intensifying US-China AI rivalry.
This report provides an in-depth analysis of the following: (1) an overview of Kimi K3, Qwen 3.8 Max, and DeepSeek V4 Flash; (2) a dissection of Kimi K3’s core technologies; and (3) a comparative analysis of AI models developed in China and the United States. This report aims to evaluate mainstream Chinese AI models, elucidate the underlying technical principles and trends of Kimi K3, and assess the competitive landscape between Chinese and American AI capabilities.
Dexterous hands are one of the most costly core modules in humanoid robots. As robot shipments grow, the market for dexterous hands expands in lockstep, and capital is rapidly flowing into specialized dexterous‑hand module manufacturers. Specialist vendors, based on requirements for degrees of freedom, payload, and sensing, have developed three main technical routes: screw‑linkage, tendon‑driven biomimetic, and high‑perception designs, and they offer multiple product lines tailored to different scenarios. Among whole‑machine manufacturers, Tesla has adopted a high‑precision in‑house R&D route; its mass‑production readiness and reliability validation still remain to be fully proven. 1X, by contrast, focuses on home scenarios and compliant interaction, using low‑gear‑ratio tendon drives combined with tactile “skin.”
Green communications leverages nuclear power, liquid cooling, and virtual power plants to overcome energy barriers for 5G and AI, integrating green hydrogen and small reactors for low-carbon power while adopting open architectures and advanced cooling.
The global humanoid robotics industry is advancing rapidly across three key hubs—China, North America, and the broader Asian supply chain—and is approaching a pivotal inflection point around 2026 as the industry transitions from technology validation to early commercialization.
The evolution of the ecosystem showcased at EAI SHOW, the growing pressure for commercial deployment highlighted at the Robotics Summit, and the advances in supply chain capabilities and computing platforms demonstrated at COMPUTEX all indicate that industry competition is shifting away from isolated technological breakthroughs toward system-level competition encompassing AI foundation models, critical components, and standard ecosystems.
Against this backdrop, humanoid robots are no longer viewed merely as an extension of industrial automation. Instead, they are increasingly emerging as the primary embodiment of next-gen Physical AI and a key catalyst for industry-wide transformation.
Following its IPO, SpaceX is building a proprietary satellite ecosystem spanning broadband, direct-to-cell, and space computing, creating fresh global supply chain opportunities. Leading operators use proprietary protocols to strengthen market moats, while facing challenges from space debris and spectrum regulation. Taiwanese suppliers leverage quality and flexibility to deepen ground equipment positioning.
Co-packaged optical (CPO) technology is entering the commercialization stage, driven by AI data center demand for higher bandwidth and lower power consumption. Volume production begins in 2026, with large-scale procurement expected in 2027–2028. Fiber alignment, thermal management, and testing costs are the three core technical barriers to production ramp. Silicon photonics wafers, InP substrates, and FAUs are highly concentrated among a limited number of suppliers, posing supply disruption risks. Taiwanese manufacturers are entering via contract manufacturing and assembly; securing upstream material access will be the defining strategic priority in the next phase.
Optical interconnects in AI data centers are shifting from legacy pluggable optics toward LPO, CPO and OCS. LPO streamlines the signal chain to improve power while keeping a pluggable model, CPO co‑packages optics for higher density and efficiency, and OCS uses optical switching for large, dynamic AI networks, together forming key next‑generation infrastructure.
Intensifying global geopolitical conflicts are driving up defense spending, and warfare is pivoting toward asymmetric and information warfare, making low-cost unmanned vehicles (UAVs) crucial. China is deepening military-civilian integration to break through technology controls, while Taiwan is fully developing localized UAV and AI defense supply chains to strengthen resilience.
With the generative AI market expanding at a CAGR of over 35%, data throughput for a single large language model (LLM) training task has reached the exabyte (EB) level. As transmission speeds evolve toward 1.6Tbps, traditional copper cables are hitting a “physical wall,” limiting transmission distances to under one meter and creating a massive energy drain. Leveraging cross-domain technologies like TSMC’s Compact Universal Photonic Engine (COUPE) advanced packaging and Micro LED mass transfer, Taiwan’s supply chain has built a comprehensive ecosystem—spanning wafer foundries and ASIC design to photoelectric testing. Consequently, Taiwan has become an indispensable strategic hub in the global AI computing infrastructure.
1. Google is consolidating its proprietary TPUs, Ironwood racks, 3D Torus topology, and the Apollo OCS optical backbone into a unified high-speed interconnect architecture. As a result, the focus of cluster planning is shifting from individual servers to modular designs centered on racks and Superpods.
2. Under this architecture, the share of 800G+ high-speed optical modules in data centers deployments is projected to grow from roughly 20% in 2024 to over 60% by 2026. No longer an optional upgrade, these modules are becoming the baseline configuration for next-generation clusters, driving annual demand into the millions.
3. For the optical interconnect supply chain, 800G+ optical modules and OCS systems are becoming critical elements deeply integrated with Google’s infrastructure. Between 2026 and 2028, supply tightness and profitability within the sector will be primarily driven by the available production capacity and yield rates of lasers and MEMS.
4. For industry strategists and investors, assessing the market outlook over the coming years requires looking beyond GPU and TPU shipments. It is equally important to track the shipment volume and penetration of 800G/1.6T optical modules to fully understand the structural shifts between computing power deployment and high-speed interconnect investments.
