AI compute surges in 2026 as ASICs rise, NVIDIA defends its ecosystem, and bandwidth becomes the bottleneck.
Strong CSP capex fuels steady AI server growth as cloud giants expand GPU rack orders and self-developed chip programs.
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...
Global Capex rises on AI buildout as China speeds self-reliant chips, easing foreign hardware reliance.
During the first half of 2026, surging demand for KV Cache coupled with constrained memory supply resulted in severe memory bottlenecks. To resolve the KV Cache bottlenecks, industry players are seeki...
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 t...
The AI industry entered a critical transition period in 2025, marked by the rapid expansion of LLMs and the large-scale commercialization of AI applications. For China’s semiconductor industry, this r...
In January 2026, NVIDIA introduced the CMX Context Memory Storage Platform, managed by the BlueField‑4 DPU, to extend the memory hierarchy between local SSD and shared storage and address the massive...
This report first explains why HBM plays a key role in the evolution of LLM and AIGC applications. It then examines the HBM upgrade roadmap and explores how heterogeneous memory integration strategies...
As technical iteration narrows the capability gap between models, Chinese open-source models are expanding their global presence by compensating weaker hardware with software optimization and scale. U...
In AI inference, MoE architectures and long-context processing have sharply increased memory-capacity requirements for model weights and KV cache, shifting the bottleneck from insufficient compute to...
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 t...
Humanoid robot manufacturers in the US and China are developing pre-trained datasets and new training architectures to enable humanoid robots to adapt to a wider range of applications.
DeepSeek’s technological innovations reduce reliance on massive computing power, but as AI applications continue to expand, long-term demand for compute resources will keep rising. Its emergence is pu...
TrendForce analyzes that DeepSeek reduces AI deployment costs through optimized algorithms, creating new opportunities for SMEs. While the short-term impact on the memory market is limited, in the lon...
