Huawei's Ascend AI Ecosystem Accelerates: Huawei announced this week new progress in its Ascend AI computing industry ecosystem, with several partners launching servers and solutions based on the Ascend 910B, aiming to further expand its influence in the domestic AI computing market.
Moore Threads Releases New Driver: Domestic GPU company Moore Threads released a significant update to its community edition driver this week, significantly improving compatibility and performance for multiple games and optimizing the experience for AI and video processing applications, demonstrating its continued investment in the software ecosystem.
Biren Technology's BR100 Series Sees Application: Reports indicate that Biren Technology's first general-purpose GPU chip series, the BR100, has been deployed in several domestic AI computing centers and has begun providing computing power for tasks like large model training.
Tencent Cloud Launches New Generation Computing Cluster: Tencent Cloud announced that its new generation of self-developed Starsea servers will integrate more efficient AI accelerators, aiming to reduce the cost of large-scale AI training and inference and improve the price-to-performance ratio of its cloud computing services.
Policy Supports AI Computing Infrastructure: The national level reiterated the acceleration of building an integrated national computing system, with several local cities announcing new AI computing center construction plans, providing a broad market outlook for domestic GPUs.
NVIDIA Releases Next-Gen Blackwell Architecture GPUs: NVIDIA officially launched the B200 and GB200 superchips based on the Blackwell architecture. Their AI performance represents a massive leap over the previous generation, notably reducing the energy consumption for training trillion-parameter models to 1/25th, drawing significant industry attention.
AMD MI300 Series Delivery Ramps Up: AMD CEO Dr. Lisa Su stated this week that customer shipments of its Instinct MI300 series AI accelerators are accelerating, with projected 2024 revenue exceeding $3.5 billion, indicating substantial progress in challenging NVIDIA's market position.
Microsoft and OpenAI Plan Massive Supercomputer: According to reports, Microsoft and OpenAI are planning a data center project codenamed "Stargate," equipped with millions of dedicated AI chips and expected to be extremely costly, highlighting the extremely high expectations for future computing demand.
Intel Gaudi 3 AI Chip Officially Debuts: Intel officially launched its new-generation AI accelerator, Gaudi 3, claiming it outperforms NVIDIA's H100 GPU in both training and inference performance, aiming to provide AI companies with a new cost-effective alternative.
AWS Launches New Self-Developed AI Chips: Amazon Web Services released its self-developed Trainium2 AI accelerator and deepened its cooperation with NVIDIA on the Grace Hopper Superchip, intensifying competition in the cloud AI computing market by offering a diversified portfolio of computing power to meet different customer needs.