DeepSeek has made headlines once again with significant developmentsChina Unicom has announced that its Unicom Cloud has adapted various specifications of the DeepSeek-R1 model, tailored for both domestic and mainstream computing capabilities on the "Xingluo" platformThis advancement serves both private and public scenarios, allowing multi-product scenario callsIt can be pre-deployed in over 270 backbone cloud pools nationwide, fully integrating with Unicom Cloud Desktop, Programming Assistant, and other product environments.

Simultaneously, the latest news from Huawei's computing official account reveals that DeepSeek-R1, DeepSeek-V3, DeepSeek-V2, and Janus-Pro have officially launched on the Ascend communityThis allows users to access the DeepSeek series of models with just a click, supporting immediate usability on Ascend hardware platforms, enabling rapid deployment for inference

Advertisements

Additionally, JD Cloud has officially introduced the DeepSeek-R1 and DeepSeek-V3 models, facilitating both online deployment on public clouds and specialized mixed private instance deployments.

Today, the stocks associated with the DeepSeek concept have surged across the board, with the average increase for the entire sector exceeding 13% at one pointThis topic continues to gain traction.

DeepSeek experiences another jolt

In just 20 days since launch, DeepSeek has achieved a record-breaking active user count of over 20 million, marking yet another milestone for the platform, and this still appears not to be the endEarlier today, significant news arrived from China Unicom’s official WeChat accountAs a leading cloud computing service provider in China, Unicom Cloud has leveraged its version 7.5 of cloud intelligence integration and scenario capability upgrades to establish a deep connection with DeepSeek across multiple scenarios, offering efficient, secure, and flexible cloud computing services to clients from various industries

Advertisements

This reflects a major breakthrough in Unicom Cloud's AI computing capacity, aimed at accelerating the equitable development of AI technology.

Currently, Unicom Cloud has adopted multiple specifications of the DeepSeek-R1 model on the Xingluo platform tailored for both private and public scenariosIt offers comprehensive operational service guarantees, providing customers with an "out-of-the-box" experience; facilitating multi-product scenario calls with pre-deployment capabilities across more than 270 backbone cloud pools nationwide; and offering robust security services, including real-time DDoS protection and enhanced security powered by meta-scenarios.

At the same time, JD Cloud has officially introduced the DeepSeek-R1 and DeepSeek-V3 models with functionalities for online deployment on public clouds and specialized mixed private instance deployments, enabling users to deploy based on their needs and quickly access resources

Advertisements

In the “AI Asset Market/Models,” JD Cloud provides multiple open-source versions of DeepSeek, such as DeepSeek-R1-Distill-Qwen-1.5BThe mixed private instance deployment model has also been synchronized with the JD Cloud vGPU AI computing platform to provide localized model inference services for finance, enterprises, and government clients without data leaving their domains.

According to the latest announcement from Huawei's computing official account, on February 4, the DeepSeek-R1, DeepSeek-V3, DeepSeek-V2, and Janus-Pro went live on the Ascend communityThese models enable one-click access and immediate usability on Ascend hardware platforms, while also offering capabilities like precision performance fine-tuning and rapid deployment of inference services, significantly enhancing the experience of AI development and application.

Concept stocks surge

Today, DeepSeek-related stocks experienced a significant rise, with as many as 19 individual stocks hitting their price limits or exceeding 10% increases, including companies like Xingheng Technology, Youke De, Ankai Micro, Anhui Information, Daily Interaction, and Wanxing Technology, all of which rose by over 20%. China Unicom also saw a sharp uptick in its stock in the afternoon.

On January 20, 2025, DeepSeek officially released the DeepSeek-R1 model, concurrently making the model weights open-source

• Microsoft's Q2 Performance Highlights Potential
• Witnessing History: Gold Breaks $2900!
• Active Cross-Border ETF Trading
• US Inflation Remains Above Target
• South Korea Ups Biopharma Investment

The DeepSeek-R1 has shown performance in mathematical, coding, and natural language reasoning tasks that matches that of OpenAI’s o1 formal versionFrom February 1-2, 2025, both Huawei Cloud and Tencent Cloud announced that they had launched related services based on the DeepSeek-R1 through their respective cloud servicesFurthermore, Microsoft, Nvidia, Amazon, Intel, and AMD have recently introduced DeepSeek model services as well.

Huaxi Securities believes that DeepSeek's success could potentially transform the existing AI industry landscapeThis transformation aligns with China's competitive stance in the global AI industry while simultaneously adding to the competition between open-source and closed-source modelsNotably: 1) In training, the use of FP8 has drawn significant attentionAccording to the Deep Learning and NLP community, DeepSeek-V3 is the first large Mixture-of-Experts (MoE) model to successfully use FP8 mixed-precision training (at least within the open-source community). 2) Unlike OpenAI’s data training model reliant on human intervention, DeepSeek-R1 adopts the R1-Zero route, applying reinforcement learning directly to the foundational model without requiring supervised fine-tuning or labeled data

3) Low-cost models are poised to lead the AI industry along a "new path": combining open-source and MoE4) The debate of open-source versus closed-source sees open-source reshaping the AI ecosystem while allowing both to flourish in downstream applications.

Guohai Securities posits that DeepSeek, supported by China’s technical teams and local computing resources, leads global technological innovation in the frontier domain of large models, garnering attention in both domestic and international marketsThis breakthrough signifies the feasibility of China catching up with overseas AI technology despite limited computing resources, thus boosting the confidence of AI workers globally, including those in China, and invigorating innovation in this fieldSimultaneously, DeepSeek is expected to foster innovation and dissemination of global AI applications and technology through cost optimization and technical innovation, potentially accelerating the arrival of the AGI era.