Huawei Challenges Nvidia’s AI Dominance, Pushes Homegrown Ascend Chips in Mideast and Southeast Asia Alternative concise options: 1. Huawei Pitches Ascend AI Chips Across Mideast, Southeast Asia in Nvidia Challenge 2. Beyond Nvidia: Huawei Markets Self-Developed Ascend AI Chips to Mideast, Southeast Asia 3. Huawei Seeks Mideast, Southeast Asia Sales for Ascend AI Chips to Rival Nvidia Selected best fit: **Huawei Challenges Nvidia’s AI Chip Dominance with Homegrown Ascend Line in Mideast, Southeast Asia**

Huawei is expanding into the global AI chip market, targeting the Middle East (UAE, Saudi Arabia) and Southeast Asia (Thailand) with its Ascend 910B chip, challenging NVIDIA. To attract clients, it’s promoting its CloudMatrix 384 super AI server, powered by the Ascend 910C chip. This server integrates 384 Ascend NPUs and 192 Kunpeng CPUs, offering high throughput (2,300 tokens/sec) and massive scalability (up to 160,000 cards). It supports stable, resilient operation. While highlighting its AI cloud services as ideal for large models, Huawei is not exporting the Ascend 910C overseas due to supply constraints, prioritizing Chinese firms impacted by US restrictions. (98 words)

CNBC News reports on July 11 that Huawei is attempting to break into the AI chip market dominated by NVIDIA with a strategic push to export its self-developed chips to the Middle East and Southeast Asia, according to sources familiar with the matter.

Insiders reveal that Huawei has recently engaged with potential clients in the UAE, Saudi Arabia, and Thailand to market its Ascend 910B AI chip. This move signals Huawei’s ambition to carve out a share in the high-growth global AI semiconductor arena.

Illustration of Huawei AI chip technology

To lure customers, Huawei is also promoting its proprietary “super AI server” CloudMatrix 384, which leverages the advanced Ascend 910C chips. As a showcase of its capabilities, Huawei Cloud demonstrated the CloudMatrix 384’s computational power in a recent video, highlighting:

384 Ascend NPUs and 192 Kunpeng GPUs interconnected to form a unified system.

Industry-leading single-card inference throughput: 2,300 tokens per second.

Unmatched cluster computing scale: 160,000 cards with 95% linear scalability.

Cloud-based resilience: 40-day stable training and 10-minute recovery times.

Visualization of Huawei CloudMatrix server

Huawei Cloud emphasizes that its latest Ascend AI cloud services represent the optimal computing solution for large-scale AI model deployments, marrying performance with efficiency. However, supply constraints mean Huawei has no immediate plans to export the Ascend 910C chips overseas, as it prioritizes allocation to Chinese enterprises seeking alternatives to U.S.-restricted advanced chips.

Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/4471.html

Like (0)
Previous 6 hours ago
Next 4 hours ago

Related News