Overenthusiastic Selloff of Nvidia Stocks Disregards Challenges from Huawei's 910D
Rewritten Article:
Last week, tech titan Nvidia (NASDAQ:NVDA) encountered renewed scrutiny after it was revealed that Huawei has reportedly asked some Chinese companies to give those 910D AI chips a spin, a device Huawei claims could outshine Nvidia's lauded H100 chip. However, objective benchmarks and precise performance specifications from independent sources have yet to be seen[3][4]. Currently, the H100 tops the charts in raw compute power, boasting 2.8 times higher FP16 and 2.6 times higher BF16 performance than the 910D[5].
Nonetheless, the racing game between Huawei and Nvidia isn't sole performance-based. The 910D, manufactured on SMIC's 7nm N+2 process, uses a dual-chiplet design to deliver competitive INT8 throughput for inference tasks, despite having lower yield rates of around 40-50%[5]. On the other hand, the H100, constructed on TSMC's 4nm-class node, enjoys higher yield rates of approximately 90% and far greater transistor density[5].
Huawei's push into the AI chip race means Nvidia may have to stay one step ahead to maintain its place at the top. The emergence of the 910D could drive Nvidia to explore new avenues of innovation[1][4]. Furthermore, trade restrictions and a predilection for domestic markets might leave Huawei at a disadvantage globally, where Nvidia's well-established ecosystem and superior manufacturing abilities matter[4][5]. All in all, technological advancements in AI chip design could accelerate, propelling the industry forward[4].
In a nutshell, while Huawei eyes dethroning Nvidia, the H100 generally outperforms the 910D on sheer calculating power. The challenges to Nvidia, then, are fierce competition and an opportunity for innovation, but their dominance in the global market remains thanks to their established network and technical excellence in manufacturing.
- Huawei's AI chip, the 910D, is being tested against Nvidia's H100, aiming to surpass its performance in the artificial-intelligence sector.
- Despite the 910D's competitive INT8 throughput for inference tasks, Nvidia's H100, manufactured on TSMC's 4nm-class node, enjoys higher yield rates and greater transistor density.
- The intense competition between Huawei and Nvidia in the AI chip race could lead to significant advancements in technology, accelerating the progress of the industry.
