China´s tech giants hesitate to adopt Huawei chips over overheating and NVIDIA software ecosystem

Despite Huawei´s ambitions, China´s leading tech firms are slow to adopt its Artificial Intelligence chips, citing overheating and strong NVIDIA ecosystem lock-in.

Huawei´s efforts to curb NVIDIA´s dominance in China´s Artificial Intelligence hardware market with its Ascend 910C GPUs are facing substantial resistance from the country´s top technology firms. ByteDance, Alibaba, and Tencent have reportedly not committed to large-scale purchases of Huawei´s chips, highlighting the challenge of breaking free from the entrenched NVIDIA CUDA software ecosystem and addressing technical limitations of Huawei´s products.

A major obstacle for Huawei is the deep investment Chinese tech companies have made in NVIDIA´s CUDA platform, which offers a mature and feature-rich environment for developing Artificial Intelligence workloads. Huawei´s alternative, CANN (Compute Architecture for Neural Networks), lacks critical features and compatibility, making transitioning costly and time-consuming. Many industry leaders expect Huawei to adapt to their custom requirements, rather than overhauling existing infrastructures.

Reliability concerns also persist, as the Ascend 910C GPUs reportedly suffer from periodic overheating, further eroding confidence among potential buyers. In addition to technical reservations, competitive dynamics make leading technology firms wary of adopting a rival´s chips, slowing market uptake. Major companies have stockpiled NVIDIA GPUs over recent years, reducing the urgency to switch despite expanding US sanctions and export restrictions that have cast a shadow on Huawei hardware, especially for firms with significant global operations.

Huawei attempts to tackle functionality gaps by introducing products like the CloudMatrix 384, designed to rival NVIDIA´s high-performance Blackwell-based supercomputers. Nevertheless, the lack of direct support for memory-efficient formats like FP8 limits appeal, and stopgap solutions fail to match NVIDIA’s efficiency. While Huawei’s Ascend 910C’s headline specifications compete with NVIDIA’s H100 GPU, inertia from inventory, compliance risks, and insufficient developer community support further blunt its competitive edge. Meanwhile, NVIDIA continues to thrive, forecasting billions in Artificial Intelligence infrastructure revenue despite headwinds from Chinese sanctions.

76

Impact Score

Deepfake porn and chatbot privacy breaches

Nonconsensual deepfake pornography is harming not only people whose faces are inserted into explicit media, but also adult creators whose bodies and likenesses are reused without permission. Generative Artificial Intelligence chatbots are also exposing private phone numbers, making personal information easier to retrieve and harder to control.

European Union Artificial Intelligence Act raises layered compliance demands for finance

Banks, insurers and financial intermediaries face a more complex compliance environment as the European Union Artificial Intelligence Act overlays existing financial regulation and the GDPR. Proposed changes in the Digital Omnibus Package may delay some obligations, but the core challenge remains managing overlapping rules, roles and regulators.

Europe and US discuss biometric data-sharing framework

European Union and US officials are negotiating a border security arrangement that could enable continuous biometric data exchanges on EU citizens. The UK says the US has also requested access to fingerprint records as part of Visa Waiver Program discussions.

Apple plans Intel 18A-P for M7 and 14A for A21

Apple is expected to use Intel’s 18A-P process for M7 chips in MacBook models and Intel’s 14A process for A21 chips in iPhones. The shift points to a broader supplier strategy as Apple moves beyond TSMC for parts of its future silicon roadmap.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.