Huawei Ascend 950 artificial intelligence accelerator revealed with custom HBM

Huawei has revealed its Ascend 950 artificial intelligence accelerator family, pairing in-house HBM memory with a focus on large-scale system performance rather than single-chip dominance.

Huawei’s next-generation Ascend 950 artificial intelligence accelerator has been shown in detail for the first time, highlighting the company’s custom silicon and high bandwidth memory. The design combines Huawei’s first self-developed HBM with a new generation of artificial intelligence acceleration, signaling a strategy built around vertically integrated components and system-level optimization.

The Ascend 950 family is planned for early 2026 and will launch with two variants. The 950PR model includes 128 GB of in-house HBM with around 1.6 TB/s of bandwidth, while the 950DT model increases memory to 144 GB and boosts bandwidth to nearly 4 TB/s. Both chips target one PetaFLOP of FP8, and two PetaFLOPS of FP4, positioning them as high-performance accelerators even though Huawei currently trails NVIDIA in per-chip performance.

Huawei’s competitive approach centers on dense packaging and aggressive networking across clusters, aiming to win through scale rather than relying solely on raw per-chip metrics. Although the company has not confirmed the manufacturing process, the accelerator will likely use SMIC’s newest N+3 node with 5 nm-class features. Chinese SMIC has officially achieved volume production of its newest 5 nm node relying on the deep ultraviolet to manufacture its silicon, and since the first customer for N+3 was Huawei with the Kirin 9030 SoC, it is seen as logical that the Ascend artificial intelligence accelerator family would also be produced on the same node.

64

Impact Score

SK Hynix warns of tight commodity DRAM supply through 2028

SK Hynix expects tight supply of commodity DRAM such as DDR5, GDDR6, and LPDDR5x to persist through 2028, putting gamers and PC buyers at risk of higher memory prices, while advanced HBM and SOCAMM lines continue to expand capacity for Artificial Intelligence hardware.

Artificial Intelligence transforms scientific research with ethical safeguards

Artificial Intelligence is reshaping scientific research through autonomous labs, hypothesis-generating systems, and cross-disciplinary applications, while sparking parallel efforts to build ethical and governance frameworks. The article tracks how industry, academia, and governments are trying to balance rapid advances with quality control, transparency, and safety.

From bytes to bedside: artificial intelligence in medicine and medical education

A new clinical obstetrics and gynecology article argues that rapidly advancing generative artificial intelligence and large language models are set to reshape both patient care and medical training, while stressing the need for ethical and safe implementation. The authors describe how these systems are already demonstrating clinical reasoning capabilities and propose a framework for integrating them responsibly into health care and education.

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.