PulseAugur
EN
LIVE 10:58:30

DeepSeek V4, Huawei Ascend 950DT Co-Design Slashes AI Inference Costs 75%

By PulseAugur Editorial · [2 sources] ·

A new analysis from SemiAnalysis reveals that the DeepSeek V4 AI model and Huawei's Ascend 950DT AI accelerator were co-designed to significantly reduce inference costs. This collaboration has reportedly cut costs by 75%, bringing the price down to 0.20 yuan per million tokens. The findings highlight the specific optimizations made within Huawei's CANN 8.5 stack and the Ascend 950DT's architecture to cater to DeepSeek's inference patterns. AI

IMPACT This co-design approach demonstrates a path to significantly lower inference costs, potentially accelerating AI adoption and deployment.

RANK_REASON Analysis report detailing technical co-design and cost savings for an AI model and hardware.

Read on Pandaily →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

DeepSeek V4, Huawei Ascend 950DT Co-Design Slashes AI Inference Costs 75%

How we ranked this

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
Analysis report detailing technical co-design and cost savings for an AI model and hardware.
Source corroboration
2 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
Topics
infra, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
130 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

Full methodology in our editorial standards.

COVERAGE [2]

  1. Pandaily TIER_1 English(EN) · [email protected] (Pandaily) ·

    DeepSeek V4 and Huawei Ascend 950DT: The Co-Designed Chip That Cut AI Inference Costs by 75%

    A groundbreaking trace-level analysis by Wall Street research firm SemiAnalysis has revealed that DeepSeek V4 and Huawei Ascend 950DT AI accelerator were co-des...

  2. Mastodon — mastodon.social TIER_1 English(EN) · [email protected] ·

    DeepSeek V4 and Huawei Ascend 950DT AI accelerator were co-designed from scratch, cutting inference costs by 75% to just 0.20 yuan per million tokens. SemiAnaly

    DeepSeek V4 and Huawei Ascend 950DT AI accelerator were co-designed from scratch, cutting inference costs by 75% to just 0.20 yuan per million tokens. SemiAnalysis found Huawei's CANN 8.5 stack and the unique dual-die 950DT architecture were built specifically for DeepSeek infere…

    LINKS pandaily.com/deepseek-v4-huawei-ascend-95…

RELATED ENTITIES

RELATED TOPICS