Skip to content

Project

Huawei Ascend

Huawei's Ascend series of AI accelerator chips (NPUs) for training and inference, positioned as a domestic alternative to Nvidia GPUs in China.

Known aliases

  • Ascend
  • Ascend 960
  • Ascend 970
  • Ascend AI chips
  • ASCEND consortium
  • Ascend NPU
  • Huawei Ascend

Current stories

investOne report1 publisher

DeepSeek plans 160,000 Huawei Ascend chips for inference while training stays on Nvidia

Huawei claims its Ascend chips have outsold Nvidia in China as of October 2026, but DeepSeek keeps its training work on Nvidia hardware. Unit sales and training workloads are separate measures, and sizing either company's China business needs both.

Reality

Evidence30
Adoption40
Hype gap+25
Incentives
Not scored
Confidence40
investConfirmed8 publishers

Huawei answers Nvidia's per-chip lead by wiring 4,000 processors into one pod

The Ascend 960 is dated 2027 and the 970 is dated 2028, but the only large order on the record is DeepSeek's 160,000 chips of the current generation. That fleet is forty pods of silicon Huawei ships today.

Perspective Coverage

8 publishers
Builder
Builder 35%
Operator
Operator 30%
Investor
Investor 35%

Reality

Evidence55
Adoption35
Hype gap+40
Incentives65
Confidence60
investOne report1 publisher

Washington think tank counts 343 immersion DUV machines in Chinese-owned fabs

Chinese-owned chip plants held 343 immersion DUV machines by early this year, about 270 of them ASML NXT:1980i units, a Washington think tank says. Its authors want all new DUV shipments stopped, and a bipartisan bill in Congress would require the Netherlands and Japan to do exactly that.

Publishers:en.sedaily.com

Reality

Evidence35
Adoption55
Hype gap+25
Incentives70
Confidence40
buildConfirmed4 publishers

DeepSeek's Ascend ports swap a CUDA dependency for Huawei's CANN

DeepSeek open-sourced six software components for Huawei's Ascend AI chips on September 30, porting infrastructure it first built for Nvidia hardware. The ports cut the software bill for leaving CUDA, and The Neuron argues Ascend's case now turns on chip supply, reliability and full-scale model performance.

Perspective Coverage

4 publishers
Builder
Builder 41%
Operator
Operator 24%
Investor
Investor 35%

Reality

Evidence60
Adoption25
Hype gap+30
Incentives70
Confidence60
buildConfirmed2 publishers

Huawei reserves the Kirin 9050 Pro for its top Mate 90 models

Huawei's Mate 90 launch gives the Kirin 9050 Pro only to its top editions, with a Kirin 9035 in the Pro and a Kirin 9030 in the standard model. Richard Yu said smartphone chip capacity remains tight and is shared with Huawei's Ascend AI chips.

Publishers:d2233.cms.socastsrm.com economictimes.indiatimes.com en.prnasia.com mezha.net

Perspective Coverage

4 publishers
Builder
Builder 40%
Operator
Operator 29%
Investor
Investor 31%

Reality

Evidence60
Adoption30
Hype gap+30
Incentives70
Confidence60
investOne report1 publisher

Huawei claims 244 gigabytes of shared HBM for each of the Atlas 960E's 4,096 chips

Huawei used HUAWEI CONNECT 2026 to publish its own specification for a 4,096-chip cluster with a petabyte of shared memory, aimed at models approaching 10 trillion parameters.

Reality

Evidence30
Adoption
Not scored
Hype gap+45
Incentives65
Confidence35
productOne report1 publisher

NHTSA and SAE launch ASCEND to write the driving test for automated vehicles

The question of when a robot driver is good enough has been the industry's to answer. NHTSA's new consortium with SAE is meant to produce the competency test that would settle it, with the agency's rulemaking to follow.

Publishers:techcrunch.com

Reality

Evidence58
Adoption20
Hype gap+22
Incentives62
Confidence55
productOne report1 publisher

Huawei sells AI data centre power gear on a promise to steady the grid

The grid-interactive stack unveiled in Shenzhen pairs a 1,280 kVA UPS cabinet and grid-forming storage with a thermal unit built for Ascend servers. Huawei wants buyers judging it on tokens per watt in place of PUE.

Publishers:thenextweb.com

Reality

Evidence36
Adoption24
Hype gap+30
Incentives82
Confidence45
securityOne report1 publisher

Washington names industrial-scale distillation, then hands the detection bill to abuse teams

A White House memo calls Chinese model distillation deliberate and industrial in scale, then answers it with information sharing, best practices and a demarche. The detection work stays with providers.

Publishers:news.risky.biz

Reality

Evidence55
Adoption38
Hype gap+32
Incentives66
Confidence58
investOne report1 publisher

Korea's 720-billion-dollar AI plan meets its first critic: the man running its research hub

Seoul wants top-three status by 2035 on more than 1,000 trillion won. The inaugural director of its National AI Research Hub says the money is budgeted on the wrong clock.

Publishers:en.sedaily.com

Reality

Evidence32
Adoption28
Hype gap+22
Incentives66
Confidence38
investOne report1 publisher

Beijing can ban Nvidia purchases faster than it can replace CUDA

China's chip directive is a procurement decision the domestic hardware cannot yet honor. The constraint is not silicon supply, it is nearly two decades of software nobody has rebuilt.

Reality

Evidence26
Adoption58
Hype gap+18
Incentives
Not scored
Confidence31