/
Navigation
C
Chronicles
Browse all articles
C
E
Explore
Semantic exploration
E
R
Research
Entity momentum
R
N
Nexus
Correlations & relationships
N
~
Story Arc
Topic evolution
S
Drift Map
Semantic trajectory animation
D
P
Posts
Analysis & commentary
P
Browse
@
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
?
Concept Search
Semantic similarity search
!
High Impact Stories
Top coverage by position
+
Sentiment Analysis
Positive/negative coverage
*
Anomaly Detection
Unusual coverage patterns
Analysis
vs
Rivalry Report
Compare two entities head-to-head
/\
Semantic Pivots
Narrative discontinuities
!!
Crisis Response
Event recovery patterns
Connected
Nav: C E R N
Search: /
Command: ⌘K
Embeddings: large
VOICE ARCHIVE

Steve Hsu

@hsu_steve
26 posts
2025-12-18
Taylor Ogan broke this story on Manifold in early 2025. Taylor: ... there are teams at Huawei in two different parts of China that are working around the clock ... literally cut off from their friends and family just because the work they're doing is so sensitive. Imagine how
2025-12-18 View on X
Reuters

Sources: China built a working prototype of an EUV machine in early 2025 with a team of former ASML engineers who reverse-engineered the company's EUV machines

In a high-security Shenzhen laboratory, Chinese scientists have built what Washington has spent years trying to prevent …

Reuters: Shenzhen team completed a working prototype of a EUV machine in early 2025 Government is targeting 2028 for working chips, but sources say 2030 is more likely. ... The team includes recently retired, Chinese-born former ASML engineers and scientists—prime recruitment [image]
2025-12-18 View on X
Reuters

Sources: China built a working prototype of an EUV machine in early 2025 with a team of former ASML engineers who reverse-engineered the company's EUV machines

In a high-security Shenzhen laboratory, Chinese scientists have built what Washington has spent years trying to prevent …

2025-12-05
This is the key step in which GPT5 proposes the main idea of the paper, completely de novo: [image]
2025-12-05 View on X
@hsu_steve

Physicist Steve Hsu says he has published a peer-reviewed theoretical physics paper whose main idea came from GPT-5

Steve Hsu / @hsu_steve :

2025-08-27
FT on SMIC ramp up in AI chip production. Qwen AI: Based on the retrieved search results, here's the available information about SMIC's Ascend AI chip production projections: For 2025: Huawei is projected to ship approximately 700,000 Ascend AI chips in 2025, though yield [image]
2025-08-27 View on X
Financial Times

Sources: a Chinese fab making Huawei's AI chips is set to start production as soon as the end of 2025, with two more due in 2026; the fabs' ownership is unclear

Fabrication plants serving Huawei push to increase production as US cuts off access to Nvidia's top processors

2025-08-22
100 AI prompts can boil a cup of water! “the median prompt—one that falls in the middle of the range of energy demand—consumes 0.24 watt-hours of electricity, the equivalent of running a standard microwave for about one second” [image]
2025-08-22 View on X
MIT Technology Review

Google says the median Gemini app text prompt consumes 0.24Wh of energy, about the same as running a microwave for a second, and emits 0.03g of CO2 equivalent

an official report confirms that Gemini consumes per query:  — 0.24 Wh of energy (~9 seconds of TV)  — 0.03 g of CO2 equivalent  — 0.26 ml of water (about 5 drops)  —  blog: cloud....

2025-08-14
DeepSeek R2 delay due to transition to Huawei Ascend chip for training? DS + HW engineers collaborating on CUDA to CANN migration is ultimately positive for HW in the long run. R2 release was originally expected last May. Since then at least one SOTA Chinese model has been
2025-08-14 View on X
Financial Times

Sources: DeepSeek R2's launch delay is due to training issues on Huawei Ascend chips, prompting a switch to Nvidia chips for training and Huawei's for inference

Difficulties of training the start-up's latest system with Huawei's semiconductors highlight dependence on Nvidia

2025-07-24
FT: Nvidia AI chips worth $1bn smuggled to China after Trump export controls No shortage of B200 in China. Mostly deployed in data centers out of reach from US regulations, but available for use by PRC bigtech companies. Qwen AI Summary of the Article: The Financial Times [image]
2025-07-24 View on X
Financial Times

Analysis and sources: $1B+ of Nvidia's AI chips were shipped to China in the three months after President Trump tightened chip export controls in May

Roaring black market for US semiconductors operates despite efforts to curb Beijing's high-tech ambitions

2025-04-20
Some details about Huawei's next AI chip. Note SMIC 6nm process - suggests 7nm yields are acceptable for 910C, as was reported earlier this year. “Ascend 920 is a direct competitor to NVIDIA's H20, given that special focus has been made on ensuring that the supply chain is
2025-04-20 View on X
Tom's Hardware

Huawei unveils the Ascend 920 AI chip, expected to hit mass production in H2 2025, after Trump included Nvidia H20 in the ban on AI chip exports to China

The Ascend 920 supposedly offers a performance comparable to the Nvidia H20. … Just one day after this, Huawei announced the Ascend 920 …

2025-04-17
Jensen in Beijing: ... Huang met Nvidia clients, including the founder of generative artificial intelligence start-up DeepSeek, to discuss new chip designs for Chinese customers. ... Huang has called Huawei China's ‘single most formidable tech company’ https://www.ft.com/...
2025-04-17 View on X
Financial Times

Nvidia CEO Jensen Huang arrived in China on April 16 and visited Beijing on April 17, a source says to meet officials to discuss Trump's new chip export rules

Meetings with Chinese vice-premier and DeepSeek founder come after US clamps down on chipmaker's sales to China

2025-02-25
@teortaxesTex I wouldn't trust the numbers from FT. It's all rumors and the reporters aren't chip experts. But directionally it's plausible that SMIC/HW are improving Ascend yields a lot. I think SemiAnalysis has even larger projections for 910C shipments in 2025.
2025-02-25 View on X
Financial Times

Sources: Huawei has increased its AI chip yield to nearly 40%, up from 20% about a year ago, meaning its Ascend chips are now profitable for the first time

Chinese company improves ‘yield’ of latest semiconductor, despite US efforts to prevent manufacturing advances

FT on Huawei / SMIC Ascend 910C chip yields: doubled to 40% and now targeting 60%. TSMC H100 yield is ~60% ? When I was in Shanghai I had dinner with one of the top SMIC guys involved with leading edge chip fabrication. The team has a big Taiwanese component with experience at [image]
2025-02-25 View on X
Reuters

Sources: Tencent, Alibaba, ByteDance, and other Chinese companies are ramping up orders for Nvidia's H20 AI chip due to booming demand for DeepSeek's models

Chinese companies are ramping up orders for Nvidia's (NVDA.O) H20 artificial intelligence chip due to booming demand …

@teortaxesTex I wouldn't trust the numbers from FT. It's all rumors and the reporters aren't chip experts. But directionally it's plausible that SMIC/HW are improving Ascend yields a lot. I think SemiAnalysis has even larger projections for 910C shipments in 2025.
2025-02-25 View on X
Reuters

Sources: Tencent, Alibaba, ByteDance, and other Chinese companies are ramping up orders for Nvidia's H20 AI chip due to booming demand for DeepSeek's models

Chinese companies are ramping up orders for Nvidia's (NVDA.O) H20 artificial intelligence chip due to booming demand …

FT on Huawei / SMIC Ascend 910C chip yields: doubled to 40% and now targeting 60%. TSMC H100 yield is ~60% ? When I was in Shanghai I had dinner with one of the top SMIC guys involved with leading edge chip fabrication. The team has a big Taiwanese component with experience at [image]
2025-02-25 View on X
Financial Times

Sources: Huawei has increased its AI chip yield to nearly 40%, up from 20% about a year ago, meaning its Ascend chips are now profitable for the first time

Chinese company improves ‘yield’ of latest semiconductor, despite US efforts to prevent manufacturing advances

2025-01-29
Another day, another frontier model from China: Alibaba this time with Qwen2.5-Max. https://qwenlm.github.io/... [image]
2025-01-29 View on X
Reuters

Alibaba releases Qwen 2.5-Max, an AI model that the company's cloud unit claims “outperforms” GPT-4o, DeepSeek-V3, and Llama-3.1-405B “almost across the board”

Chinese tech company Alibaba (9988.HK) on Wednesday released a new version of its Qwen 2.5 artificial intelligence model …

2024-12-27
DeepSeek V3 is out. Big improvements. This model is fast, inexpensive to run, and seems to beat Claude, 4o etc. on a broad set of benchmarks. Perhaps most importantly V3 is still open source! Keep in mind DS is a hedge fund whose founder (educated in AI at Zhejiang University)
2024-12-27 View on X
VentureBeat

DeepSeek releases DeepSeek-V3, an open-source MoE model of 671B total parameters, with 37B activated per token, claiming it outperforms top models like GPT-4o

Chinese AI startup DeepSeek, known for challenging leading AI vendors with its innovative open-source technologies, today released a new ultra-large model: DeepSeek-V3.

2024-10-14
Alex Gerko in FT. Not on X anymore, but he posts a lot of fun stuff on LinkedIn https://www.linkedin.com/... https://www.ft.com/... [image]
2024-10-14 View on X
Financial Times

A profile of Alex Gerko, who used an Icelandic supercomputer and 25K AI chips to build XTX, an algorithmic trading firm that handles $250B of daily trades

XTX Markets conquered foreign exchange trading and made its Russian-born founder a multibillion-pound fortune X: @hsu_steve and @tonytassell X: Steve Hsu / @hsu_steve : Alex Gerko ...

2024-10-09
Wow! DeepMind Nobels for AlphaFold! Jumper's background is in Physics. To be honest I wan't aware that Hassabis was very involved in this at a technical level... https://en.wikipedia.org/...
2024-10-09 View on X
Bloomberg

The Nobel in Chemistry goes to David Baker “for computational protein design” and DeepMind's Demis Hassabis and John Jumper “for protein structure prediction”

- Demis Hassabis, John Jumper share half the $1.1 million award  — Remainder goes to David Baker for building new proteins

2024-09-13
If this article is at all accurate the situation for venture-backed startups in China has deteriorated horribly in the last couple of years. Very bad for the innovation ecosystem. https://www.ft.com/... I seem to hear constantly about Chinese startups in chips and AI, just to
2024-09-13 View on X
Financial Times

Sources paint a bleak picture for startup founders in China; IT Juzi report: only 1,202 startups were founded in the country in 2023 compared to 51,302 in 2018

Venture capital finance has dried up amid political and economic pressures, prompting a dramatic fall in new company formation X: @carnage4life , @tonytassell , @michaelxpettis , @...

2024-08-26
Durov on US pressure to install backdoors in Telegram. Now the French have him in custody.
2024-08-26 View on X
Reuters

Sources: Telegram CEO Pavel Durov was arrested in France as part of a probe into Telegram allegedly allowing a range of crimes due to a lack of moderation

Related by education: Saint Petersburg State University BBC : Telegram CEO Pavel Durov arrested at French airport Jones Hayden / Politico : Telegram CEO reportedly arrested at Fren...

Durov on US pressure to install backdoors in Telegram. Now the French have him in custody.
2024-08-26 View on X
Politico

French President Emmanuel Macron says the government was not involved in the arrest of Telegram CEO Pavel Durov and “this is in no way a political decision”

French president condemns “false information” as Pavel Durov's detention in Paris triggers global outcry.