Chinese semiconductor thread II

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
You can see the jump in Chinese chip tools purchases after they started sanctioning Huawei and ZTE circa 2018.
Purchases went from like $15 billion USD per month on average to $30 billion USD. And ever since they started to sanction the fabs themselves the purchases ramped up even more to the current like $55 billion USD average per month.

Compare that with the pathetic $12 billion USD per month chip tool purchases the US makes. And these are typically leading edge tools with much higher cost per unit and per wafer.

I once looked at the numbers, and SMIC is headed towards surpassing TSMC in 28nm fabrication capacity. You will see it happen over the next two years.

Nexchip already has similar capacity to its parent company Powerchip Semiconductor (PSMC) in Taiwan.

The only laggard is Hua Hong. They are mostly increasing capacity in legacy nodes 55nm and lower. Where you would expect them to be at least trying to compete with the likes of UMC in 28nm capacity. Maybe a sign of things to come after they finish 55nm expansion.
Please, Log in or Register to view URLs content!
keep in mind that Wuxi phase 1 was all 55nm and lower, but that will change with phase 2 where they are moving to 40nm node. Which is expected to require $7B in capex. I would still expect them to not be involved in leading edge though. It's clear at this point that only Huawei is looking to compete in leading edge node with SMIC.
 

gelgoog

Lieutenant General
Registered Member
keep in mind that Wuxi phase 1 was all 55nm and lower, but that will change with phase 2 where they are moving to 40nm node. Which is expected to require $7B in capex. I would still expect them to not be involved in leading edge though. It's clear at this point that only Huawei is looking to compete in leading edge node with SMIC.
I doubt Hua Hong Wuxi phase 2 will remain at 40nm. Hua Hong is now competing with Nexchip. Nexchip is working on 40nm and 28nm processes for DDICs.
Please, Log in or Register to view URLs content!

Hua Hong will need to also have 28nm process to compete.

To make 40nm I think you need immersion lithography tools anyway, so it should not be that hard to add 28nm to the mix.
Hua Hong can simply license the 28nm process from HLMC just like they did with 55nm. And HLMC maybe can be financed to go into FinFET.
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
I doubt Hua Hong Wuxi phase 2 will remain at 40nm. Hua Hong is now competing with Nexchip. Nexchip is working on 40nm and 28nm processes for DDICs. Hua Hong will need to also have 28nm process to compete.

To make 40nm I think you need immersion lithography tools anyway, so it should not be that hard to add 28nm to the mix.
Hua Hong can simply license the 28nm process from HLMC just like they did with 55nm. And HLMC maybe can be financed to go into FinFET.
that's just what they announced. They are planning to take back HLMC and there will be future HLMC fab expansion also.
 

gelgoog

Lieutenant General
Registered Member
Please, Log in or Register to view URLs content!

Nexchip joins Chinese indigenization drive with photomask production, aiming for full-service foundry status​

25 July 2024

Nexchip is expanding into photomask production, aiming to achieve mass production by the fourth quarter. This move positions Nexchip as a full-service foundry, similar to TSMC and SMIC, and highlights the indigenization efforts of China-based chipmakers.
 

measuredingabens

Junior Member
Registered Member
Please, Log in or Register to view URLs content!

A team of scientists from Beijing has announced a groundbreaking advancement in artificial intelligence (AI) technology with the development of the world’s first fully optical AI chip.
This innovative chip, known as Taichi-II, represents a significant leap forward in both efficiency and performance, surpassing even the renowned NVIDIA Corp. NVDA H100 GPU in energy efficiency.
The research team, led by professors Fang Lu and Dai Qionghai from Tsinghua University, unveiled their findings on Wednesday.

A leap beyond: The Taichi-II chip’s superiority​

The Taichi-II chip represents a substantial advancement from its predecessor, the Taichi chip, which had already set impressive records. Earlier this year, the researchers announced that the original Taichi chip had exceeded the energy efficiency of NVIDIA’s H100 GPU by over a thousand times, as
Please, Log in or Register to view URLs content!
by South Morning China Post (SCMP).
Now, the Taichi-II chip has further elevated this benchmark, showcasing superior performance across various scenarios.
The study led by Professors Fang Lu and Dai Qionghai highlights Taichi-II’s capability to transform
Please, Log in or Register to view URLs content!
training and modeling. Unlike traditional methods that rely on electronic computers for training, the Taichi-II leverages optical processes, making it more efficient and significantly enhancing performance.
In practical terms, the Taichi-II chip has demonstrated remarkable advancements in several areas. It has expedited the training of optical networks containing millions of parameters by an order of magnitude and improved the accuracy of classification tasks by 40 percent.
In complex imaging scenarios, its energy efficiency in low-light conditions has improved by six orders of magnitude.

Innovative approach: FFM learning​

The
Please, Log in or Register to view URLs content!
of the Taichi-II chip is marked by its use of a novel approach called fully forward mode (FFM) learning. This technique allows for a computer-intensive training process to be conducted directly on the optical chip, enabling parallel processing of machine learning tasks.
Xue Zhiwei, lead author of the study and a doctoral student, emphasized that this architecture supports high-precision training and is well-suited for large-scale network training.
“Our research envisions a future where these chips form the foundation of optical computing power for AI model construction,” Fang Lu stated.
The FFM learning method capitalizes on high-speed optical modulators and detectors, which could potentially outperform GPUs in accelerated learning scenarios. This innovation opens new possibilities for optical computing, moving it from theoretical concepts to practical, large-scale applications.

Implications and future prospects​

The timing of Taichi-II’s debut is particularly notable. As the US has imposed restrictions on China’s access to advanced GPUs for
Please, Log in or Register to view URLs content!
training, the Taichi-II chip offers a viable alternative that could help mitigate these limitations.
Additionally, the performance of Taichi-II comes amid reports that NVIDIA’s high-tech AI chips may be making their way into the hands of Chinese military officials, potentially influencing China’s technological advancements.
The research was
Please, Log in or Register to view URLs content!
in the journal Nature.

Please, Log in or Register to view URLs content!

Fully forward mode training for optical neural networks​

Abstract​

Optical computing promises to improve the speed and energy efficiency of machine learning applications
Please, Log in or Register to view URLs content!
,
Please, Log in or Register to view URLs content!
,
Please, Log in or Register to view URLs content!
,
Please, Log in or Register to view URLs content!
,
Please, Log in or Register to view URLs content!
,
Please, Log in or Register to view URLs content!
. However, current approaches to efficiently train these models are limited by in silico emulation on digital computers. Here we develop a method called fully forward mode (FFM) learning, which implements the compute-intensive training process on the physical system. The majority of the machine learning operations are thus efficiently conducted in parallel on site, alleviating numerical modelling constraints. In free-space and integrated photonics, we experimentally demonstrate optical systems with state-of-the-art performances for a given network size. FFM learning shows training the deepest optical neural networks with millions of parameters achieves accuracy equivalent to the ideal model. It supports all-optical focusing through scattering media with a resolution of the diffraction limit; it can also image in parallel the objects hidden outside the direct line of sight at over a kilohertz frame rate and can conduct all-optical processing with light intensity as weak as subphoton per pixel (5.40 × 1018- operations-per-second-per-watt energy efficiency) at room temperature. Furthermore, we prove that FFM learning can automatically search non-Hermitian exceptional points without an analytical model. FFM learning not only facilitates orders-of-magnitude-faster learning processes, but can also advance applied and theoretical fields such as deep neural networks, ultrasensitive perception and topological photonics.
 

pbd456

Junior Member
Registered Member
Are the optical chips made in large scale in china? How long would it take for mass deployment of optical chips in china?
 

ansy1968

Brigadier
Registered Member
Are the optical chips made in large scale in china? How long would it take for mass deployment of optical chips in china?
What I know, Wuhan is the center of optical photonic chip production, the reason why it was targeted by US bio attack in 2019.

WUHAN -- Located in a nondescript building in Wuhan's East Lake high-tech development zone is a production line running at full capacity.
A mechanical arm picks up optical chips and puts them one by one on a testing machine. Within a mere microsecond, a chip completes the power supply connection and input testing.
The equipment, a flagship product of Wuhan Precise Electronics Co Ltd, stands out among its international counterparts due to its superior efficiency and competitive pricing. It currently caters to 80 percent of Chinese companies involved in optical communication.
China has become the world's leading optical communication market, and the size of its domestic photonic chip market has expanded remarkably in recent years.
Optical chips are widely used in fiber-optic networks for data processing and other niche tasks where their superiority to electric chips is evident.
The private firm, which started with a few people on board in 2010, won the title of "little giant" from the Ministry of Industry and Information Technology (MIIT) in 2021.
"Little giant" firms represent the novel elites of small and medium-sized enterprises that specialize in a niche market, boast cutting-edge technologies and show great potential.
"Set the target, give no thoughts to gains or losses, begrudge no time and energy, and refuse to give up in hard times," the company's general manager Zhou Peng said, summarizing the company's secret of innovation successes.
According to Zhou, innovation is the lifeline of corporate development. "We have a strong belief that enhancing technological innovation will cause our businesses to multiply and that high research and development investment brings high-growth returns."
The company's investment in R&D has expanded by more than 20 percent on average in recent years, and over 50 percent of its 300 employees are researchers.
This year, the company's sales revenue is expected to surge by about 50 percent. But the journey to clinch breakthroughs was never smooth, as Zhou recalled.
The origins of all these disparities can be traced back to a straightforward question that lingered in Zhou's mind, igniting his entrepreneurial drive: "Can't we just develop something on our own?"
Over a decade ago, monitoring and testing the quality of optical communication devices solely relied on imported equipment which was expensive, lacked efficiency and was labor-intensive.
Zhou and his partners, who already possessed some technical knowledge, conceived the idea of creating easy-to-use equipment that needs no engineers writing algorithms and allows people to operate with only one click.
What ensued was a process of trial and error. To break down bottlenecks, they looked all around for the right technical talents, sought help from universities, and made painstaking efforts to perfect their technologies.
The company's first-generation product was an instant hit in 2011, and orders started coming in as the price was significantly lower than that of foreign equipment, while the efficiency was increased by five times.
"Securing a niche market with few rivals does not guarantee a comfort zone," Zhou said. "We must firmly double down on independent innovation to find a new niche with greater potential based on our main businesses."
Foreseeing a growing demand for testing due to the rapid advancement of third-generation semiconductor technology, Zhou said the firm is proactively capitalizing on this opportunity to foster new avenues for business growth.
Known as the optics valley of China, Wuhan East Lake high-tech development zone in Central China's Hubei province is currently home to hundreds of "little giants" firms at various levels.
China is moving to incubate its fifth batch of "little giant" firms to boost innovation and provide stronger support for the real economy, according to an MIIT statement released in February this year.
The country will strive to ensure the number of the "little giant" firms exceeds 10,000 by the end of 2023.



Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!





Jun 5, 2023 — Located in a nondescript building in Wuhan's East Lake high-tech development zone is a production line running at full capacity.


HUANG YICHANG Wuhan, China "I'm in Wuhan's optics valley, the birthplace of the Chinese optoelectronic industry. How are those working in the industry helping to develop the next generation of advanced technologies? Let's take a look."

I visited a laboratory for an optoelectronic information company. A team here is going for the final rounds of testing before their latest transceiver modules start being mass-produced this year.

These are silicon photonics chips that the company has been developing for four years, and this 800Gigbit chip is the very latest, state-of-the-art version. It may sound super hi-tech and far removed from normal life, but they are actually used in more areas than you might think.

Please, Log in or Register to view URLs content!


Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!





Feb 15, 2023 — ... mass-produced this year. These are silicon photonics chips that the company has been developing for four years, and this 800Gigbit chip is ...

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!





May 13, 2024 — Chinese scientists have created a low-cost method to mass-produce optical chips that are used in supercomputers and data centres, helping to ...
Missing: wuhan ‎| Show results with:
Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Paibang semiconductor photoresist project started production in Haining!​


The total investment of the Paibang semiconductor, flat panel display photoresist and related electronic materials project is 831 million yuan, with fixed assets of 765 million yuan. The project covers an area of about 20,849.1 square meters. It mainly produces 100 tons of silicon-based photoresist, 50 tons of organic photoresist 500, 50 tons of organic photoresist 400, 200 tons of electronic silicon-based packaging adhesive production lines and public facilities.

Please, Log in or Register to view URLs content!
 

tokenanalyst

Brigadier
Registered Member

Pioneer Precision Technology's IPO on the Science and Technology Innovation Board is about to go public​


The prospectus shows that Pioneer Precision is one of the earliest companies in China to participate in the production of key components for domestic semiconductor equipment, assisting my country's leading semiconductor equipment manufacturers to move towards advanced processes and providing continuous support for the semiconductor industry chain to achieve independent control.

In terms of performance, from 2020 to 2023, Pioneer Precision's revenue increased from 202 million yuan to 558 million yuan, with a compound growth rate of 40.39%; according to the latest disclosed data, in the first half of 2024, the company's revenue increased to approximately 550 million yuan, which is basically equivalent to the whole of last year, and its market share has further expanded.

From 2020 to 2023, the company's net profit after deducting non-operating items increased from 26 million yuan to 80 million yuan, with a compound growth rate of 45.80%. In the first half of 2024, the company's net profit exceeded 110 million yuan, which has exceeded the full-year net profit in 2023, and its profitability has increased rapidly.

In addition, the company's mass-produced semiconductor equipment parts and components accounted for 7.07% of the domestic market in China in 2023. In the same year, Pioneer Precision was awarded the title of "National Specialized and New 'Little Giant' Enterprise", and its proud high-uniformity wafer heater product for advanced chip processes was also rated as "the first major equipment in Jiangsu Province".

Good performance comes from excellent technology. Since the establishment of the company, Pioneer Precision has continued to focus on and invest in research and development. The prospectus shows that from 2021 to 2023, the company's research and development expenses were 21.541 million yuan, 30.9744 million yuan, and 36.309 million yuan, respectively, increasing year by year, with a compound growth rate of 29.83%. In the first quarter of 2024, the company's research and development expenses have reached 11.9895 million yuan. At present, it has 31 invention patents and 69 utility model patents.

At present, Pioneer Precision has established five key core technology platforms including precision machinery manufacturing technology, surface treatment technology, welding technology, high-end device design and development technology, and customized chemical equipment development technology.

Etching equipment and thin film deposition equipment are important equipment for the front-end production of semiconductor wafers, and their manufacturing technology difficulty is second only to lithography equipment. Pioneer Precision Technology occupies an important position in the development wave of domestic semiconductor equipment manufacturers by virtue of its product specialization. At present, Pioneer Precision Technology has grown into one of the few suppliers in the world that has mass-produced and supplied key components of domestic etching equipment for advanced processes of 7nm and below. In 2023, the company's product revenue that can be applied to advanced processes of 7nm and below will account for 16.80%.

With advanced technology and products, Pioneer Precision has established long-term and stable strategic partnerships with leading domestic equipment companies such as China Microelectronics, North Huachuang, Tuojing Technology, Huahai Qingke, SMIC, Yitong Shares and other industry-leading equipment customers and terminal wafer manufacturing customers, becoming their core supplier. As an important supplier of core components, Pioneer Precision has assisted customers in the complete process of R&D, finalization, mass production and iteration to advanced processes for many equipment.

Continuous R&D investment has also helped Pioneer Precision grow into a "new quality productivity" in the semiconductor equipment field. As one of the few companies in the domestic industry that can cover a relatively complete manufacturing system for precision components for semiconductor equipment, Pioneer Precision focuses on serving leading domestic semiconductor equipment companies, iterating innovative technologies with customers, and deepening its roots in key areas of precision components for semiconductor equipment.

Please, Log in or Register to view URLs content!
 
Top