China’s DeepSeek Is America’s AI Sputnik Moment

there is a strong evidence the chinese team has taken data illegally from the AI team in the US. that’s standard industrial espionage. that accelerates the software development.
the necessary hardware and chips come from Singapore so the investigation by the FBI.
in the months, years leading to deepseek lots of high performance chips are imported into Singapore.
that’s very suspicious.

I installed deepseek on my iphone.
here what I get from deepseek. I do a quick Q&A. confirmed. all hardware are made by US companies.

“who has developed deepseek? who are the developers?”

DeepSeek is developed by DeepSeek Artificial Intelligence Co., Ltd., a Chinese company dedicated to making AGI a reality. The specific developers or team members are not publicly disclosed, but the company focuses on advancing AI technologies and applications. For more details, you can visit their official website: [https://deepseek.com]

“what hardware is used by deepseek?”

DeepSeek, like other advanced AI systems, relies on high-performance hardware to train and run its models efficiently. While specific hardware details aren't publicly disclosed, typical setups for AI development and deployment include:

1. **GPUs (Graphics Processing Units)**:
- NVIDIA GPUs (e.g., A100, V100, H100) are widely used for deep learning due to their parallel processing capabilities and support for frameworks like CUDA and cuDNN.

2. **TPUs (Tensor Processing Units)**:
- Google's TPUs are specialized hardware for accelerating machine learning workloads, particularly for large-scale models.

3. **CPUs (Central Processing Units)**:
- High-performance CPUs (e.g., Intel Xeon, AMD EPYC) are used for preprocessing, inference, and tasks that don't require GPU acceleration.

4. **Distributed Computing**:
- Large-scale AI training often involves clusters of GPUs/TPUs connected via high-speed networks (e.g., InfiniBand) for distributed training.

5. **Storage**:
- Fast storage solutions (e.g., NVMe SSDs) are used to handle large datasets and model checkpoints efficiently.

6. **Cloud Infrastructure**:
- Many AI companies, including DeepSeek, leverage cloud platforms (e.g., AWS, Google Cloud, Azure) for scalable computing resources.

If you're interested in DeepSeek's specific infrastructure, you might find more details in their technical documentation or by reaching out to their team.
微信图片_20250131211510.png
 

Tech war: Huawei cloud unit powers DeepSeek models in China’s bid for AI autonomy​

Huawei’s cloud unit teamed up with Beijing-based AI infrastructure start-up SiliconFlow to make the models available to end users​

Reading Time:2 minutes

The cloud-computing unit of Huawei Technologies has worked overtime with a local company during the Lunar New Year holidays to make DeepSeek’s artificial intelligence (AI) models available to end users on its platform in an efficient and affordable way.

Huawei’s cloud unit teamed up with Beijing-based AI infrastructure start-up SiliconFlow to make DeepSeek’s large language model V3 and reasoning model R1 available to end users through the telecom giant’s Ascend cloud service, with performance matching “DeepSeek models run on global premium graphic processing units [GPUs]”, Huawei Cloud said in a statement on Saturday.

According to a separate statement from SiliconFlow, which hosts the DeepSeek models, the charge for access to V3 on its platform is discounted to 1 yuan (US$0.13) for 1 million input tokens and 2 yuan for 1 million output tokens, while the same charges for the R1 model are 4 yuan and 16 yuan.


The move by Huawei and SiliconFlow comes as the open-source DeepSeek models are gaining popularity in both China and the US. OpenAI investor Microsoft earlier this week launched support for R1 on its Azure cloud-computing platform and GitHub. This allows clients to build AI applications that run locally on Copilot+ personal computers. E-commerce giant Amazon.com has enabled developers to create applications with the “powerful, cost-efficient” R1 through Amazon Web Services.
Huawei’s Ascend cloud service used its home-grown Ascend solution for compute resources, which could involve various types of hardware including Huawei’s self-developed server clusters, AI modules and accelerator cards, a company website said.

Huawei and SiliconFlow did not specify the chips used in the Ascend cloud service.

The move to launch DeepSeek’s models on a home-grown hardware backbone highlights China’s progress in cutting dependency on foreign technology and bolstering its domestic AI industry amid growing efforts by the US to choke off China’s access to high-end chips that the US government said could be used to advance military aims.

Huawei and SiliconFlow did not immediately respond to a request for comments during the Lunar New Year holiday.

Advertisement
AD

Hangzhou-based start-up DeepSeek has sent shock waves across Silicon Valley and Wall Street for its claimed ability to develop AI models at a fraction of the cost of its larger US competitors.

SiliconFlow, which also provides access to other models including Meta’s Llama and Alibaba’s Qwen, said in a statement that DeepSeek’s models “are a great gift from the DeepSeek team to all mankind”. Alibaba owns the South China Morning Post.

Chinese social media and video gaming giant Tencent also adopted DeepSeek’s reasoning model R1 on its cloud-computing platform, touting a three-minute easy set-up process.


Advertisement

If this is true... the end of USA AI monopoly.

USA restriction on China is like a double-edged sword.

USA doesn't want China to develop AI, but at the same time USA wants China to keep using USA AI ecosystem.

Everything is so wrong.
 
If this is true... the end of USA AI monopoly.

USA restriction on China is like a double-edged sword.

USA doesn't want China to develop AI, but at the same time USA wants China to keep using USA AI ecosystem.

Everything is so wrong.


Pretty much game over for US as Huawei Ascend 910C is already out(H100 performance) and Huawei is developing Ascend 910D(B100 performance).

US companies like Nvidia and OpenAI will have massive revaluations as they are still quite a lot overvalued.

Nvidia should never have surpassed 1 trillion US dollars in stock value.
 
Pretty much game over for US as Huawei Ascend 910C is already out(H100 performance) and Huawei is developing Ascend 910D(B100 performance).

US companies like Nvidia and OpenAI will have massive revaluations as they are still quite a lot overvalued.

Nvidia should never have surpassed 1 trillion US dollars in stock value.

Valuations is always based on future revenue and profit.

Given China will be a competitor for at least 95% of their product - it will hit profits.

China has demonstrated they can flood the market and destroy competition.
 
Pretty much game over for US as Huawei Ascend 910C is already out(H100 performance) and Huawei is developing Ascend 910D(B100 performance).

US companies like Nvidia and OpenAI will have massive revaluations as they are still quite a lot overvalued.

Nvidia should never have surpassed 1 trillion US dollars in stock value.

DeepSeek will expedite the development of the AI model for the world.

DeepSeek can completely bypass those expensive hardwares produced by Nvidia and other US companies.

It is going to break the US monopoly nor China wants a duopoly with the US.

In order to suppress the multiple new competitors not just from China, they have to ban the sale of Nvidia chips for other countries as well. And they are going to kill their own companies by doing so.

It is completely a checkmate move from China to prick their AI bubble, and they will have no chance to keep inflating their GDP in the coming years.
 
DeepSeek will expedite the development of the AI model for the world.

DeepSeek can completely bypass those expensive hardwares produced by Nvidia and other US companies.

It is going to break the US monopoly nor China wants a duopoly with the US.

In order to suppress the multiple new competitors not just from China, they have to ban the sale of NVIDIA chips for other countries as well. And they are going to kill their own companies by doing so.

It is completely checkmate move from China to prick their AI bubble, and they will have no chance to inflate their GDP in the coming years.


The US keeps making one mistake after another.

If they did not ban H800 and H1000 to China, Ascend 910C may not have existed yet.
 
The US keeps making one mistake after another.

If they did not ban H800 and H1000 to China, Ascend 910C may not have existed yet.

They have really overestimated themselves, and they tended to believe that their technology is at alien level and no other humans on Earth is capable of producing it.

The American exceptionalism has really intoxicated their mind and judgment.

In fact, once China made the breakthrough on the 7nm in 2023 with maximum sanction from the Biden admin should already be regarded as a wake up call.

Yet they keep ignoring these warnings and doubling down their mistakes.
 
They have really overestimated themselves, and they tended to believe that their technology is at alien level and no other humans on Earth is capable of producing it.

The American exceptionalism has really intoxicated their mind and judgment.

In fact, once China made the breakthrough on the 7nm in 2023 with maximum sanction from the Biden admin should already be regarded as a wake up call.

Yet they keep ignoring these warnings and doubling down their mistakes.

Ascend 910C not only is cheaper but more energy efficient than H1000.

Ascend 910B went up to 50% yield which is not bad for such a large die. SMIC is likely to improve on this for 910C.

Trump will make US decline even faster as he is the retard of all retards.
 
Ascend 910C not only is cheaper but more energy efficient than H1000.

Ascend 910B went up to 50% yield which is not bad for such a large die. SMIC is likely to improve on this for 910C.

Trump will make US decline even faster as he is the retard of all retards.

Yes, all those anti-China haters believe that the Biden admin was too soft on China because they couldn't make China collapsed in the last 4 years. So they believe that Trump could finish the job.

With the grievance and pure destructive hatred from these losers, they are going to kill their beloved USA.

The body of Uncle Sam couldn't handle anymore incoming collision with China and it is surely going to collapse.
 
They have really overestimated themselves, and they tended to believe that their technology is at alien level and no other humans on Earth is capable of producing it.

The American exceptionalism has really intoxicated their mind and judgment.

In fact, once China made the breakthrough on the 7nm in 2023 with maximum sanction from the Biden admin should already be regarded as a wake up call.

Yet they keep ignoring these warnings and doubling down their mistakes.

USA is in panic mode.

They know they can't fight against China in fair competition.

Remember DeepSeek attacked a couple days ago.

Who can mobilize a massive attack equal to European traffic capacity for a week in a single day if it's not done by CIA.

USA is desperate, they are playing dirty right now.
 
Pretty much game over for US as Huawei Ascend 910C is already out(H100 performance) and Huawei is developing Ascend 910D(B100 performance).

US companies like Nvidia and OpenAI will have massive revaluations as they are still quite a lot overvalued.

Nvidia should never have surpassed 1 trillion US dollars in stock value.

Huawei’s attempts to fill the Nvidia-sized gap in China’s AI market are being hindered by performance issues and slower connectivity speeds with its hardware alternative.

Huawei has promoted its Ascend AI chips as a potential alternative to Nvidia's hardware, as the U.S. company is prohibited from exporting its GPUs to China under export controls by the Biden administration.

The FT reports, however, that Chinese AI firms using the Ascend chips have complained over the hardware’s performance.

The Ascend line of chips reportedly suffers from stability issues, a key issue as developers building AI models require hardware to run for extended periods during training efforts.

Users have also reported Ascend suffers from slower inter-chip connectivity.

Inter-chip connectivity enables developers to combine hardware into clusters that power inference and training workloads for AI models. Slower connectivity would mean training processes take far longer.

In addition to hardware troubles, users have reportedly expressed concerns over the related software, dubbed Cann.

Cann is meant to be Huawei’s equivalent of Nvidia’s Cuda software platform which helps developers intuitively during data processing.

However, the FT reports that even Huawei’s staff have expressed concern over Cann’s performance, suggesting it makes Ascend hardware “difficult and unstable to use.”

Another engineer with reported knowledge of Baidu’s use of the Ascend chips told the FT that the hardware crashed frequently.

Huawei had touted its latest AI chip, the Ascend 910B, as being comparable to Nvidia's A100 GPU in terms of computing power and performance.

However, the A100 has since been replaced by the H100, which itself will be usurped by Nvidia’s Blackwell series in 2025, after the company fixed design flaws that had delayed production.

Zhang Ping'an, the CEO of Huawei Cloud Computing Technologies, said in July at the World Artificial Intelligence Conference in Shanghai that China can still be a leader in AI despite restrictions on its access to top-level hardware.

Demand for the company’s Ascend chips to power that leadership role has been strong, with Huawei reporting a 34% increase in revenues in the first half of 2024.

However, the company is reportedly struggling to meet customer demands. Chinese chipmakers face restrictions on importing machinery essential for chip production.

In July, the Korean news site Chosun.biz reported that a low yield rate at Semiconductor Manufacturing International Co. (SMIC), which manufactures the chips for Huawei, impacted production, with a reported four of five processes being defective.

Huawei declined to comment when contacted by Capacity.


Relative to NVIDIAs product, the Ascend series is trash. Poor performance, yield, stability, and SMIC can’t mass produce them.

And NVIDIA is moving to 3nm Rubin architecture within a year.
 
DeepSeek will expedite the development of the AI model for the world.

DeepSeek can completely bypass those expensive hardwares produced by Nvidia and other US companies.

It is going to break the US monopoly nor China wants a duopoly with the US.

In order to suppress the multiple new competitors not just from China, they have to ban the sale of Nvidia chips for other countries as well. And they are going to kill their own companies by doing so.

It is completely a checkmate move from China to prick their AI bubble, and they will have no chance to keep inflating their GDP in the coming years.

NVIDIA/TSMC will be in full production on 3nm Rubin architecture in a year.

Meanwhile China has no access to EUV machines.🤣

There’s not enough copium in the universe for this thread.
 
Pretty much game over for US as Huawei Ascend 910C is already out(H100 performance) and Huawei is developing Ascend 910D(B100 performance).

US companies like Nvidia and OpenAI will have massive revaluations as they are still quite a lot overvalued.

Nvidia should never have surpassed 1 trillion US dollars in stock value.

IMG_4547.jpeg

Huawei isn’t achieving anything near this performance
 

Users who are viewing this thread

Back
Top