Preface: DeepSeek and NVIDIA's relationship
Who is DeepSeek?
DeepSeek is a startup from China that recently developed an AI model that competes with OpenAI GPT-4 at a very low cost. DeepSeek's success is mainly attributed to its innovative AI training method, which breaks through the high cost of AI training through a combination of older NVIDIA A100 GPUs and low-cost chips.
NVIDIA’s Role
NVIDIA is a leader in the global AI hardware market, and its high-end GPUs are widely used in AI training and inference. In the past, NVIDIA was the core supplier of almost all large-scale AI projects, and companies had to rely on their expensive GPUs to train AI models.
DeepSeek and NVIDIA
DeepSeek currently still relies on NVIDIA's hardware to run AI models, but DeepSeek has proved that AI training does not necessarily require the use of NVIDIA's latest H100 or Blackwell architecture GPUs, which has also made the market begin to rethink the necessity of NVIDIA's high-end hardware.
Three key insights
1. How does DeepSeek shake up NVIDIA’s market position?
The success of DeepSeek has led the market to re-evaluate the cost structure of AI training. In the past, companies needed to invest large sums of money to purchase NVIDIA's most expensive GPUs to train AI, but DeepSeek adopted a more cost-effective strategy, using older A100 GPUs and optimized algorithms to successfully reduce the computing requirements for AI training. This has caused the market to question whether it still needs to rely on NVIDIA's high-end GPUs, or whether there are other cheaper alternatives?
Moreover, DeepSeek’s approach could inspire other AI companies to follow suit, further eroding NVIDIA’s monopoly in the AI training market. This will not only affect NVIDIA's hardware sales model, but may also change the technological development direction of the entire AI industry.
2. Is NVIDIA’s response strategy sufficient?
Faced with the challenge brought by DeepSeek, NVIDIA did not sit idly by and actively strengthened its competitive advantage through the following three methods:
- Strengthening the AI software ecosystem：NVIDIA AI Foundry and NIM Microservices(NVIDIA's AI microservices) mean that even if people who develop AI use different hardware, they still have to use their technology, making it difficult to be completely independent of NVIDIA.
- AI Foundry: Like an AI tool factory, it helps developers train and build AI models.
- NIM Microservices: A small tool that makes it easier for AI models to run in different applications.
- Deepening cooperation with technology giants：Large technology companies such as Amazon and Meta still plan to expand their AI hardware spending, and NVIDIA is also ensuring that these companies continue to use its GPUs by providing more powerful computing solutions.
- Introducing the new generation of Blackwell GPU: NVIDIA plans to launch Blackwell architecture GPUs with significantly improved computing power to ensure that it continues to lead in the computing power competition.
These strategies show that NVIDIA does not view the DeepSeek challenge as a simple threat, but as aAn opportunity to promote the evolution of the AI hardware market.
3. From an investor’s perspective, how should we view NVIDIA’s future?
Although NVIDIA's stock price has fallen in the short term due to the rise of DeepSeek, the market still maintains high confidence in NVIDIA in the long run. Two major factors support NVIDIA's long-term growth:
- Tech giants continue to expand AI investments
- Amazon plans to invest $100 billion in cloud and AI infrastructure by 2025, including a large number of NVIDIA H100 GPUs.
- Meta also increased its AI hardware spending from $37 billion to $60-65 billion and expects to purchase more NVIDIA GPUs.
- NVIDIA Remains Strong in Multiple Markets
- Gaming industry recovery: NVIDIA gaming GPU sales grow from $1.5 billion in the third quarter of 2023 to $3.2 billion in 2025, with annual revenue reaching $12 billion.
- Autonomous driving market expansion: NVIDIA's automotive division has seen significant revenue growth, and as autonomous driving technology develops, the market potential could reach billions of dollars.
- Medical and Robotic AI Applications: NVIDIA's GPU applications in the medical and robotics fields are developing rapidly, becoming a new driving force for the company's future growth.
Overall, although the emergence of DeepSeek brings challenges to NVIDIA, NVIDIA still has strong growth potential thanks to its complete ecosystem, strong partnerships and diversified market layout.
Why does the emergence of DeepSeek shake up NVIDIA's stock price?
The emergence of DeepSeek has led many people to discuss its impact on the AI industry, because DeepSeek uses a different approach than traditional methods to train AI, which may change the demand for NVIDIA hardware in AI training.
Market panic and stock price crash
When the market discovered that DeepSeek had successfully developed cutting-edge AI models at low cost, NVIDIA's stock price plummeted by 17% in just a few days, and its market value evaporated by US$552 billion. This means that investors and the market are beginning to worry whether the AI industry is shifting from a "computing power race" to an "efficiency race," and whether NVIDIA's traditional high-end hardware advantages will be weakened as a result?
How does DeepSeek train AI?
DeepSeek's AI training method is different from traditional methods. It is like a very smart kitchen that can flexibly dispatch resources according to demand instead of having all chefs work at the same time.
- Mixed of Experts (MoE) Architecture：Each chef is responsible for a specific cuisine. When a customer places an order, the system will assign the most suitable chef to handle it based on the demand, rather than involving the entire kitchen. This can save ingredients and manpower (computing resources) and improve efficiency.
- Multi-Latent Attention (MLA) Technology：It's like a professional sports analyst watching multiple games at the same time, quickly grasping the key actions, and then making the best decision, which makes DeepSeek's AI more competitive in language generation and dialogue applications.
- Efficient computing power utilization: DeepSeek primarily uses NVIDIA A100 GPUs, rather than the latest H100 or Blackwell GPUs. It's like optimizing software.Make an old phone run close to the performance of the latest flagship phone, lower cost but similar effect.
Changing rules of the AI market
In the past, companies training AI had to use NVIDIA’s most expensive GPUs because they were the only options on the market. However, DeepSeek reduces the hardware requirements for AI training through innovative algorithms, allowing companies to rethink whether they still need to rely on NVIDIA's expensive products.
For example, in the past everyone believed that only luxury super cars could provide the ultimate driving experience, but now the advancement of electric vehicle technology has allowed many consumers to discover options with higher CP value.
What does the emergence of DeepSeek mean for NVIDIA?
In recent years NVIDIA's position in the AI industry is becoming increasingly solid, because many technology companies rely on its GPUs to train AI. However, the emergence of DeepSeek may change this situation and affect NVIDIA's market advantage. Therefore, we need to first understand NVIDIA's current main products and markets, we can see the challenges brought by DeepSeek:
1. NVIDIA's current main products and markets
NVIDIA's core products are its high-end GPUs, which are mainly used in AI training, reasoning, data centers, games, and autonomous driving. Currently, NVIDIA relies mainly on its H100 and the upcoming Blackwell GPUs as core products for AI training, while its gaming market relies on the RTX 40 series GPUs. These high-performance chips have become the standard choice for technology companies and cloud service providers.
NVIDIA's largest buyers include Amazon AWS, Microsoft Azure, Google Cloud and Meta. These tech giants have invested heavily in AI training infrastructure, spending billions of dollars each year on NVIDIA’s GPUs to support AI model training, such as OpenAI’s GPT-4 and Meta’s Llama 4.
2. Potential threat of DeepSeek to NVIDIA
DeepSeek's success shows that companies may no longer need the latest and most expensive NVIDIA GPUs for AI training, which poses two major threats to NVIDIA's core business:
(1) Demand for high-end GPUs may decrease
DeepSeek leverages older A100 GPUs and innovative AI training methods to enable high-performance AI training at a lower cost. This has caused companies to wonder, do they still need expensive H100 or Blackwell GPUs, or can they accomplish the same AI training work through more cost-effective methods?
(2) Intensified competition in the AI market
The success of DeepSeek may lead more AI companies and technology enterprises to adopt similar strategies and reduce their dependence on NVIDIA's high-end GPUs. For example, if more companies choose to use older GPUs or other non-NVIDIA chips to train AI in the future, NVIDIA's monopoly in the AI training market will be challenged.
(3) The choice of technology giants is key
Currently, NVIDIA still relies on the purchasing power of technology giants such as Amazon, Meta, and Microsoft to maintain its market position. However, these companies are also actively looking for ways to reduce the cost of AI training. If DeepSeek or other competitors can offer more cost-effective solutions, these companies may reduce their reliance on NVIDIA GPUs.
It is worth noting that Amazon AWS and Google Cloud are also developing their own AI chips (such as AWS Trainium and Google TPU), which means NVIDIA may faceMore internal competition. If these cloud providers choose cheaper alternatives in the future, NVIDIA may lose some market share.
In summary, the emergence of DeepSeek did not immediately subvert NVIDIA, but it definitely made the market rethink the role of GPU in AI training. This is still a potential risk for NVIDIA and deserves close attention.
Does DeepSeek really pose a threat to NVIDIA?
A true analysis of DeepSeek's impact on NVIDIA's stock price
The rise of DeepSeek has caused the market to worry that NVIDIA's monopoly in the AI training market will be challenged, so its stock price has fluctuated sharply in the short term. However, does this mean that NVIDIA will really be replaced? We can analyze it from several levels:
1. In-depth analysis of the reasons for the stock price decline
NVIDIA's stock price fell in just a few days after DeepSeek announced its technological breakthrough. 17%, the market value evaporated more than $552 billion. This is mainly due to two major concerns in the market:
- Concerns about the decline of monopoly in the AI training market: DeepSeek demonstrates that AI training does not necessarily require the latest NVIDIA hardware, which may affect future sales of NVIDIA's high-end GPUs.
- Has the long-term investment direction of technology giants changed?: Will Amazon, Google, and Meta turn to lower-cost solutions instead of relying on NVIDIA's latest GPUs?
2. Analyze the responses of NVIDIA's major customers
Currently, NVIDIA's major buyers are the world's largest cloud service providers, including:
- Amazon AWS: Currently still planning to invest in 2025 $100 billion The company’s AI and cloud infrastructure includes a large number of NVIDIA H100 GPUs.
- Microsoft Azure: Still investing heavily in NVIDIA technology and using it as the core of our AI products.
- Google Cloud：Google is also developing its own AI chip TPU, but still relies on NVIDIA hardware in some areas.
- Meta (Facebook’s parent company): Plans to increase AI equipment spending from $37 billion Upgrade to $60-65 billion, which means it will remain NVIDIA's main customer in the short term.
These tech giants are unlikely to completely abandon NVIDIA hardware in the short term, as switching to new technologies requiresTime and high cost. However, if other competitors continue to offer more cost-effective solutions, NVIDIA may face more competitive pressure in the next few years.
3. Impact of the AI Inference Market
Currently DeepSeek mainly affects AI Training Market,but AI Inference Still dominated by NVIDIA.The demand for the inference market is much higher than that for the training marketMany companies use AI models for real-time applications, such as ChatGPT, translation tools, self-driving cars, etc. These applications are still highly dependent on NVIDIA's software and hardware ecosystem.
It's like building a skyscraper. DeepSeek has shown that cheaper materials can be used to build the infrastructure (AI training), but to make the building run properly (AI inference), you still need advanced equipment and services provided by NVIDIA.
4. Can DeepSeek really threaten NVIDIA?
- Limited impact in the short term：DeepSeek's technological breakthrough is amazing, but it is still Depends on NVIDIA hardware To run its AI models, NVIDIA's market dominance remains solid in the short term.
- Long-term challenges may increase: If DeepSeek or other competitors can further optimize algorithms in the future, or even develop solutions to replace GPUs, NVIDIA may lose some of its share in the AI training market.
Summary: NVIDIA still has strong market competitiveness
The rise of DeepSeek has indeed affected NVIDIA's stock price, but its actual threat is still limited. NVIDIA still controls the high-end AI inference market and has a strong enterprise customer base. Investors need to pay close attention to the development of the AI market, especially the future investment directions of technology giants, which will determine whether NVIDIA can continue to maintain its market leadership.
NVIDIA's response
Although NVIDIA's stock price has been affected in the short term, does this mean that it will really be replaced? After analyzing the impact of DeepSeek, we also need to understand how NVIDIA responds to such challenges to better understand future trends.
1. Strengthening the AI software ecosystem
NVIDIA is building an "Apple ecosystem" in the field of AI. Through Nvidia AI Foundry and NIM microservices, it is difficult for enterprises and developers to completely get rid of NVIDIA's technical environment even if they choose other AI chips. This is like the closed ecosystem of Apple iOS. When users get used to the seamless integration of iPhone, Mac and iCloud, it is difficult to switch to Android or Windows. In this way, NVIDIA ensures that AI developers still need to rely on its CUDA software development platform, making NVIDIA's technology an "essential standard" in the AI field. Even if the hardware market changes, NVIDIA can still maintain its own advantages.
2. Develop more cost-effective AI chips
In the past, NVIDIA's strategy was like selling "supercars", and high-performance GPUs were mainly targeted at financially strong technology giants. But now competitors such as DeepSeek are proving that "electric cars" can also run fast, making the market begin to pay attention to more cost-effective AI computing methods. Therefore, NVIDIA began to develop a new generation of low-cost AI training and inference GPUs, allowing small and medium-sized enterprises and developers to obtain powerful AI computing capabilities at a lower cost. This strategy is like Tesla's launch of the Model 3, which makes efficient electric vehicles affordable to more people, further expands market share, and prevents customers from leaving for DeepSeek or other competitors.
3. Strengthen cooperation with cloud service providers
AWS, Google Cloud, and Microsoft Azure still rely on NVIDIA's technology, but these companies are also starting to look for alternatives. Therefore, NVIDIA has recently been actively negotiating with these technology giants to ensure that they will continue to purchase NVIDIA's GPUs in large quantities in the short term. In addition, NVIDIA also emphasized the competitive advantages of its GPU in the field of AI reasoning, attempting to consolidate its dominant position in the AI reasoning market while the AI training market is affected.
Future Outlook: How this AI war will affect us
This AI hardware war is not just a competition between technology giants, it will also have an impact on investors, technology industry practitioners, and even ordinary consumers. We can focus on the following aspects:
1. What should investors pay attention to?
- Short-term volatility vs. long-term value: NVIDIA's stock price may fluctuate due to market sentiment, but there is still stable market demand in the long run. Long-term investors should focus on the long-term development of its AI hardware instead of being scared off by short-term market fluctuations!
- Procurement strategies of tech giants：If companies like Amazon and Meta continue to invest heavily in NVIDIA, it means that NVIDIA's market position remains solid. On the other hand, if these companies start turning to other solutions, it could affect NVIDIA's revenue growth.
- AI Training vs. AI Inference Market: DeepSeek mainly affects AI training, while AI reasoning is still highly dependent on NVIDIA. Investors can pay attention to NVIDIA's layout in the field of AI reasoning, such as cloud AI services, autonomous driving and other applications.
2. How should technology industry practitioners respond?
- AI Engineer & Data Scientist：This competition means that developers need to pay attention to different AI training methods and understand how to optimize AI models in different hardware environments.
- Cloud and Infrastructure Engineer: Whether NVIDIA can maintain its dominance as a cloud provider will affect the future of cloud AI computing, which will also affect IT infrastructure choices.
- Product Managers & Business Decision Makers：Enterprises need to think about how to use more cost-effective AI training and reasoning solutions, and adjust technology choices according to market changes.
3. Acceleration of AI Popularization
Whether NVIDIA is affected or not, this competition will make AI technology more popular, which may bring several key changes:
- AI training costs fall：It is easier for enterprises and individual developers to use AI technology, making more innovative applications possible.
- AI products are closer to consumers：In the future, AI may become a part of daily life like smartphones, affecting education, medical care, finance and other fields.
- More companies are entering the AI industry：When AI training becomes cheaper, more competitors will appear in the market, accelerating the innovation and application of AI technology.
This AI competition has just begun. Whether you are an investor, a technology industry player, or a general consumer, the AI revolution will have a significant impact on our future. Now is the best time to pay attention to this revolution!
Related reports
related articles
Taiwan’s first AI unicorn: What is Appier, with a market value of US$1.38 billion, doing?
What is DNS? Introduction to Domain Name System – System Design 06
Introduction to System Design Components Building Block – System Design 05
Back-of-the-envelope Back-of-the-envelope Calculation – System Design 04
Non-functional features of software design – System Design 03
Application of abstraction in system design – System Design 02
Introduction to Modern System Design - System Design 01
English 
繁體中文 
简体中文 
한국어 
日本語 
Español