DeepSeek Innovation and Endogenous Dilemma: Crypto Navigates in the Mist

Reprinted from panewslab
02/21/2025·2MAuthor: YBB Capital Researcher Zeke
1. AI is inevitable, but Crypto x AI?
At the beginning of 2025, DeepSeek created by Huanfang Quantitative dropped a "nuclear bomb" on the AI circle. A Chinese AI model that uses only 2,048 Nvidia H800 GPUs and has a training cost of US$5.58 million (about one-tenth of Meta) has achieved the level of arm wrestling with GPT-4o and Llama 3.1 in benchmark tests such as MMLU and GPQA. , even slightly surpasses these top large models from Silicon Valley in latitudes such as complex reasoning and Chinese semantic understanding. The United States has been blocking China for many years, but it has encountered a dramatic deconstruction on the computing power sandbox of DeepSeek. Chinese AI under the siege has finally developed a technological path that is in line with national conditions but is comparable to cutting-edge, a set of open source and low-level The combination of price and homogeneity directly penetrated the US computing power moat.
Chinese technology products that are always slightly inferior to the United States in terms of performance always give me a cheap and plagiarized stereotype. I believe this is also the intuitive perception of most people about Chinese Internet companies' products. But this time DeepSeek is indeed different. I won’t talk about whether the user experience is higher than ChatGPT, which is a strong subjective feeling. Judging from the tense reactions of the frequent mentions of this matter by American political and technology giants, China is no longer the role of a technology follower this time, and the global shock caused by this is also very strong.
Although the first thing that was affected was my wallet, to some extent, this was indeed a misjudgment on the development of traditional AI, I still want to talk about my opinion, that is, the impact of DeepSeek on Crypto.
1. Nvidia was most injured in this incident. First, the demand for AI computing power was questioned. Second, Nvidia's unified software and hardware computing architecture "CUDA" was bypassed. Friends who are familiar with the AI field should not be CUDA. Strangeness is one of the important cornerstones to promote the development of modern AI. When big model developers use NVIDIA GPUs, they generally do research and development based on CUDA. Using CUDA has low requirements for developers, because some functions have been encapsulated in CUDA, and there is no need to pay attention to too many details when using it, but it will definitely lose execution efficiency.
Since CUDA is a general-purpose programming framework, it will result in some flexibility being lost when training the model. DeepSeek's approach is to directly use PTX (the intermediate instruction set framework designed by Nvidia for GPUs) to bypass the hardware's limitations on training speed and shorten the training time. The training time for other models takes 10 days, and DeepSeek can be completed in 5 days. This also means that if DeepSeek intends to adapt to Chinese GPUs in the future and will be more handy in hardware adaptation, Nvidia's AI chip throne will also be shaken. (This paragraph comes from the report of South Korea Future Asset Securities on the DeepSeek training session)
In addition to the fact that the decline in stock prices will have a strong impact on the currency circle that has strong correlation with the US stock market, I personally think that in the long run, it is a good thing for decentralized computing power projects. First of all, more personal GPUs can also exert their remaining energy in the future. Secondly, if DeepSeek's small and beautiful open source model path can be successful, it will force a large number of AI companies to be forced to open source. The computing power required for local deployment and secondary development in the future It will become more and more vigorous. Judging from the hardware requirements of the minimum parameter 1.5B to the maximum parameter 70B of DeepSeek R1, from the lowest NVIDIA GeForce GTX 1660 Super to the 40 Series and 50 Series, to the professional-grade A100 and H800 GPUs, they have contributed again. Opportunity for extra computing power. For decentralized computing power projects with relatively transparent presence and a bit useless computing power supply, it may be an opportunity to turn things around (of course, if the delay can be low enough).
2. The AI framework project is the hottest emerging track in Crypto before DeepSeek dropped the "nuclear bomb" and the last track I wrote before the Spring Festival. Now under the bombing of DeepSeek, almost everyone is moving closer to zero. After all, others can compete with OpenAI for less than 6 million yuan, and our leading projects with more than one billion yuan in FDV have not yet produced some of them. It can be called a practical AI Agent.
We have been pursuing assetization almost paranoid since the inscription era. At present, the currency circle is quite open to assetization. AI framework projects that are not listed at all require an open source Github code base and a social account to The price of issuing coins is that you must be able to accept the dimensionality reduction of the two-way foil thrown by traditional AI companies one day.
In the golden age of AI development, traditional Internet companies can use the only killer moves this time. The development of AI between the Sino-US game will only get faster and faster. Crypto needs to be in the upstream and downstream of AI. The key to the problem is to combine it with which directions can highlight the advantages of decentralization and not be beaten to death by an inexplicable AOE. The technology stack of Crypto x AI can be divided into computing power layer, data layer, middleware layer, and application layer. In these layers today, I have never felt the necessity of Crypto. If we consider privacy and security from a future perspective, it may be a good perspective. After all, AI Agent has become a reality today. How to ensure privacy of work information and private information processed may be a problem that traditional Internet companies cannot solve. Further, if an AI agent can have the permission to pay, how to ensure the security of the wallet will also become a problem. Taking blockchain as the compliance audit layer of the AI model should be our main development direction in the future.
On the other hand, it is to motivate in that aspect. In addition to stimulating the power layer, model sharing, etc., it can actually teach humans how to interact with the virtual world. Unlike LLM training, it naturally has the whole world for decades. The Internet corpus knowledge teaches AI how to act correctly actually requires humans to constantly mark the correct position. Just like teaching visual models to identify what animals are and what cars are, this is not something that can be accomplished by outsourcing companies to find a group of college students. The creation of an AI agent that can interact with the virtual world requires a complex decentralized network. Teaching AI through countless human individuals is also a direction. This is also discussed in more detail in my past articles, and what else can be done around motivation? Combined with DePin, teach the agent to interact with reality, motivate AI to gain attention, motivate secondary innovations about AI (for example, the incentive method of Bittensor is a good example), and the token incentive mechanism is adjusted automatically by AI (about this point derived from the self in A question in the previous article, the original words are as follows: When a decentralized project becomes extremely large and enters the mainstream, what should we do in deflation and inflation? Should we rely on simple codes, or do we follow only a few people or A project party with dozens of people? Or those soul figures? Oh, by the way, we still have governance tokens. But governance tokens are meaningless before solving the witch problem. Democratic voting can never be reflected in governance proposals, after all a16z can reject a large community's vote of favor with just a few wallets, so what's the point of voting? Wait.
We really cannot gather a group of high-end AI talents like traditional Internet companies, buy or rent a large-scale GPU cluster for training, and it is naturally a dream to create a DeepSeek in the blockchain. The meaning of Crypto's existence is to give another field an irreplaceable decentralized feature, just as we have given financial freedom. AI is an inevitable narrative for humans, but what role can Crypto play in it?
3. This is the first time I have talked about Wordcoin in my article. This crypto utopia project initiated by Sam Altman seems a bit nonsensical even now. Whether to enter the iris means you need to make a choice between national monitoring and enterprise monitoring, which is like being in the Matrix to choose to take a red or blue pill.
However, the concept of national income or inclusive finance may no longer be a joke at this stage. DeepSeek, which can locally deploy cutting-edge models, has allowed agents to appear in hospitals and government units in China. According to McKinsey's 2024 forecast report, 50% of jobs will be replaced by AI as soon as the next six years. In the future, the government may even issue a unified distribution of Wordcoin. If this process continues to intensify, the tokens related to the concept of inclusive finance may appear in large quantities and be repeatedly hyped. According to the five- or six-year window period, If it matches Trump's term, will the crypto president issue a similar token? I think it's very likely.
4. According to Elon Musk's recent remarks, AI may win the Nobel Prize for 25 years. So I think it is more interesting and effective than DeSci now, and maybe I can call it Decentralized AI Science? That is, DeAIS.
2. Meme Coin is no longer Meme
We once analyzed that Meme Coin was talking about subculture, community consensus, and communication effects. Now, I am sitting in front of GMGN, and every day I analyze the cabal, the front address, and the Dev's rat warehouse. When a CA is sent to various local dog groups, it is the moment of charge. Today's Meme is more outrageous than any other time in the past. In the current Pump.fun, let alone find a token that can sleep peacefully, you may Just go to the toilet and the K-line fell off the cliff.
The continuous simplification of the threshold for asset issuance and the high anonymity of blockchain have prompted this crazy casino culture to become more and more intense. An unknown grass-roots team can use the currency circle as an ATM. The evolution direction of Meme is becoming more and more casual, not only the above mentioned coins, but any event, anyone or even any AI can convert coins. Without a cultural core and without consensus cohesion, the so-called leader of the founding sect can only be completely forgotten in a few weeks. The celebrity coins trend that started with Trump only lasted for a month. Under a Twitter Twitter of President Mile, hundreds of millions of funds flowed out of the circle again from Sol, and Meme began to recede. President Mile's reaction to this was also very simple. He deleted Twitter and replied to everyone: I don't know.
The rapid development of AI has taken away too much attention from the world, and the development of the technical school is difficult to move forward. Retail investors who give up value investment can only bet on the pig-killing market that they are the lucky ones. The increasingly scarce liquidity is repeatedly extracted, and what is reflected in reality is the red K-lines day after day within Cex and Dex, as well as the disdain of traditional capital and outsiders in the circle for copycats.
3. Do not ask for a sword when carving a ship
The law of cycles has obviously failed, and it is difficult to find the sword when carving a boat to find a sword. BTC’s bullish do not mean that it’s bullish do, but BTC’s bullish do must have bears when it’s bullish do. Our understanding of counterfeits must be innovative. The counterfeit market is no longer a market that can be supported by a white paper. A large-scale project launched in the top Cex must be mature enough to support its own currency price.
Looking back at the growth of tokens over the past seven years, there were less than 2,000 tokens listed in the 2017 market, while nearly 25,000 tokens listed in 2024 (data originated from CoinGeko, including tokens that have been removed from the shelves. ), the exponential expansion of the number of tokens is essentially the irreversible evolution of the blockchain low-entropy value system to a high-entropy noise system. When each token still carries the ideal of "subverting the world" in 2017, the token in 2024 has evolved into a bargaining chip for liquidity exit. The emergence of more tokens has not brought more innovation and implementation, but the high valuation of celebrity projects has doubled the demand for market liquidity.
As mentioned above, if you do not gain recognition from the outside world, retail investors cannot support the valuation of these projects. Most copycat projects are often listed at the highest point in history, while Binance is the last stop. The currency circle needs an innovation, and celebrity projects should be worthy of their huge financing. Public disclosure of project financial reports piloted by Bybit may be a solution. But in my personal foolish opinion, the market needs to reshape the valuation system and listing standards of today's counterfeit projects.
4. Confused
I've seen a ray of light on Ton, thinking that the beginning of Crypto consumer-grade applications has arrived, but this brief light disappears as the Tap to Earn tide fades. The liquidity mining derived from DeFi five years ago made the currency circle experience a peak of great glory. Five years later, the only area we succeeded is DeFi.
The topic of communication with people in the circle is very simple now. Have you bought BTC? Shorted? Give me a CA. Everyone is confused and we can no longer find the right direction. Apart from BTC, it is the current situation in the currency circle that you can’t sleep well if you buy any token. Diamond Hands is not a compliment today. If you don't buy BTC, it's more like a synonym for fools.
And when I opened the various link media on my phone, I seemed to be watching the New York Times and gossip media. All the phenomena reflected that most of the hopes in this circle have been placed on policies and attention. From the perspective of a VC, we may only invest in tool products in the future, and the asset issuance platform will become a seller and rent collector in order to survive.
Conclusion
Obviously this is not the status quo that we are willing to see. Although Crypto seems to be in the fog of lost direction at present, DeepSeek's success proves to us that reforming the past and making new things is still the most effective way to break through the dilemma of failure. Crypto has the best policy, attention, capital volume and solid infrastructure in history. In the near future, there are still many counterfeit ETFs that can inject liquidity into the currency circle again. We are clearly in the mainstream, but we are trapped under our own besiege. There may be a turning point behind the ebb of Meme Coin today, and there may not be only AI in the future of mankind.