What does the re -disposal of Deepseek for the future of Crypto can create its effective approach to the decentralization AI applications and the ecosystems that depend on GPU?
Dibsic gets all attention
On January 27, the global technology industry faced an earthquake when Deepseek, an Amnesty International, launched the R1 model, which raises excitement Reactions This shook the investor's confidence and the heart of more than one trillion dollars from the Nasdak compound index.
The repercussions struck some of the largest names in technology, with NVIDIA an experience The largest market loss for one day in the history of the United States-the amazing value of $ 600 billion. Google, Alphabet, lost $ 100 billion, while Microsoft got $ 7 billion.
Deepseek's launch disrupted the narration of American hegemony in artificial intelligence, and difficult perceptions that modern Amnesty International models can only be developed with huge investments and more advanced chips.
Send the DEPSEK R1 to the world because of its ability to match the performance of competitors such as Openai 'S O1-MINI while using less resources and less less.
According to the company, the R1 was developed at a cost of only $ 5.6 million, and it is a blatant contrast with estimates of similar models from American companies, which ranged between $ 100 million and 1 billion dollars.
Even the CEO of Openai, Sam -German called the work of "activation", praising his ability to provide impressive results for the price.
So why did Dibsic become the focus of global interest? How does this disrupt not only of artificial intelligence but also the markets of technology and wider encryption? Let's take a closer look at the noise and the differences that make Deepseek the name of the world cannot ignore it.
Noise
Deepseek's success lies in breaking the barriers that have long been considered unparalleled. The company's ability to build a competitive model for Amnesty International without reaching the United States' advanced chips has transformed heads and raised questions about the necessity of huge artificial intelligence investments.
US sanctions banned the export of H100 first -class chips in NVIDIA to China, but Deepseek found a way to overcome this restriction, using the least powerful H800 chips instead.
According to artificial analysis, Debsic delivers results This exceeds even some of the leading artificial intelligence models of Google, Meta and Anthropic.
Analysts have already started expressing concerns about whether it is estimated AI investment of $ 1 trillion over the next few years will lead to adequate returns.
Goldman Sachs warned early in August 2024 that artificial intelligence spending could be excessive, and that Debsik's ability to achieve such exploits with the lowest resources have swamped these doubts.
What distinguishes Deepseek is not only his competence but his philosophy. The founder Liang Winfing, the hedge fund manager who began to get rid of artificial intelligence as a hobby, followed an unconventional approach by offering a Depsic assistant and the basic code for free.
Liang's step flaws contrasts with Western competitors, who depend heavily on marketing and Paywalls to recover high development costs.
Liang emphasizes that "artificial intelligence should be accessible to everyone and are available to everyone" puts Deepseek as a formed not only in technology but also in business models.
Despite the noise, the cybersecurity that he faced on its first day highlighted the weaknesses in its infrastructure, which raised questions about the company's willingness to expand.
Dispieces and allegations surrounding Depsic
While the first time was praised in Dibsic as a technological teacher, it did not escape from the storm of doubt and controversy.
One of the most indicative allegations comes from speculation that Deepseek may not be completely transparent about the capabilities of its devices.
Elon Musk suggested that Dibsic "clearly" has approximately 50,000 NVIDIA H100 chips despite the US export controls that are restricted to China.
Musk's comment indicates that Deepseek may circumvent these sanctions, either through unpopular channels or by obtaining chips before tightening restrictions.
If this is true, these allegations may be undermined that the R1 model has achieved its criteria using the lowest H800 chips, which are explicitly designed as an alternative to a reduction in the Chinese market.
Doubts also remain about the costs of the company reported to the company. The most prominent Zastocks, the financial commentator, the ability to build a large Deepseek language model for only $ 5.6 million.
"You can not even buy a home beach house in California for $ 6,000,000, but apparently, the CCP is worthy of LLM better than Meta for this price," he said, noting China's excessive history of technological developments.
Zastocks also indicated that American technology companies such as Meta and Google have historically recognized artificial intelligence to avoid knees. Deepseek's ability to match or outperform these companies with limited resources challenge the base of this industry.
"There is no snapshot admitting the actual cost of this LLM because it will not add with the limited technology they can reach (legally)."
Mining shares, artificial intelligence symbols, and beyond
The final tremors of the destructive Deepseek appeared on technology shares such as NVIDIA; It resonated through the encryption markets, especially those that affect GPU -based mining companies and codes centered on artificial intelligence.
Deepseek's advanced possibilities of artificial intelligence, high performance with low devices requirements, and raised questions about the sustainability of GPU industries, which leads to an impact of uncertainty.
Crypting mining companies bears the implications of repercussions. Riot control platforms (RIOT), a pioneering Bitcoin mine, saw his shares decreased by 15.5 % on January 27, while CIFR (CIFR), another main player in the sector, faced a 25 % more severe decrease.
The disorder was not limited to mining shares. The codes that focus on artificial intelligence, a suitable but increasing segment of the market, faced a large sale.
Symbols like the offer (MakeWhich facilitates the display of decentralized graphics processing unit, has decreased by 11 % over the past seven days. Likewise, the graph)GRT), Blockchain indexing protocol, and the artificial SuperTelligence Alliance (Fet), Which integrates artificial intelligence and Blockchain to create independent factors, decreased by 15 % and 14 %, respectively.
The decline in both coding and symbols that focus on artificial intelligence highlight their deep dependence on graphics processing units in NVIDIA, or graphics processing units, which are specialized chips designed for parallel treatment.
The graphics processing units alone are very important to deal with the intense calculations required in industries such as cryptocurrencies and AI. For example, NVIDIA GPUs uses to solve complex mathematical problems that achieve the health of Blockchain transactions, and gain rewards in this process.
Likewise, artificial intelligence projects such as the processing process of processing graphics processing units for decentralized display salads, allowing the creation of 3D images, while Fetch.AI depends on GPU to train artificial intelligence models and operate independent factors.
The Dibsic promise to achieve advanced artificial intelligence with a decrease in the efficiency of devices has caused doubts about the long -term cost and sustainability structures of GPU.
By demanding better results with fewer devices resources, Deepseek indirectly threatens the high -cost GPU infrastructure on which these industries depend.
However, it appears that the sales drive by panic is a more reaction than the knee than the basic transformation. the AI-Crypto SectorAlthough Dynamic, only 1 % of the rating of the broader coding market is worth $ 3.65 trillion, which reduces its total effect.
In addition, early installation signs have appeared during the past 24 hours, as minor recovers were observed through many AI's symbols.
Moreover, the wider overall factors may be surprising. On January 23, President Trump issued an executive order establishing the presidential working group in the digital asset markets, indicating a long -term student batch of unified federal encryption regulations.
It also clearly prevents federal agencies from follow -up Digital currencies of the Central BankIt is a step in line with the commitment of the decentralized encryption industry.
More optimism stems from the administration's pledge of up to $ 500 billion in joint artificial intelligence research, a commitment that can indirectly enhance the AI-Crypto projects on the mathematical infrastructure.
Incions in databases, power stations and advanced devices are closely with the needs of the GPU intense AI-Crypto initiatives, which are likely to compensate for the short-term opposite winds caused by Deepseek sabotage.
What is the next encryption?
Deepseek's success in providing advanced artificial intelligence capabilities with a simple part of the cost of confidence in American domination of artificial intelligence.
However, Edgy, a prominent encryption analyst, indicates that this disorder can already be optimistic about the encryption market, especially the AI-Crypto sector.
Despite the short term chaos, Edgy argues that the nature of Deepseek is open source and reduce the costs of inference can serve as a blessing for AI-Crypto symbols.
By reducing the cost of running Blockchain applications that operate in Amnesty International, Deepseek carries the ability to make encryption projects more sustainable and profitable in the long run.
For example, frames like AI16Z, which have already integrated Deepseek technology, reflects how these developments can reduce the financial barriers that prevent the spread of AI solutions into the ecosystem for encryption.
While Deepseek has sparked short -term concerns, its ability to increase the cost and increase innovation in the encryption space indicates that this disorder can finally benefit the industry.
Source link