DeepSeek动摇了AI市场，挑战了美国的统治地位，并为加密货币带来了新的机会

DeepSeek的R1模型在全球范围内发送了冲击波，因为它能够匹配Openai的O1-Mini等竞争对手的性能，同时使用较少的资源和更少的资金。

DeepSeek, a Chinese AI company, has made waves with the launch of its R1 model, which has raised questions about the cost-effectiveness of AI and its implications for the tech and crypto markets.

中国人工智能公司DeepSeek随着其R1模型的推出而引起了轰动，该模型引发了有关AI的成本效益及其对技术和加密市场的影响的疑问。

DeepSeek’s success stems from its ability to build a competitive AI model without access to advanced U.S. chips. Despite U.S. sanctions banning the export of Nvidia’s top-tier H100 chips to China, DeepSeek found a way to work around this restriction by using the less powerful H800 chips instead.

DeepSeek的成功源于其在不访问Advanced US芯片的情况下建立竞争性AI模型的能力。尽管美国制裁禁止出口NVIDIA的顶级H100芯片到中国，但DeepSeek通过使用较小的H800芯片而找到了一种解决这一限制的方法。

According to Artificial Analysis, DeepSeek delivers results that outperform even some of the leading AI models from Google, Meta, and Anthropic. This has led to concerns about whether the estimated $1 trillion AI investment over the next few years will yield sufficient returns.

根据人工分析，DeepSeek提供的结果甚至超过了Google，Meta和Anthropic的一些领先的AI模型。这引起了人们对未来几年中大约1万亿美元的AI投资是否会产生足够回报的担忧。

Goldman Sachs had already warned as early as August 2024 that AI spending might be excessive, and DeepSeek’s ability to achieve such feats with minimal resources has now amplified those doubts.

高盛萨克斯（Goldman Sachs）早在2024年8月就已经警告说，AI支出可能会过高，而DeepSeek通过最少的资源实现此类壮举的能力现已扩大了这些疑问。

Adding to the hype, DeepSeek’s founder, Liang Wenfeng, a hedge fund manager who started dabbling in AI as a hobby, has taken an unorthodox approach by offering DeepSeek’s assistant and underlying code for free.

DeepSeek的创始人Liang Wenfeng是对冲基金经理Liang Wenfeng，他开始作为业余爱好涉足AI，他通过免费提供DeepSeek的助手和基本的代码来采取非正统的方法。

This move starkly contrasts with Western competitors, who rely heavily on commercialization and paywalls to recoup their high development costs. Liang’s assertion that “AI should be affordable and accessible to everyone” positions DeepSeek as a disruptor not only in technology but also in business models.

这一举动与西方竞争对手形成鲜明对比，西方竞争对手在很大程度上依靠商业化和付费墙来弥补其高发展成本。 Liang的断言“ AI应该负担得起，每个人都可以使用”职位，不仅在技术中，而且在商业模式中也是破坏者。

However, despite the hype, DeepSeek also faced a cyber-attack on its debut day, highlighting vulnerabilities in its infrastructure and raising questions about the company’s preparedness to scale.

然而，尽管大肆宣传，DeepSeek在首次亮相的日子也面临着网络攻击，突出了其基础设施中的脆弱性，并提出了有关公司对规模缩小的准备的疑问。

One of the most pointed allegations comes from speculation that DeepSeek might not be fully transparent about its hardware capabilities. Elon Musk suggested that DeepSeek “obviously” possesses approximately 50,000 Nvidia H100 chips despite U.S. export controls that restrict their sale to China.

最尖锐的指控之一来自猜测DeepSeek可能对其硬件功能不完全透明。埃隆·马斯克（Elon Musk）建议，尽管美国出口控制限制了他们向中国的销售，但DeepSeek“显然”拥有大约50,000个NVIDIA H100芯片。

Musk’s comment implies that DeepSeek may have circumvented these sanctions, either through undisclosed channels or by acquiring the chips before restrictions tightened.

马斯克的评论表明，DeepSeek可能通过未公开的渠道或通过在限制加剧之前获取芯片来规避这些制裁。

If true, this could undermine claims that the R1 model achieved its benchmarks using only the less capable H800 chips, which were explicitly designed as a downgraded alternative for the Chinese market.

如果是真的，这可能会破坏R1模型仅使用功能较低的H800芯片实现其基准，这些芯片是在中国市场上明确设计的。

Doubts also linger about the company’s reported development costs. ZaStocks, a financial commentator, highlighted the implausibility of building a large language model of DeepSeek’s calibre for just $5.6 million.

对公司报告的开发成本的怀疑也存在。财务评论员扎斯托克（Zastocks）强调了建立DeepSeek口径大型语言模型的不可思议性，仅为560万美元。

“You can’t even buy a beachfront home in California for $6,000,000, but apparently, the trustworthy CCP built a better LLM than Meta for that price,” he quipped, referencing China’s history of overpromising on technological advancements.

他打趣说：“您甚至不能以600万美元的价格在加利福尼亚购买海滨房屋，但显然，值得信赖的CCP以该价格建立了比Meta更好的LLM。”

DeepSeek.I’m not buying it. Yes, they have a great model but the cost just doesn’t add up.You can’t even buy a beach front home in California for $6,000,000 but apparently the trustworthy CCP built a better LLM than Meta for that price.China has a long history of lying…

deepseek.i不买它。是的，他们有一个很棒的模型，但成本并没有加起来。您甚至无法以600万美元的价格购买加利福尼亚的海滩前房，但显然值得信赖的CCP以该价格以比Meta建立了更好的LLM。说谎的历史…

ZaStocks also noted that U.S. tech companies like Meta and Google have historically admitted to overspending on AI to avoid falling behind. DeepSeek’s ability to match or outperform these firms with limited resources seems to defy this industry norm.

Zastocks还指出，像Meta和Google这样的美国科技公司历史上已经承认超支了AI，以避免落后。 DeepSeek匹配或​​超越这些资源有限的这些公司的能力似乎无视这一行业规范。

As ZaStocks put it, “There’s no shot they’d admit to the actual cost of this LLM because it wouldn’t add up with the limited technology they’re able to access (legally).”

正如扎斯托克（Zastock）所说，“他们没有承认该LLM的实际成本，因为它不会加起来他们能够访问的有限技术（合法）。”

The unveiling of DeepSeek’s advanced AI capabilities, promising high performance with reduced hardware requirements, also raised questions about the sustainability of GPU-dependent industries, triggering a ripple effect of uncertainty.

DeepSeek高级AI功能的揭幕，有望通过减少硬件要求的高性能，也提出了有关GPU依赖行业可持续性的问题，从而引发了不确定性的连锁反应。

Crypto mining firms bore the brunt of the fallout. Riot Platforms (RIOT), a leading Bitcoin miner, saw its shares plummet by 15.5% on Jan. 27, while Cipher Mining (CIFR), another key player in the sector, faced a sharper decline of 25%.

加密矿业公司首当其冲。领先的比特币矿工Riot Platforms（Riot）在1月27日看到其股票下降了15.5％，而该行业的另一个关键参与者Cipher Mining（CIFR）面临的幅度下降了25％。

The turbulence wasn’t confined to mining stocks. AI-focused crypto tokens, a niche but growing segment of the market, also faced a significant sell-off.

湍流不仅限于采矿股。以AI为重点的加密代币（一个利基市场），但不断增长的市场也面临着大量的抛售。

Tokens like Render (RENDER), which facilitates decentralized GPU rendering, dropped 11% over the last seven days. Similarly, The Graph (GRT), a blockchain indexing protocol, and Artificial Superintelligence Alliance (FET), which integrates AI and blockchain to create autonomous agents, declined by 15% and 14%, respectively.

像渲染（渲染）这样的代币促进了分散的GPU渲染，在过去的七天中下降了11％。同样，图形（GRT），一个区块链索引方案和人工超级智能联盟（FET），该联盟（FET）集成了AI和区块链以创建自主药物，分别下降了15％和14％。

The downturn in both crypto mining stocks and AI-focused tokens highlights their deep reliance on Nvidia’s GPUs, or graphics processing units, which are specialized chips designed for parallel processing.

加密矿业库存和以AI为中心的代币的低迷强调了它们对Nvidia的GPU或图形处理单元的深厚依赖，这些芯片是专门用于并行处理的专业芯片。

GPUs are critical for handling the intensive computational tasks required in industries like cryptocurrency mining and AI. For example

GPU对于处理加密货币采矿和AI等行业所需的密集计算任务至关重要。例如