After two years of US tech giants such as OpenAI, Google, Meta, and Amazon dominating the artificial intelligence (AI) discourse, a little-known and scrappy Chinese AI lab—DeepSeek—has entered the buzzy space.What is DeepSeek?DeepSeek is a Chinese AI startup specialising in open-source large language models (LLMs). It recently garnered significant attention following the successful launch of its LLM, DeepSeek V3, which demonstrated impressive capabilities.When launched in December 2024, DeepSeek V3 demonstrated superior performance across benchmarks compared to leading models from OpenAI, Anthropic, and Meta. This includes excelling in coding, mathematical problem-solving, and even identifying code errors. A fortnight later, the company unveiled DeepSeek R1, which caused a stir in the tech world. It showcased advancements in reasoning and problem-solving that were on par with or better than many existing models. The R1 uses large-scale reinforcement learning (RL) to process data and create responses. Its performance is comparable to OpenAI’s O1 model in areas such as mathematical ability, coding proficiency, and general knowledge comprehension. At the same time, it excels in creative writing, general question answering, editing, and summarisation tasks.Low-costWhat has shocked Silicon Valley players is that it was built for a fraction of the cost required to develop the top-performing OpenAI and Meta models.The startup claims that it developed this outperforming, low-cost model within two months and with an investment of less than $6 million ($5.58 million, to be precise). This starkly contrasts the $100 million OpenAI reportedly spent on training its GPT-4 model.The Chinese breakout AI company has used graphics processing units (GPUs) regarded as last-generation in the US (such as Nvidia's H100 AI GPUs) to train its models. A significant reason is the export controls imposed by the Joe Biden administration in October 2022 on advanced computing chips, which sought to hinder China’s advances in AI. While this may have limited Chinese companies' access to advanced GPUs, it didn’t stop them from maximising the abilities of the chips they had, as DeepSeek has shown.Besides the scrappy ‘garage mindset’ mentality that DeepSeek may embody, its sudden popularity is attributed to its affordability. Its R1 model offers developers, researchers, and organisations seeking AI solutions $0.55 per million input tokens and $2.19 per million output tokens. For comparison, OpenAI charges $15 per million input tokens and $60 per output token.DeepSeek’s ability to dramatically reduce inference costs is causing concern among US tech giants. Perplexity AI CEO Aravind Srinivas commented on X, "DeepSeek has effectively replicated O1-mini and made it open-source." Microsoft CEO Satya Nadella also remarked that DeepSeek's arrival should be taken seriously.Open source fightDeepSeek has released several models, including DeepSeek Coder, DeepSeek LLM, and DeepSeek R1. These models are free of charge for researchers and commercial users. The company is gaining popularity for its innovative approach to AI development and its commitment to open source, which Elon Musk claims OpenAI was intended to embody. This development comes amid the Altman-led company's efforts to transition to a for-profit model, which resulted in a legal dispute with the CEOs of Tesla and SpaceX.In the broader context, China is moving towards open sourcing. For instance, Alibaba Cloud has released over 100 open-source large language models (LLMs) as part of its Qwen 2.5 series. These models range from 7 billion to 72 billion parameters and support more than 29 languages.Versus USThis occurs as the US-China "chip war" escalates into a significant geopolitical battleground. Due to national security and technological dominance concerns, the Biden administration has imposed strict export controls on advanced semiconductor manufacturing equipment for China, placing Chinese tech firms at a disadvantage against their Western counterparts.There is also growing concern, particularly in the West, that Chinese LLMs are subject to government review, specifically concerning their adherence to “core socialist values," as reported by the Financial Times in July last year. The report added that regulators would test responses to politically sensitive topics such as the 1989 Tiananmen Square massacre, Taiwan’s independence, and Chinese President Xi Jinping.The High-FlyersThe company's CEO, Liang Wenfeng, founded DeepSeek in 2023 as an AI lab within his quantitative hedge fund. Wenfeng studied artificial intelligence as an undergraduate at China’s Zhejiang University before establishing High-Flyer in 2015, a hedge fund focused on algorithmic trading.In 2021, High-Flyer relied exclusively on AI for its trading operations, which interestingly accounts for its possession of Nvidia chips. In April 2023, the fund announced the formation of a separate entity dedicated to artificial general intelligence (AGI) research. This new entity would operate independently from High-Flyer's financial business, even as High-Flyer remains DeepSeek’s sole investor.For more news like this visit The Economic Times.

But the U.S. companies have one major disadvantage: secrecy. For years, Silicon Valley has operated on a closed-door model, keeping AI breakthroughs locked behind proprietary systems. DeepSeek’s decision to make R1 open source has flipped that narrative.

DeepSeek-V3 stands out because it offers performance similar to that of other leading AI models, but it is created on a much smaller budget.

Can the $500B Stargate Project secure U.S. AI dominance? This is a 21st-century moonshot the U.S. cannot afford to miss.

When Chinese quant hedge fund founder Liang Wenfeng went into AI research, he took 10,000 Nvidia chips and assembled a team of young, ambitious talent. Two years later, DeepSeek exploded on the scene.

DeepSeek's new R1 model matches or beats OpenAI's performance while being free and open-source—and it got there in a fascinating way.

Stargate is just one piece in the race for data, AI and global power. A glimpse into the future of U.S. regulation.

OpenAI is focusing on AI infrastructure with Stargate as rivals like China's DeepSeek close the gap on its AI models.

SINGAPORE—A Chinese artificial-intelligence company has Silicon Valley marveling at how its programmers nearly matched American rivals despite using inferior chips.

It was recently revealed that OpenAI secretly funded and accessed data related to the FrontierMath AI benchmark. The controversy raises questions about the legitimacy of

OpenAI, the company behind ChatGPT, has released its "Economic Blueprint" for AI to outcompete China, boost economic prosperity and benefit U.S. education.

China startup DeepSeek just released the first Open Source Reasoning Model that matched the OpenAI o1 reasoning model. OpenAI was charging $200 per