DeepSeek AI—The Hedge Fund-Backed AI Model Making Big Tech Sweat

China’s DeepSeek launch of DeepSeek-R1, an MIT-licensed, fully open-source reasoning model, a strong contender to OpenAI’s o1, has tongues wagging across the media and boardrooms in what’s being called a David-Goliath story. AI has been emerging as the new battleground for global powers to show their dominance. Days after Trump announced US$500 B in building AI infra across the US generating over 100k jobs, Chinese startup DeepSeek rolled out R1 challenging OpenAI’s model. And developers are loving it. Contrasting OpenAI’s closed ecosystem, one can modify DeepSeek-R1 to one’s own comfort, be it business or research. *“We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” *— Jim Fan, senior research manager and lead of embodied AI (GEAR Lab) at NVIDIA “We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” Jim Fan, senior research manager and lead of embodied AI (GEAR Lab) at NVIDIA told AIM. “Our goal is to explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure RL process,” AIM quoted the DeepSeek team. The AI community are certainly sitting up and taking notice. *“This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” *— Bruce Keith, CO-Founder and CEO, InvestorAi Bruce Keith, CO-Founder and CEO, InvestorAi, says, “DeepSeek R1 has definitely challenged the dominance of a few players in the models and data ecosystem – OpenAI, Google, and Meta will feel it the most. R1 will have a significant impact on the AI landscape. The announcement drives home the importance of innovation and focusing on the applications and data rather than just the processing power. This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” Within days, the Chinese-built AI model has upended the industry, surpassing OpenAI’s o1, dethroning ChatGPT in the App Store, while NVIDIA’s market cap plunged by US$589 B. Unlike OpenAI’s closed ecosystem, DeepSeek-R1 is open-source, free to use, and radically efficient. It achieves state-of-the-art performance without requiring massive GPU clusters, forcing the industry to rethink the high-cost arms race in AI. Donald Trump already called it a “wake-up call” for American AI companies. It’s Cheap The Chinese AI startup has sent shockwaves through Silicon Valley, bringing into question US big tech dominance in the AI market. Built entirely on open-source technology and lower-end chips, DeepSeek sidesteps the need for high-end hardware restricted by US export controls and claims to have developed the model for just US$5.6 million. As a result, DeepSeek is available at a cost that is just 2% of what users would spend on OpenAI’s O1 model. *“DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” *— Christian Struve, Cofounder & CEO of Fracttal “DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” Christian Struve, Cofounder & CEO of Fracttal. *“DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications.” *— JD Raimondi Head of Data Science at Making Sense “DeepSeek took the initiative that Meta had taken internally: competing with the big private models with public models that can be used by everyone at low cost. But DeepSeek’s differentiator with r1 is a new one: besides being free, it is efficient, low cost, and with performance comparable to the big boys (GPT o1, GPT o4). DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications,” says JD Raimondi Head of Data Science at Making Sense. The fact that they created this platform with under US$6 M investments has shaken the tech CEOs globally highlighting that game-changing innovations don’t necessarily need billion-dollar investments. “However, from a supply/demand perspective the GPU market which Nvidia dominates is still far away from hitting peak demand,” Bruce Keith adds. Is This AI’s TikTok Moment? But DeepSeek isn’t without controversy. The application is raising questions such as data privacy, national security, and geopolitical implications. DeepSeek’s privacy policy confirms that user data is stored in China. Former Indian IT minister Rajeev Chandrasekhar, one of India’s tech savvy politicians, has already asked, if DeepSeek is the TikTok of AI. “…it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients.” — Harold Barber, COO and CFO at Escala 24×7 “DeepSeek is disrupting the AI ??market with an innovative proposition that challenges the main players in the area and could radically transform the way companies access and use this technology. However, it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients,” warns Harold Barber, COO and CFO at Escala 24×7. The US Navy has already banned DeepSeek. Still, while open-source, the Chinese version of the AI works with strict censorship, and doesn’t answer politically sensitive questions. At the same time, its open-source nature allows developers to run it locally, without restrictions, a formidable point in its favour. A David Vs. Goliath AI Story Paramdeep Singh, Co-founder of Shorthills AI, says DeepSeek changes the complete GenAI narrative. “It is like David has defeated Goliath. The old GenAI story was that only the large models could win. It required super-specialized skills, huge compute, thousands of latest GPUs, web-scale data, trillions of nodes, and huge amount of electricity to train a foundational language model. All of this translated to millions of dollars to train the model. That meant companies and countries with deep pockets were going to monopolize that market. “It is like David has defeated Goliath. The old GenAI story was that only the large models could win… Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters,” — Paramdeep Singh, Co-founder of Shorthills AI “Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters. The model is frugal and trained on spare capacity of GPUs that were lying idle. The model is so small that it can literally run on your browser. It is simply the best value for money model. We have a ray of hope where Large Language Model training and usage can be democratized. It’s not people sitting in ivory towers, but talent with frugal hardware that can train the best model. We all love this David vs Goliath story,” he says. *“The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” *— Mike Capone, CEO of Qlik The beauty of DeepSeek’s lies in its ability to aid and not just wow. As Mike Capone, CEO of Qlik, says, “The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” And the story of glory continues.DeepSeek just announced Janus Pro, an AI image model that rivals OpenAI’s DALL·E 3. If, or maybe we should say, when DeepSeek gets into video generation, competitors like Google’s Veo and OpenAI’s Sora could be in trouble next. Meanwhile, Alibaba released its Qwen 2.5 AI model it says surpasses DeepSeek. The pressure is on not just big tech or just the US, but also on smaller players and countries like India. Navanwita Bora Sachdev, Editor, The Tech Panda China’s DeepSeek launch of DeepSeek-R1 , an MIT-licensed, fully open-source reasoning model, a strong contender to OpenAI’s o1 , has tongues wagging across the media and boardrooms in what’s being called a David-Goliath story. DeepSeek-R1 OpenAI’s o1 AI has been emerging as the new battleground for global powers to show their dominance. Days after Trump announced US$500 B in building AI infra across the US generating over 100k jobs, Chinese startup DeepSeek rolled out R1 challenging OpenAI’s model. announced And developers are loving it . Contrasting OpenAI’s closed ecosystem, one can modify DeepSeek-R1 to one’s own comfort, be it business or research. developers are loving it *“We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” *— Jim Fan, senior research manager and lead of embodied AI (GEAR Lab) at NVIDIA *“We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” *— Jim Fan, senior research manager and lead of embodied AI (GEAR Lab) at NVIDIA *“We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” * — *“We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” * — “We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive—truly open, frontier research that empowers all,” Jim Fan , senior research manager and lead of embodied AI (GEAR Lab) at NVIDIA told AIM . OpenAI Jim Fan AIM “Our goal is to explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure RL process,” AIM quoted the DeepSeek team. The AI community are certainly sitting up and taking notice. *“This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” *— Bruce Keith, CO-Founder and CEO, InvestorAi *“This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” * — Bruce Keith, CO-Founder and CEO, InvestorAi *“This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” * *“This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” * Bruce Keith, CO-Founder and CEO, InvestorAi , says, “DeepSeek R1 has definitely challenged the dominance of a few players in the models and data ecosystem – OpenAI, Google, and Meta will feel it the most. R1 will have a significant impact on the AI landscape. The announcement drives home the importance of innovation and focusing on the applications and data rather than just the processing power. This genuinely democratises AI and gives countries who don’t have the existing infrastructure a huge leap forward to experiment and be part of the frontier.” InvestorAi Within days, the Chinese-built AI model has upended the industry, surpassing OpenAI’s o1, dethroning ChatGPT in the App Store , while NVIDIA’s market cap plunged by US$589 B . Unlike OpenAI’s closed ecosystem, DeepSeek-R1 is open-source, free to use, and radically efficient. It achieves state-of-the-art performance without requiring massive GPU clusters, forcing the industry to rethink the high-cost arms race in AI. App Store NVIDIA’s market cap plunged by US$589 B DeepSeek-R1 Donald Trump already called it a “wake-up call” for American AI companies. Donald Trump already called it a “wake-up call” It’s Cheap It’s Cheap It’s Cheap The Chinese AI startup has sent shockwaves through Silicon Valley, bringing into question US big tech dominance in the AI market. Built entirely on open-source technology and lower-end chips, DeepSeek sidesteps the need for high-end hardware restricted by US export controls and claims to have developed the model for just US$5.6 million . As a result, DeepSeek is available at a cost that is just 2% of what users would spend on OpenAI’s O1 model. US$5.6 million *“DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” *— Christian Struve, Cofounder & CEO of Fracttal *“DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” * — Christian Struve, Cofounder & CEO of Fracttal *“DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” * *“DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” * “DeepSeek is not the first open source model, but its MIT license and R1 thinking model stand out for their accuracy and low training cost, democratizing access to advanced AI, either locally or through accessible APIs, thus driving innovation.” Christian Struve , Cofounder & CEO of Fracttal . Christian Struve Fracttal *“DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications.” *— JD Raimondi Head of Data Science at Making Sense *“DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications.” * — JD Raimondi Head of Data Science at Making Sense *“DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications.” * *“DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications.” * “DeepSeek took the initiative that Meta had taken internally: competing with the big private models with public models that can be used by everyone at low cost. But DeepSeek’s differentiator with r1 is a new one: besides being free, it is efficient, low cost, and with performance comparable to the big boys (GPT o1, GPT o4). DeepSeek very easily positioned itself at the same level as Meta as a good competitor to the big boys for the “winning” (prevalent) model in the world of AI-powered applications,” says JD Raimondi Head of Data Science at Making Sense . JD Raimondi Making Sense The fact that they created this platform with under US$6 M investments has shaken the tech CEOs globally highlighting that game-changing innovations don’t necessarily need billion-dollar investments. “However, from a supply/demand perspective the GPU market which Nvidia dominates is still far away from hitting peak demand,” Bruce Keith adds. Is This AI’s TikTok Moment? Is This AI’s TikTok Moment? Is This AI’s TikTok Moment? But DeepSeek isn’t without controversy. The application is raising questions such as data privacy, national security, and geopolitical implications. DeepSeek’s privacy policy confirms that user data is stored in China. Former Indian IT minister Rajeev Chandrasekhar, one of India’s tech savvy politicians, has already asked , if DeepSeek is the TikTok of AI. already asked “…it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients.” — Harold Barber, COO and CFO at Escala 24×7 “…it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients.” — Harold Barber, COO and CFO at Escala 24×7 “…it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients.” “…it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients.” “…it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients.” “DeepSeek is disrupting the AI ??market with an innovative proposition that challenges the main players in the area and could radically transform the way companies access and use this technology. However, it is crucial to analyze in detail its opportunities and challenges, as well as its real potential to drive the digital transformation of our clients,” warns Harold Barber , COO and CFO at Escala 24×7 . Harold Barber Escala 24×7 The US Navy has already banned DeepSeek . banned DeepSeek Still, while open-source, the Chinese version of the AI works with strict censorship, and doesn’t answer politically sensitive questions. At the same time, its open-source nature allows developers to run it locally, without restrictions , a formidable point in its favour. run it locally, without restrictions A David Vs. Goliath AI Story A David Vs. Goliath AI Story A David Vs. Goliath AI Story Paramdeep Singh, Co-founder of Shorthills AI , says DeepSeek changes the complete GenAI narrative. Shorthills AI “It is like David has defeated Goliath. The old GenAI story was that only the large models could win. It required super-specialized skills, huge compute, thousands of latest GPUs, web-scale data, trillions of nodes, and huge amount of electricity to train a foundational language model. All of this translated to millions of dollars to train the model. That meant companies and countries with deep pockets were going to monopolize that market. “It is like David has defeated Goliath. The old GenAI story was that only the large models could win… Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters,” — Paramdeep Singh, Co-founder of Shorthills AI “It is like David has defeated Goliath. The old GenAI story was that only the large models could win… Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters,” — Paramdeep Singh, Co-founder of Shorthills AI “It is like David has defeated Goliath. The old GenAI story was that only the large models could win… Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters,” “It is like David has defeated Goliath. The old GenAI story was that only the large models could win… Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters,” “It is like David has defeated Goliath. The old GenAI story was that only the large models could win… Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters,” “Now we have Deepseek that completely flipped this story. We have a hedge fund manager releasing a model that beats the big daddies of GenAI on all parameters. The model is frugal and trained on spare capacity of GPUs that were lying idle. The model is so small that it can literally run on your browser. It is simply the best value for money model. We have a ray of hope where Large Language Model training and usage can be democratized. It’s not people sitting in ivory towers, but talent with frugal hardware that can train the best model. We all love this David vs Goliath story,” he says. *“The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” *— Mike Capone, CEO of Qlik *“The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” * — Mike Capone, CEO of Qlik *“The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” * *“The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” * The beauty of DeepSeek’s lies in its ability to aid and not just wow. As Mike Capone , CEO of Qlik , says, “The AI race won’t be won by creating the most sophisticated model; it’ll be won by embedding AI into business systems to generate tangible economic value. Global leadership in AI should be measured by its application, not just its invention.” Mike Capone Qlik And the story of glory continues.DeepSeek just announced Janus Pro , an AI image model that rivals OpenAI’s DALL·E 3. If, or maybe we should say, when DeepSeek gets into video generation, competitors like Google’s Veo and OpenAI’s Sora could be in trouble next. Janus Pro Google’s Veo OpenAI’s Sora Meanwhile, Alibaba released its Qwen 2.5 AI model it says surpasses DeepSeek . The pressure is on not just big tech or just the US, but also on smaller players and countries like India. Alibaba Qwen 2.5 AI model surpasses DeepSeek Navanwita Bora Sachdev, Editor, The Tech Panda Navanwita Bora Sachdev , Editor, The Tech Panda Navanwita Bora Sachdev Navanwita Bora Sachdev The Tech Panda