Chinese AI startup DeepSeek’s R1, a new open source reasoning model that surpasses OpenAI’s o1 and Google’s Gemini on multiple benchmarks, has sparked a global market turmoil.
The Hangzhou-based startup founded by Liang Wenfeng in 2023, which released the model quietly, shot to No. 1 position on Apple’s App Store, outperforming US tech players’ products.
Andreessen Horowitz cofounder Marc Andreeson describes DeepSeek R1 as AI’s ‘Sputnik moment,’ a turning point when a country feels the need to catch up to another country’s technological or scientific advancements.
Tech giants like Nvidia, Microsoft, and Alphabet saw their stock prices plunge, while the Nasdaq 100 and Europe’s Stoxx 600 tech sub-index bled nearly $1 trillion in market cap.
US stocks, particularly in the semiconductor sector, saw significant losses, while Chinese markets held steady, with their strongest performance relative to US markets in over two years. Chip maker Nvidia’s shares were down nearly 15% on NYSE in early trade on Monday.
Speaking at the World Economic Forum earlier this month, Satya Nadella said, “We should take the developments out of China very, very seriously.”
It’s all About Challenging US’s AI Dominance
For years, the AI debate centered on open-source vs. closed models. But today, it’s moving beyond and questioning who has or will have AI dominance.
Meta’s chief AI scientist Yan LeCun said, “To people who see the performance of DeepSeek and think China is surpassing the US in AI. You are reading this wrong. The correct reading is: Open source models are surpassing proprietary ones.”
He added that DeepSeek’s success stems from open research and tools like PyTorch and Llama, which were developed by Meta. By building on existing work and contributing back, they’ve shown the power of open-source collaboration in advancing the AI ecosystem.
Additionally, HuggingFace is replicating the entire DeepSeek R1 pipeline for the open-source community.
Now, it’s about who actually controls AI’s future. Despite US trade restrictions, the country, with limited resources, proved to the world that it could develop a cost-effective alternative in a short span. However, at the WEF in Davos, Scale AI CEO Alexandr Wang said that DeepSeek has around 50,000 NVIDIA H100s that they cannot openly discuss due to US export controls.
Interesting to note that while OpenAI is charging a whopping $200 a month for its version of the o1 reasoning model, DeepSeek is offering it for free. DeepSeek has in a way proved that the US has no AI moat.
“While NVIDIA has a strong foothold, AI dominance cannot be taken for granted,” warned Charu Chanana, chief investment strategist at Saxo Markets. “The emergence of China’s DeepSeek indicates that competition is intensifying. Future competitors will evolve faster and challenge established companies more quickly.”
Meanwhile, OpenAI’s o3 model—its next big release—still isn’t public.
India’s AI Awakening
DeepSeek’s disruption isn’t just shaking Wall Street—it’s compelling India to reassess its AI strategy. VCs, echoing Andrew Ng, consider this the prime time for application layers. Per Ng, the “DeepSeek sell-off” highlights the potential in applications, and fierce foundation-model rivalry ultimately favours app builders.
For years, Indian tech leaders have argued over whether to build foundational models or just focus on AI applications. But now?
“We need to build, not just wrap existing AI,” said Perplexity AI CEO Aravind Srinivas, after seeing DeepSeek’s success.
Srinivas is now offering $10 million to any Indian team that can build a competitive AI model from scratch. Others, like Turing’s Dream, Sarvam AI and Zoho’s Sridhar Vembu, are shifting their focus to deep AI research.
Ajai Chowdhry, HCL co-founder, stressed the need for strategic autonomy and said: “This should be a huge wake up call for India, and we must work in emergency mode to win the AI race. In this new unipolar world, every country is on its own.”
Meanwhile, Jio’s Gaurav Aggarwal echoed this urgency, saying, “I left Google Research to ensure India doesn’t get AI-colonised.” Sarvam AI’s Pratyush Kumar is already building sovereign AI models and inviting collaboration from Srinivas. Researchers like Paras Chopra and Zerodha’s Kailash Nadh emphasised grassroots efforts and deeper research funding as crucial for long-term success.
India’s IT giants, including TCS, Infosys, Wipro, and HCLTech, are working on AI projects, mostly focusing on the applications and client side of things, not building foundational technologies for the country.
But there’s one exception. Tech Mahindra is probably the only system integrator that has built a foundational model from scratch.
Last year, at MachineCon GCC Summit 2024, former Tech Mahindra chief CP Gurnani revealed that it was able to develop an Indian LLM for local languages and 37+ dialects in just 5 months, spending less than $5 million.
[Must Watch] Check out the latest episode of ‘What’s the Point?,’ where we unpack what AI for Bharat really means.
The message is clear: AI dominance isn’t about who spends the most money—it’s about who moves the fastest and smartest.
