Chinese AI company DeepSeek is already shaking up the AI industry thanks to the launch of its new chatbot, sending tech stocks from the likes of AI giant Nvidia tumbling.
Although DeepSeek, founded in 2023, is a new entrant to the AI race, the Chinese firm behind it claims the latest model R1 is open-source and uses reinforcement learning without the need for human supervision.
But for the likes of Nvidia, which has dominated the global AI hardware market, it doesn’t appear to be great news.
Similarly for US rivals like Meta AI or ChatGPT, the hype around DeepSeek signals what many are now calling an “AI wake-up call” with its potential to cast a shadow over the US’s leadership in AI, thanks to the Chinese AI firm’s ability to develop AI models at a significantly lower cost – breaking down the barriers to entry for many.
But what do leaders across the North think – is it good news as the AI race hots up?
READ MORE: Is DeepSeek the beginning of the end for Nvidia’s AI dominance?
“DeepSeek is a big deal because it is basically as good as OpenAI’s o1 reasoning model (the latest publicly available most intelligent AI), it was trained really quickly (< 4 months), it used way fewer GPUs during training and is up to 98% cheaper to use than o1,” Luke Budka, AI director at Definition, tells Prolific North.
DeepSeek’s rise could potentially have “huge ramifications” for the AI market too.
“The US may now be controlling the compute supply chain, but DeepSeek’s R1 model shows this may not matter; there may already be enough compute out there to get to superintelligence,” he explains.
“It also has ramifications for Nvidia and companies buying its chips – if you can make these kind of intelligence breakthroughs with a more efficient training architecture, then the demand for compute could waver.
READ MORE: Demystify the myths and risks as an agency using AI
“It shows us that the AI leaders in the US might not be that far ahead. And because DeepSeek has open-sourced its new models (can see the code and accompanying detailed technical reports, not the training data though) and you can see the chain of thought process taking place (the literal ‘thinking’ the model is doing), the rest of the world has a chance to examine the tech in detail and catch up. Smart play if you’re committed to eroding the US’s AI moat.”
Richard Potter, CEO at Peak, says DeepSeek’s rise is big news for every sector.
“The rise of DeepSeek across major markets underscores the real – and rapidly growing – AI wave, which is quickly becoming ubiquitous in our daily lives,” he explains.
“Northern tech firms, as well as all others, must have a robust AI strategy that can truly leverage this technology. Without this, firms will be unable to stay competitive in a global landscape dominated by agile and innovative players.”
Behind the scenes, the fact DeepSeek has relied on less advanced chips paired with cutting-edge training methods has ruffled feathers as it’s now rivaling major AI players – but without the high-cost infrastructure.
DeepSeek is the brainchild of Liang Wenfeng, a Chinese hedge fund manager.
