While the “Magnificent Seven” tech giants have spent billions developing their artificial intelligence (AI) systems, DeepSeek’s model reportedly uses much less power and operates on far less expensive hardware, making its rise to prominence all the more remarkable.
The start-up has raised eyebrows by claiming to have developed its ChatGPT-like platform in under two months for just $6 million, but nevertheless, a professor believes its arrival could signal a seismic shift in how AI is developed and deployed, lowering entry barriers for start-ups globally.
Speaking on a recent episode of the Relative Return podcast, Professor Geoff Webb from Monash University’s Department of Data Science and AI said if DeepSeek’s claims hold true, it would show that AI innovation doesn’t have to come with a billion-dollar price tag.
“If it’s possible for other players to come in at a lower investment than was previously thought necessary, there’s much lower barriers to entry and so there are likely to be many more players,” Webb said.
“Until now, I had thought there was, there was really no prospect of Australia having its own homegrown equivalent to one of these systems. And now it actually looks like for a middle power like Australia, it is actually reasonable thing,” he added.
According to Webb, the model, which reportedly doesn’t rely on the powerful hardware that Nvidia has built its dominance on, challenges the very foundations of how AI is deployed today.
AMP’s Shane Oliver agrees, stating that if DeepSeek’s claims hold true, it would have “huge implications” for the AI market.
“If DeepSeek can do it – the same output with less input – then others will as,” the chief economist said in a market note on Friday.
“Yes, the supplier for a new technology may do well but then competition comes along and costs collapse and it becomes ubiquitous. Think of what happened to flat screen TVs which collapsed in price over the last two decades or traditional computer chips.”
It could also mean that the billions being spent on AI development, including Trump’s US$500 billion Stargate announcement, may not be necessary if the same can be done with less.
China as a disruptor
Another fascinating wrinkle is the role of China in this disruption.
Webb pointed out that DeepSeek sends a clear message to the world – despite US efforts to limit China’s access to cutting-edge AI tech like Nvidia’s A100 chips, Chinese companies are proving they can still compete on the global stage.
“There’s been a very serious attempt by the US to keep China out by denying them the most recent AI hardware, the Nvidia A100 chips, and DeepSeek claims not have used these, but instead to have just used the deliberately less capable chips that were made explicitly for the Chinese market,” Webb said.
“That shows that this is no longer the preserve of the US and that’s a very significant development.”
Oliver also suggested that Chinese companies may be further ahead in developing AI than assumed, questioning whether US tech export bans are actually working.
“I don’t think this really threatens the US lead though, which ultimately derives from its inventiveness, which China, like the USSR generations ago, will struggle to replicate.”
The rise of low-cost AI models could also have significant environmental implications.
If DeepSeek’s approach is as energy-efficient as it claims, it could pave the way for more sustainable AI development, reducing the carbon footprint of data centres worldwide.
“The environmental implications of the massive power being required and the amounts of power required are extraordinary and do have clear environmental impacts,” Webb said.
“I don’t think it’s widely known that Microsoft has recommissioned the Three Mile Island nuclear power plant in order to provide the power that these systems need. And there’s investments from other big AI companies in modular nuclear power … So, if slightly less power is required, then that’s certainly going to be a good thing.”
As for the stock market, which reacted sharply to DeepSeek’s emergence – sending the Nasdaq 100 down 3 per cent in hours – Webb noted that while some of this was a natural correction, it also raised deeper questions about whether tech stocks are overpriced.
Looking forward, the professor said given tech will only continue to innovate and evolve, “there’s no doubt” that companies operating in this space will continue to attract substantial returns.
“There’s no doubt that the US is still in the lead. So DeepSeek is impressive, but it’s not by any stretch of the imagination at quite the same point as the leading US systems and certainly not ahead of them … They still clearly have very substantial value. So I wouldn’t be ditching my investments in them, but maybe hedge,” Webb said.
Oliver advised investors to prioritise diversification.
“Tech and AI related shares have provided a good ride and it will be really hard to time any correction in them, but history tells us that sooner or later it will have one, but at the same time it may not necessarily drag down the rest of the market if its gradual,” the chief economist said.
To hear more from Webb, click here.
