• Chinese AI startup Deepseek is shaking up global market
  • Investors wonder now if smaller AI players can outshine big tech
  • Experts debate the long-term impact of DeepSeek on the AI landscape

 

This week, a little-known Chinese startup called DeepSeek dropped a bombshell that shook the entire AI market.

The company, whose name probably rang zero bells to most people until Monday, said it has managed to create a chatbot that rivals OpenAI’s GPT in performance.

DeepSeek’s R1 is an open-source AI model designed to mimic human reasoning. The company claims R1 rivals or even outperforms leading US developers across key industry benchmarks, including mathematical tasks and general knowledge.

Following the news, DeepSeek’s app quickly skyrocketed to #1 on the Apple App Store in the US.

But for the markets, it wasn’t just the technology causing the buzz, it was the price tag.

DeepSeek claims it can do all this for around 20 to 40 times cheaper than its US rivals, having forked out just US$5.6million to train its model using lower quality Nvidia chips.

Investors clearly didn’t like hearing that, and it’s a major problem if you’re Nvidia, which has been the go-to supplier for top-of-the-line AI chips.

Nvidia’s stock dropped a stomach-churning 17% on Monday. In fact, at US$590bn it was the largest single day loss by a company in share market history.

Tech, data centre and even uranium stocks on the ASX also took a hit as local investors became worried about reduced demand for power-hungry data centres.

Most analysts are, however, recommending that investors let this all play out and wait for the dust to settle.

“In the aftermath you will find out how much [of this selloff] was an overreaction,” said moomoo’s Jessica Amir.

Meanwhile, President Trump, never one to miss a chance to take a swing at China, has jumped in, calling DeepSeek’s breakthrough a “wake-up call” for American companies.

“I’ve been reading about China and some of the companies in China, one in particular, coming up with a faster method of AI and much less expensive method,” Trump said.

“And that’s good because you don’t have to spend as much money.”

He has a point.

If DeepSeek really can pull this off, it could set the stage for a more competitive and cheaper global AI race.

 

Who is DeepSeek anyway?

DeepSeek was founded in 2023 by Liang Wenfeng, the former co-founder of China’s top hedge fund, High-Flyer.

We don’t know much about it, but it’s understood that Liang tapped into local talent and, crucially, stockpiled around 10,000 Nvidia A100 GPUs – the older version of the latest H100 GPUs – before the US slapped export restrictions on them.

The company’s rapid rise onto the global AI stage suggests that China might not be as far behind in the AI race as many had previously thought.

And it’s rattled some much bigger players.

Only last week, companies like OpenAI, SoftBank, and Oracle bragged about throwing hundreds of billions into Project Stargate, a huge AI infrastructure initiative backed by Trump.

But now, DeepSeek has got the tech world re-thinking: Could this mark the beginning of a whole new wave of smaller AI players with less deep pockets?

It’s a tough question because, at the moment, nobody knows for sure how DeepSeek’s model is really working.

Are they just taking existing AI models and tweaking them with outdated Nvidia chips? Or have they cracked some secret sauce?

Elon Musk, for one, reckons DeepSeek’s been bluffing about using outdated Nvidia chips, claiming the company has deliberately kept quiet about it due to the strict US export controls.

 

 

Smaller AI players to carve out their niche

Is this a major blow to AI players in the west, including Australia?

Paul Chan is the founder and CEO of Decidr, a subsidiary of ASX-listed Live Verdure (ASX:LV1).

He doesn’t see the lower costs of AI models as much of a threat to larger players. In fact, he argues the costs of AI are already constantly decreasing.

“I do not believe that AI costs from the larger players are expensive,” Chan told Stockhead.

“In fact, they constantly go down and when compared to equivalent outputs from people, they truly are many orders of magnitude better.”

He also believes that because training methods are open-source, it won’t take long for larger model providers to adopt impactful techniques.

When asked if DeepSeek’s rise could push a broader diversification in AI investments away from big tech, Chan is somewhat split.

“Yes and no,” he said, acknowledging that there will always be lower-cost providers and more premium suppliers of AI intelligence.

“That is the way all markets work for reasons that won’t change anytime soon.”

Chan, however, sees opportunities for smaller AI stocks like LV1/Decidr in the face of competition from larger global players.

“There will always be various amounts of data that lower-cost models can be applied to,” he said. “The more competition and diversity of models, the better.”

He suggested that as competition increases, so will innovation and the potential for smaller players to carve out their own space in the market.

 

Should investors now shift focus? Here’s what experts say

While DeepSeek’s rise has sparked talks about investors shifting focus altogether, or away from the Magnificent Seven tech giants, it seems the jury’s still out for now.

Charu Chanana, Saxo’s Chief Investment Strategist said that if DeepSeek’s cheaper AI model gains momentum, it could indeed put a lot of pressure on US AI giants, whose valuations are already stretched.

“….This development serves as a reminder that competition in the global AI arena is intensifying, and Nvidia may not be in the pole position forever,“ she said.

Cameron Gleeson, a senior investment strategist at Betashares, agreed and called DeepSeek’s emergence a warning of just how volatile US mega tech stocks could be.

“First of all, it’s a timely reminder of the volatility of Nvidia stock. The company is up an eye-watering 90% over the past year,” he said.

Gleeson now sees a shift in the industry, similar to what we saw during the dot com boom, where hardware (like Cisco routers) initially dominated, only for software (search, social, etc.) to become the real long-term winners.

He believes cheaper, more efficient AI training might be bad news for hardware companies like semiconductor producers, but in the long run it could really benefit software companies.

Richard Clode, portfolio manager at Janus Henderson, said that despite DeepSeek’s low costs, he was still expecting big spending on AI.

But he’s also cautioning investors to be more selective about which companies will benefit from this spending.

 

The views, information, or opinions expressed in this article are solely those of the interviewee and do not represent the views of Stockhead.

Stockhead has not provided, endorsed or otherwise assumed responsibility for any financial product advice contained in this article.

At Stockhead we tell it like it is. While Live Verdure is a Stockhead advertiser, it did not sponsor this article.

 

 

 

Read MoreArtificial Intelligence - AI

Featured Companies

You might be interested in