DeepSeek, an artificial intelligence (AI) lab from China that is only a year old, has made headlines after launching their AI model that performs at the same level as other companies in the industry, but for a lower cost.
The company launched a free, open-source large language model that it developed in only 2 months for less than $6 million, using reduced-capability chips from Nvidia called H800s. (CNBC)
This launch has already sent US tech stocks into a bit of a frenzy, as it undermines the value that companies like Nvidia can supply moving forwards if AI businesses are able to create high-functioning, low-cost processors without its latest (and more expensive) tech.
Here, we look at everything you need to know about the company, including who they are, how their tech works and what it might mean for the US’ AI dominance.
Who Are DeepSeek?
DeepSeek is an AI company that was founded in 2023 by Liang Wenfeng. Liang is also the founder of High-Flyer, a hedge fund managing around $8 billion in assets (FT).
The DeepSeek lab is based in Hangzhou, China and is an independent company funded by High-Flyer. This has allowed the company to stay nimble in its research, without having to answer to boards of investors.
The company is made up of employees from China’s top universities and has focused on creating better, cheaper and faster open-source models.
What Do DeepSeek Do?
DeepSeek, though still relatively new, has caught attention for a number of reasons, including:
Building high-performance AI models at low costs: Their latest language model, the DeepSeek-V3 was developed for less than $6million. To put this into perspective, it cost OpenAI an estimated $100 million to train GPT-4. This is largely due to its use of older Nvidia chips to power its AI, rather than their latest releases. (WIRED)
Using older technology: The company used older Nvidia chips to power its AI, this pulls into question the need for more expensive and newer chips. The ability to leverage older technology not only brings the price of development down, it also undermines the value that companies like Nvidia can provide through their newest releases.
Fast development: Despite being a relatively new company, DeepSeek has already released a number of AI models including language processing models, a competitor to ChatGPT and models that use less power. This fast development has not compromised on quality, as their models either match or outperform those created by major AI players.
More from News
- Trump Lifts Sanctions in Syria: What Does This Mean For Syrian Businesses?
- Retail Cyber Attacks: Cartier And North Face Are The Next Retailers Affected
- A Look At The Different Technologies Volvo Is Bringing To Its Cars
- Klarna Launches Debit Card To Diversify Away From BNPL
- T-Mobile Now Has Fibre Internet Plans Available For Homes
- Bitdefender Finds 84% of Attacks Use Built In Windows Tools, Here’s How
- Japan Starts Clinical Trials For Artificial Blood Which Is Compatible With All Blood Types
- UK Unicorn Monzo Breaks £1 Billion in Revenue
Why Are They A Threat?
Just last week, in his first week of office, Donald Trump unveiled plans to make the US an AI superpower.
With companies like OpenAI, Nvidia, Meta and Google based there, and plans to open a billion dollar data centre called Stargate with Sam Altman, it seemed like this domination could be achievable. Until now.
The timing is particularly interesting, with Trump taking office and historical relations between the US and China hanging by a thread. As Trump works to impose tariffs on countries like China, finding new ways to innovate without the latest US tech is certainly a good way to fight back.
Not only that, but given Nvidia is a US company, the government is able to control which countries are able to receive exports of its latest innovations. By relying on older hardware, DeepSeek is taking back some of the power.
The DeepStack app has already been downloaded on mass in the US, surpassing ChatGPT in US app downloads. Given the recent TikTok ban that focused on protecting US citizens’ data from Chinese technology, it might raise further concerns about how the data will be used geopolitically.
In addition to this, DeepSeek’s ability to produce high-performing AI models at low costs is certainly turning heads. It’s proof that good AI doesn’t need expensive hardware or billions of investment, devaluing a lot of the work that has come out of Silicon Valley. This has already sent stocks falling. (BBC).
What Does This Mean for the Future of AI?
DeepSeek has already sent ripples through the AI world, and it doesn’t look like it’s stopping anytime soon.
The company’s ability to use older tech to produce cheaper, high-performance AI models has certainly shaken up the industry.
Not only does it pull the US’ AI dominance into question, it also shows that the billions of dollars of investment may not have been necessary in the first place.
As the AI race continues, the world will be watching DeepSeek closely.