DeepSeek, a powerful new open-source AI from China, may have put a pin to the AI valuation bubble. While that’s bad news for investors, it’s good news for marketers.
What it is. DeepSeek is said to work as well as OpenAI’s o1 while costing 95% less. It was built without the state-of-the-art NVIDIA chips used by OpenAI, Google and most other big GenAI companies. These chips can cost up to $70,000 each. In addition to using less expensive equipment, DeepSeek’s developers created a less expensive way to train it.
“DeepSeek came out and is beating the pants off everyone and is literally 1/100th of the cost of OpenAI,” said Chris Penn, co-founder and chief data scientist of TrustInsights. “It is also open source. They’re giving away the model so that you can try it out. They’ve published all the research and done a really good job with that.”
Penn said this is a showdown between open-source models and closed-source proprietary ones like ChatGPT. So far, the closed-source model makers are having trouble keeping up. He said DeepSeek’s open-source model, which allows many people to use and improve it, is going to spur major AI breakthroughs.
“I expect in the next six to 12 months we’ll see very, very rapid innovation along the lines of reasoning models because of the work that DeepSeek has done,” he said.
Dig deeper: Watch an excerpt of MarTech’s conversation on DeepSeek with Chris Penn
DeepSeek has disrupted the AI market because it doesn’t have the resources the established players have. The U.S. has banned exports of the most advanced chips to China for national security reasons. So, the company had to win by innovation.
“DeepSeek appears to have achieved a breakthrough in resource efficiency—an area that has quickly become the Achilles’ Heel of the industry.” said Mali Gorantla, chief scientist at AI security company AppSOC. “Companies relying on brute force, pouring unlimited processing power into their solutions, remain vulnerable to scrappier startups and overseas developers who innovate out of necessity. By lowering the cost of entry, these breakthroughs will significantly expand access to massively powerful AI.”
‘Embarrassed’ Silicon Valley
“[They’ve] basically embarrassed a whole bunch of Silicon Valley,” Penn said. “That’s good for us. That’s good for marketers. That’s good for consumers because it allows us access to reasoning technology at very, very low cost.”
Penn is very impressed with the results he’s already getting from DeepSeek. At the start of a talk last Thursday, he put a set of project requirements and some best practices into a coding environment in DeepSeek’s reasoner.
“I hit go, and it wrote an app start to finish that ran the first time in 20 minutes,” he said. “That changes pretty much everything. Because if you think about that in software, imagine that in content creation and ads in and copywriting, where the model can sit there and just chug along and say, okay, I’m going to think through how I’m going to do this.” And it does this according to the requirements it’s given.
DeepSeek is also different from ChatGPT and company in that it shows its reasoning while working on something. Today, the app is on top of the free downloads chart on iPhones in the U.S. and is among the most downloaded productivity apps in the Play Store. Reviews on both sites praised the app’s transparency.
‘Has the industry been wildly overspending?’
DeepSeek’s cost, quality and open-source model threaten to undermine the technology firms that have spent hundreds of billions of dollars on the technology.
As Andrew Sorkin of The New York Times wrote in his newsletter this morning, “The super-efficient, open-source software is raising questions about the valuations of tech giants, including the chip maker Nvidia, with their stocks getting crushed today. Has the entire industry been wildly overspending?”
Investors clearly think so. At the end of trading Monday, Nvidia was down 17%, wiping out more than half a trillion dollars in market value. Other big losers included fellow chipmakers Marvell Technology (27%) and Broadcom (17%); and networking technology provider Arista Networks (28%). Even Google parent Alphabet dropped 4.2%.
Dig deeper: Are AI tools shaping your intentions more than you realize?
The massive investment was spurred by FOMO on a risky tech, according to Paul Roetzer, CEO of The Marketing AI Institute.
“Microsoft and Amazon and Google are all basically saying ‘We’re going to spend tens of billions of dollars,’ and it may not work,” Roetzer said. Because “the risk of not doing it is far greater than burning some cash. And if we end up losing a hundred billion dollars, so be it because the upside is a trillion.”
This, along with the large amount of money invested in AI by venture capital, creates what seems like an asset bubble. That’s when the price of an asset increases rapidly and becomes detached from its intrinsic value. Some market observers were concerned about this issue well before DeepSeek’s arrival.
Marketers should keep an eye on the impact of the AI bubble’s collapse and how it affects the overall economy. Will it be like the very short-term Dot Com bubble in 2000 or the Mortgage Meltdown of 2008? At this point, it’s looking more like the first of those two.
See an excerpt of the conversation with Christopher Penn below:
