on January 31, 2025
If you missed the news, DeepSeek is a Chinese version of OpenAI. They made the headlines a couple of weeks ago, and everybody was already burying all the American companies. And the Europeans smelted into a collective depression. I have seen and read so many rumors. I am sharing a bit of analysis. Let’s have a little critical thinking here.
Rumors
Here are the recent rumors I heard, if you heard more, let’s chat!
Rumor: They did it for dirt cheap.
Like anything coming from China, there is a natural lack of transparency. USA Today reports an investment of under $6m to create it. However, even if it’s true, what does it entail? Did they get help from universities? Did they get “free” credits from Alibaba Cloud?
Rumor: They do not have access to prime hardware.
Anyone can easily open an account on any cloud provider. Since Jack Ma left Alibaba in 2019, the Chinese government has been even more involved in the company. They have data centers worldwide, including in the United States. Besides diplomats, do you know what goes through borders without control? Data.
Rumor: OpenAI claims they stole their data.
Besides the irony, it could be true. OpenAI lost a lot of brains recently. Training data and models are big, but so are USB flash drives or hard drives in those days.
Rumor: There are sanctions in place.
Sure, ASML, the leader in microprocessor machinery, cannot export its latest technology to China. But do they need it (for this project)? Black market, remote computing, and illegal importations
Rumor: They’re doomed anyway; it’s open source.
This one cracked me down. Let’s remind everybody that IBM acquired Red Hat for $34bn. In its earnings call, IBM’s CEO, Arvind Krishna, confirmed that Red Hat (along with AI) contributed to the double-digit growth of their revenue. There are (way too many) free models on Hugging Face. There is even a definition of what Open Source AI and Open Source AI models are. Do they match those definitions? Particularly the one who lists the sources.
Rumor: The stock market tanked.
I’ll admit that the stock market did take a hit. But look at the Nvidia stock: Yup, there has been a loss over the last five days, but over the last six months and yesterday, it was all green, Mr. Giant. Will it go back to its previous value? Were the stocks too high anyway? I don’t know. I am a data guy, not a predictive analytics guy.
Rumor: DeepSeek is much cheaper than OpenAI.
OpenAI and DeepSeek have similar cryptic pricing models based on tokens, which vary depending on whether they are input or output. These prices apply to APIs, not the “human” service, which is free (so far) on DeepSeek. Think about the long-term strategic alliance you are building using a model provider. Today’s price will not be the price tomorrow: it’s easy to start with a low price and raise it afterward. This strategy is a typical case of predatory pricing.
Rumor: DeepSeek is as good as OpenAI.
Not close. First, the content is sanitized. Ask any controversial question regarding China. I asked it to describe the political regime in China and who the Uyghurs are. Of course, those questions are provoking the machine. You can imagine that the Chinese government was involved in something as strategic as AI. Second, the SLA, although OpenAI does not have a stellar record, DeepSeek has no track record yet.
Nevertheless, seeing the reasoning in progress and the harsh filter at the end was funny. I should’ve recorded a video. It’s obvious for those two questions, but what happens when they are slightly less obvious? Remember: who controls the model controls the narrative.
Question: What about AI regulations?
With DeepSeek making waves, is the US or EU considering new AI regulations? There’s been some chatter about the West cracking down on Chinese AI models, much like TikTok and Huawei faced restrictions.
Rumor: Chinese coders are super good.
I would not generalize, but I work with some, and, as we would say in French, they don’t have broken arms (it means they are pretty good). However, it was the same for Eastern European developers in the 80s and 90s. Scarcity drives optimization from early on while we transfer petabytes of data in JSON…
The Bigger Picture: A Race Worth Running
Whether DeepSeek is a real threat to OpenAI or just another headline-driven hype cycle, the global AI race pushes everyone to innovate faster, think smarter, and collaborate better (Have you checked the definition of a Data Product?).
The U.S. remains a powerhouse in AI, with OpenAI, Anthropic, Google DeepMind, IBM, and others leading breakthroughs. However, challenges remain, such as funding, regulations, and ethical concerns. Meanwhile, Nvidia is breaking records, the U.S. is investing in domestic chip manufacturing, and companies like Global Foundries, IBM, Microsoft, and startups alike are doubling down on open-source AI, ethical frameworks, and enterprise adoption. The key question isn’t whether DeepSeek will “win” or “lose”, but how the AI ecosystem will evolve globally — and what role we each play in shaping it.
What’s your take? Let’s discuss. 👇
Is DeepSeek an Enormous Geopolitical Teasing? was originally published in jgpai on Medium, where people are continuing the conversation by highlighting and responding to this story.