DeepSeek And Ye Shall Find
A few days ago on LinkedIn, the panic was palpable. The aspirational mountain-climbing selfies and Q1 progress reflections that usually make up my timeline were replaced by posts about Chinese AI company DeepSeek and the downfall of NVIDIA. Fortunately, before I had to parse through the news alone, Eric Kavanagh announced his latest webinar, “Don’t Panic! A Deep Dive on the DeepSeek $Trillion Crash.” Featuring Chief Strategy and Business Development Officer at Hammerspace Tony Asaro, the webinar addressed whether or not the panic of DeepSeek’s debut is warranted or not within the larger context of AI innovation.
“The reason there’s so much uproar about DeepSeek is that it was so cheap to train,” explained Kavanagh. Because DeepSeek’s models are of comparable quality to OpenAI for a fraction of the price, many people conclude that American AI is not as valuable as we’ve been made to believe. The rapid plummeting of American tech stocks threatens our faith in this phase of our exploration. But AI scientists argue that this fear response, and even the hope and promise driving it, is likely exaggerated. And Asaro more or less agreed: “It’s not good to be impressed with a single point in time.”
Coverage in Time Magazine made a helpful distinction, saying that “while DeepSeek does represent a genuine advancement in AI efficiency, it is not a massive technology breakthrough…the American AI industry still has key advantages of China’s.” Machine learning algorithms are supposed to become cheaper over time and there’s even talk that there are flaws in the reporting of DeepSeek’s expenses. Their startup costs likely were offset by a hidden stash of NVIDIA chips that they neglected to reveal to uphold compliance with U.S. export controls.
What’s undeniable is DeepSeek’s improvement in accessibility for users of the model, with fees 30 times less than those of OpenAI o1. This could prove significant for proponents of open source models in general, but as Asaro stated, the culture backing such models determines their success. “Just because it’s open source doesn’t mean it’s better. It has to be from a healthy community and adopted widely,” he said.
Ultimately, I’d say that DeepSeek’s arrival on the AI scene is less a death knell for American AI than a reminder of how quickly the landscape is evolving. While its cost efficiency is impressive, long-term success in AI isn’t just about who can train models the cheapest—it’s about infrastructure, talent, and sustained innovation. The real story here isn’t panic or collapse but adaptation and as the AI race continues, the winners won’t be those who react the fastest, but those who build the most resilient, forward-thinking ecosystems.