• The release of DeepSeek R1 last week has turned the AI sector on its head.
• Tasked with creating an AI model with minimal access to technology, DeepSeek’s R1 out performs or matches OpenAI’s model at a lower cost.
• On Monday, the value of the AI market in the US with NVIDIA alone losing $600 billion in value.

Since OpenAI launched ChatGPT, we have been told that in order for artificial intelligence to get better, more and more compute power would be needed. It’s an argument that makes sense, or rather made sense until this week.

AI startup DeepSeek has now turned that thinking on its head and in the process, deleted value from the biggest players in the AI space, including NVIDIA. In just one day, NVIDIA’s stock value plummeted, losing approximately $600 billion in value in a single day.

As you may be aware, NVIDIA’s stock price climbed dramatically as AI firms started using its hardware to drive their models and platforms. This lead to gargantuan gains lead largely by the belief that the more demand there was for AI, the more costly hardware would be needed to drive it.

Then, DeepSeek-R1 arrived.

Released last week, DeepSeek claimed that its R1 release was on par with OpenAI-o1 but in reality, R1 outpaces OpenAI’s tech in a number of benchmarks. The tech world quickly began using the open-source model and found that R1 lived up to its creator’s claims.

What makes the tech so attractive though is that it compared to OpenAI, the price of API access is very low and more importantly, the model can be run on consumer grade hardware. Seriously, somebody managed to get the model running on a few Macs, putting the entire hardware industry being propped up by AI crashing to its knees.

But NVIDIA isn’t the only company weak at the knees. Microsoft, Alphabet, and Amazon which have all invested heavily into AI and were seemingly only competing with OpenAI, now face real competition from an outside entity.

What makes this even more alarming for the Silicon Valley techbros is that DeepSeek wasn’t even trying to compete with the likes of OpenAI. In fact the platform seemingly started out as a side project of hedge fund High-Flyer that wanted to develop AI for trading. By 2023 DeepSeek had been spun off as its own company, but it struggled to court investors beyond High-Flyer which funded the company in full thanks to its reported valuation of $8 billion.

Tensions between the US and China saw the US block the sale of tech to China but DeepSeek hoarded GPUs in order to train its AI models. With minimal access to tech, DeepSeek had to be creative and as such innovated on innovations to reduce the compute power overheads required by the bigger players in AI. This makes DeepSeek more cost effective to run while offering better or equal performance to OpenAI’s models.

Many have pointed out that by imposing restrictions on China, a model like DeepSeek was bound to happen as China would still aspire to compete with American firms and would have had to eventually get creative in developing models that could compete. In effect, the US shot its AI sector in the foot by trying to help it.

While DeepSeek has the attention of the AI sector, it’s not without its problems. The Conversation notes that the R1 model will still censor responses in line with Chinese law and the fact that data is stored in China may worry US companies.

But that matters little outside of the US and especially in developing regions where the sort of compute power need to run AI models is fiscally untenable.

The capabilities and prowess of DeepSeek has however upended the AI market and likely popped the bubble that showed no signs of shrinking. In fact, we suspect that OpenAI, Microsoft, Alphabet, AWS and other AI firms that courted billions in investment are being asked some tough questions right now.

We look forward to seeing how else AI can be disrupted now that DeepSeek has shown that with a bit of clever thinking, you can turn an entire trillion dollar industry on its head.