The rising popularity of the new artificial intelligence model R1 from Chinese firm DeepSeek has not only shaken traditional U.S. technology stocks, but has also sent shockwaves through the crypto AI sectors, dramatically impacting the prices of AI-related tokens.
While the price of tech stocks, such as Nvidia, Arm, and Broadcom have declined, AI-related cryptocurrencies have seen substantially sharper drops as the worst-performing categories in the crypto space in the last 24 hours are all related to artificial intelligence, data from CoinGecko shows.
Virtuals Protocol, a tooling platform on Base with plans to run on Solana, allows people to make AI agents or smart computer programs that can act on their own such as transacting onchain or posting content on social networks. As a category, tokens that have emerged from the Virtuals ecosystem, have dropped 33.7% to a market capitalization of $2.4 billion.
Read More: Coinbase-Incubated L2 Base Sees Record Transactions Thanks to AI Memecoin Deployer Clanker
‘Validation’ of Decentralized AI
The selloff comes one week after DeepSeek released its R1 model and as DeepSeek became the number one mobile application on the Apple App Store, outdoing OpenAI’s ChatGPT.
Part of DeepSeek’s rising popularity stems from how its newest open-source model is reportedly able to match the abilities of other notable AI models such as ChatGPT at a fraction of the cost by using less computational power.
Jake Brukhman, the co-founder of investment firm CoinFund which has been investing in decentralized AI, told Unchained over email that DeepSeek’s model operating more efficiently than OpenAI’s Chat GPT and leading mobile application downloads “is the biggest validation story of decentralized AI to date.”
“For a long time, the narrative has been that GPU-rich Big Tech companies have achieved an unassailable first mover advantage and that only a few huge monolithic models would take the market.” GPUs, or graphics processing units, are hardware components aimed at handling complex computational tasks, typically used for cryptocurrency mining and running advanced AI models.
Read More: GOAT, How AI Agents Talking Turned Into a $268 Million Memecoin ‘Religion’
The advent of DeepSeek’s more efficient and open-source model, which enables anyone to modify and distribute the code without charge, challenges that GPU-rich narrative and “shows the cracks in OpenAI’s traditionalist web2 strategy.” Brukhman noted.
DeepSeek highlights that the pre-training of AI models is becoming less of a competitive advantage or barrier for entry. “After DeepSeek, it appears that pretraining of models is proving less of a moat, and we can posit that the moat of data erodes over time as well,” Brukhman wrote.
While OpenAI released code for its AI model in 2019, the organization has not released the code for its recent models such as GPT-4o, which was launched in May.
The underperformance of the crypto AI sector is likely related to the DeepSeek news, “but the underpinning logic behind the selloff doesn’t make sense to us here at Ryze,” Matthew Graham, managing partner at venture capital firm Ryze Labs, wrote to Unchained over Telegram.
Graham’s reasoning as to why Monday’s selloff doesn’t make sense includes the economic concept Jevons paradox, which states that improvements in resource efficiency can lead to an increase in overall consumption of that particular resource.
DeepSeek’s ability to match leading AI models with greater efficiency, Graham argues, will make it easier for people to launch AI agents. “The barrier to entry will trend to zero, just as it did with launching memecoin tokens. But as with pump.fun’s memecoin cash cow, that won’t preclude building powerful AI-centric businesses,” Graham added.
