Everyone in the AI industry can’t seem to stop talking about the seismic shock that Chinese company DeepSeek has created, and for good reason.
Last week, the startup unveiled an AI model that appears to rival industry leaders like OpenAI and Anthropic, at a fraction of the cost. Consequently, this has sent ripples through both public and private markets, challenging long-held assumptions about the economics of AI development.
VC’s $300 Billion Question
DeepSeek’s claimed efficiency that it used just $6 million to train its latest model is simply stunning when compared to OpenAI’s estimated $100 million investment in GPT-4. This dramatic cost difference has raised eyebrows across Silicon Valley and beyond, particularly given the model’s sophisticated capabilities.
The impact of this was felt immediately as the DeepSeek AI app shot to the top of Apple’s App Store, while public markets reacted strongly, with Nvidia losing nearly $600 billion in market value and the Nasdaq dropping 3% in a single day.
This development couldn’t come at a worse time for venture capital firms, which have invested over $300 billion in US AI startups over the past five years. The stakes are especially high for those who backed foundational model companies, with 2024 alone seeing $40 billion flowing into the sector.
OpenAI’s recent $6.6 billion financing round pushed its valuation beyond that of any previous VC-backed tech company at IPO, making it incredibly vulnerable to shifting market sentiments.
Split Reactions from Silicon Valley
Unsurprisingly, the venture capital community has split into two distinct camps in response to the emergence of DeepSeek.
- The Optimists: Y Combinator’s Garry Tan argues that lower training costs will democratize AI development, potentially helping more startups reach $10 million in annual recurring revenue. Others, like investors at Battery Ventures and Andreessen Horowitz, see this as a catalyst for innovation rather than a threat.
- The Concerned: On the flip side of the coin, some investors are questioning the astronomical valuations of companies like OpenAI, which is losing $5 billion while generating $3.7 billion in revenue. Thomvest Ventures’ Umesh Padval acutely asked, “How do you make money? I can’t see a path.”
DeepSeek’s breakthrough could represent what many have referred to as “the Sputnik moment for AI”—a wake-up call for the US tech industry. While questions remain about DeepSeek’s actual costs and potential national security implications, it’s clear the assumption that building competitive AI models requires billions in investment has been fundamentally challenged.
In time, investors who focused on AI applications rather than foundational models will be vindicated, drawing parallels to the dot-com era where the ultimate winners weren’t the infrastructure providers but companies that built powerful products on top of that infrastructure.
The AI race has entered a new phase, where the spoils will go to those who join rather than fight progress.
