Recent weeks have witnessed a flurry of excitement around DeepSeek, a Chinese artificial intelligence model that sent ripples through stock markets and investment communities worldwide. As speculation mounted about its potential to upend US dominance in AI, investors scrambled to reassess their positions. But is DeepSeek truly the game-changer it’s been made out to be?
Let’s cut through the hype and examine what DeepSeek really means for the AI landscape and investment opportunities.
Debunking the $6 Million Myth
At the center of the DeepSeek narrative is the claim that its latest large language model (LLM) was trained for a mere $6 million using a relatively modest number of GPUs, while supposedly matching the performance of industry titans like OpenAI’s GPT-4 and Anthropic’s Claude.
This figure has captured headlines, but it fails to tell the complete story.
Think about it this way: claiming you’ve built a cutting-edge AI model for $6 million is like saying you’ve built a car for $1,000 while conveniently omitting the costs of the factory, engineers, and decades of automotive research that made it possible.
The $6 million figure doesn’t account for:
- The salaries of top-tier AI researchers
- The value of building upon existing research from leading AI labs
- Potential underestimation of the actual number of GPUs used
- Infrastructure maintenance costs
The Real Innovation: Reasoning Models
While cost discussions dominate the conversation, we’re missing the forest for the trees. DeepSeek’s true innovation lies in its reasoning model, known as R1.
Reasoning models represent the next evolution in AI capabilities — systems that can break down complex problems into manageable steps, moving beyond standard LLM prompting. What DeepSeek appears to have demonstrated is that creating reasoning models may not be as resource-intensive as previously believed.
However, creation is just the first step. Performance matters, and current assessments suggest DeepSeek’s accuracy doesn’t yet match that of OpenAI or Google. Reports indicate that DeepSeek’s R1 model hallucinates (generates false information) at significantly higher rates than comparable reasoning and open-source models.
Why This Matters for AI Infrastructure
For investors and industry watchers, DeepSeek’s emergence points to a counterintuitive reality: reasoning models becoming more accessible and common will likely drive increased demand for AI infrastructure, not less.
This follows a pattern economists call Jevons Paradox — when technology becomes more efficient, people typically end up using more of it, not less.
Consider these factors:
- Reasoning models require significantly more computing power than basic AI models
- As AI becomes cheaper and more capable, usage increases dramatically
- Companies continuously discover new AI applications that demand even more computing resources
Quality and Performance: The True Drivers of AI Adoption
The narrative around DeepSeek has focused heavily on cost reduction, but this misses the point. The real driver of AI adoption is improvement in quality and performance.
Reasoning models represent a significant leap in AI capabilities. They’re more accurate and more versatile, handling complex tasks that standard LLMs cannot approach. From an investment perspective, this increased capability drives a higher level of required AI infrastructure capital expenditure (capex):
- More AI hardware
- Enhanced cooling systems
- Greater energy requirements
- Expanded data center capacity
Is AI Model Development a “Good Business”?
DeepSeek’s approach raises important questions about the business models around AI development. Companies like Meta and DeepSeek are willing to open-source their AI models because they generate revenue through other means.
But major technology companies aren’t scaling back their AI investments in response to DeepSeek. They view AI as a transformative technology that will penetrate virtually every economic sector:
- Healthcare and drug discovery
- Automotive
- Financial services
- Manufacturing
- Entertainment
- Education
This dynamic resembles the airline industry, where carriers often struggle for profitability, but airplane manufacturers and airport operators can still run lucrative businesses.
The Bigger Picture: Evolution, Not Revolution
The most accurate way to understand DeepSeek is as part of the continual wave of innovations improving AI performance while reducing costs. Most such advances go unnoticed by investors and the public. DeepSeek simply happened to catch attention and trigger an intense market reaction.
Within the context of AI development, DeepSeek represents an evolutionary step, not a revolutionary leap. The broader story remains about making AI more capable and accessible, which historically leads to increased — not decreased — demand for computing power.
Investment Implications
For those looking at the AI space from an investment perspective, these moments of uncertainty may create opportunities. The broader AI infrastructure ecosystem remains well-positioned for growth, particularly for those maintaining a diversified approach to AI exposure.
Remember that short-term market reactions often reflect incomplete understanding of complex technological developments. Looking beyond the headlines to understand the technical nuances and business dynamics provides a more solid foundation for investment decisions.
As AI continues its rapid development, we’ll likely see more DeepSeek-like moments — innovations that trigger market volatility based on incomplete narratives. The key is to understand these developments in their proper context.
DeepSeek isn’t signaling the end of US AI dominance or a dramatic reduction in demand for AI infrastructure. Instead, it’s another step in making AI more powerful and accessible — a development that historically drives increased computing demands and creates new opportunities across the AI ecosystem.
For long-term investors, maintaining perspective during these moments of uncertainty is crucial to identifying and capitalizing on the real opportunities in AI’s continued evolution.