In 2024, over 500,000 new enterprises entered the AI domain, sparking discussions about the opportunities in AI entrepreneurship. The AI revolution has turned OpenAI into a global sensation, but NVIDIA undoubtedly stands as the paramount beneficiary of the AI era. Their financial results for the third quarter of 2024 showcase a significant surge in both revenue and net profit, with their market value surpassing that of Microsoft.
AI companies and NVIDIA have been in a mutually beneficial relationship, with NVIDIA investing in AI startups to cultivate a customer base. In 2024, NVIDIA allocated approximately $1 billion in investments, with projections to increase in 2025, while other internet companies are also ramping up their AI investments.
However, entering the AI market is no easy feat. Talent, GPU computing power, and data are the three major costs for AI companies. For instance, Xiaomi’s creation of a massive computing cluster comes with hefty expenses. Data scarcity is also an issue, with even OpenAI needing to generate data for model training.
Here’s the content formatted for a WordPress blog post:
In a contrasting view, DeepSeek has successfully offered a method for low-cost training of large models. The DeepSeek-V3 model’s training costs are significantly lower than those of GPT-4 and GPT-5, albeit with a trade-off in precision due to its technical nuances.
Challenges in AI Entrepreneurship
– **Talent Cost**: The competition for AI talent is fierce, leading to soaring costs.
– **Computing Power Cost**: The demand for GPU computing power is immense and expensive.
– **Data Cost**: Public data is exhausted, leading to insufficient data for training large models.
DeepSeek’s Solution
– **Technical Know-How**: Utilizing MLA architecture, MoE architecture, FP8 mixed-precision training framework, and model distillation to reduce costs.
– **Cost and Defects**: While costs are lowered, precision is sacrificed, and technical defects exist.
Below are the key points about the future of the AI industry:
The Future of AI Industry
The AI industry is still accessible, particularly suited for companies with substantial financial resources or those not pursuing ultimate functionality. Risks and opportunities coexist, but the entry barrier has been raised.
The AI industry remains a playground for the ambitious, but caution is advised for those looking to enter. Companies with deep pockets or those pursuing low-cost models still have a chance. Here are two types of enterprises suitable for entering the market:
1. **Well-Heeled Enterprises**: Those capable of bearing high risks in pursuit of the industry’s pinnacle.
2. **Low-Cost Model Seekers**: Companies that can create large models with decent experiences at lower costs by compromising on precision.
Below is the core content of the article, formatted for your WordPress post:
In 2024, the number of new AI-related enterprises exceeded 500,000. NVIDIA’s investments in the AI field are increasing, but AI entrepreneurship faces significant challenges in talent, computing power, and data costs. DeepSeek offers a solution for low-cost training of large models, albeit with some defects. The AI industry still presents opportunities, but only for a limited type of entering enterprises.
