AI’s Economic Value: The 830 Trillion Won Question
1. The Reality of AI Infrastructure and Changing Revenue Models
In September 2023, “The $20 Billion AI Question” was released, highlighting the significant gap between the enormous costs of building and operating AI infrastructure and the actual revenue generated. The analysis at that time pointed out that the AI ecosystem’s revenue fell short of expectations, with an annual shortfall of approximately $125 billion. This remains one of the biggest challenges facing the current AI ecosystem.
2. NVIDIA and AI Data Center Costs
Recently, NVIDIA has emerged as one of the most valuable companies in the world. By calculating the total costs of AI data centers based on NVIDIA’s annual revenue forecasts, it becomes clear that the $20 billion AI problem has now expanded to an 830 trillion won ($600 billion) issue. This figure includes not only expensive hardware like GPUs but also operational costs such as energy, buildings, and backup generators.
Supply Shortages and Inventory Increases: Changes in AI Infrastructure
1. Resolving GPU Supply Shortages
By the end of 2023, GPU supply shortages had reached their peak, but they have now been resolved. This is a major boon for AI data centers and cloud service providers. With the stabilization of GPU supply, the expansion of AI computing has become more feasible.
2. Increasing GPU Inventory
More than half of NVIDIA’s data center revenue comes from large cloud providers, with Microsoft accounting for about 22% of NVIDIA’s Q4 revenue. As hyperscale CapEx reaches historical levels, GPU inventory is increasing. This means that securing the resources needed for building AI infrastructure has become easier.
OpenAI’s Revenue Share and the Limitations of AI Products
1. OpenAI’s Growth
OpenAI’s revenue increased from $1.6 billion at the end of 2023 to $3.4 billion currently. This reflects the high demand for AI technology. However, beyond ChatGPT, there are not many AI products that consumers actually use. This indicates that the commercial application of AI technology is still in its early stages.
2. Limitations of AI Products
The limited number of AI products that consumers actually use suggests that the commercialization of AI technology has a long way to go. This shows that more innovation and application are needed for AI technology to generate economic value.
Future Prospects of AI Economic Value
1. Increasing Costs of Building AI Infrastructure
The costs of building AI infrastructure continue to rise. While large companies like Google, Microsoft, Apple, and Meta are expected to each generate new AI-related revenue of $10 billion annually, the gap, which was initially $125 billion, has now expanded to $500 billion.
2. The Emergence of B100 Chips
NVIDIA recently announced the B100 chip, which offers 2.5 times better performance than the H100 with only a 25% increase in cost. This is expected to lead to a surge in demand for the B100. This can contribute to reducing the costs required for building AI infrastructure.
Key Counterarguments and the Long-Term Value of AI Technology
1. Lack of Pricing Power
Unlike physical infrastructure, GPU data centers have less pricing power. GPU computing is increasingly becoming commoditized and is metered by the hour. This suggests that the commercial value of AI technology could change over time.
2. Investment Burn and Depreciation
Like physical infrastructure such as railroads, speculative investment booms have high capital burn rates. Semiconductors improve in performance over time, so depreciation of previous-generation chips occurs rapidly. This indicates that continuous innovation is needed to maintain the economic value of AI technology.
Conclusion: The Economic Value of the AI Revolution
AI will create tremendous economic value. NVIDIA should be credited for enabling this transition. Speculative booms are part of technology, and there’s no need to fear them. However, one should not be under the illusion that AGI will arrive soon. The road ahead is long and bumpy, but it is almost certainly worth it. The economic value of AI will be realized through continuous growth and innovation.