The AI industry is experiencing a significant growth bubble, raising questions about its future trajectory and revenue generation. The author previously highlighted a gap between the expected revenue from AI infrastructure development and the actual growth in the AI ecosystem. With Nvidia becoming the most valuable company in the world, the author revisits the revenue question, estimating it to be $600B. Several changes have occurred since the last analysis, including the subsiding GPU supply shortage, the growth of GPU stockpiles, and OpenAI’s dominance in AI revenue. However, the $125B revenue gap has now increased to $500B. Additional factors, such as lack of pricing power, potential investment incineration, and chip depreciation, contribute to the complexities of the AI industry. Despite the challenges, the author believes that the long-term potential and value creation of AI make it a worthwhile and rewarding technology wave.
| Signal | Change | 10y horizon | Driving force |
|---|---|---|---|
| AI’s $600B Question | Increase in revenue expectations for AI industry | Increased revenue generation in the AI ecosystem | Need to fill the revenue gap and meet the cost of AI data centers |
| Supply shortage has subsided | Easier access to GPUs | Increased availability of GPUs with reasonable lead times | Decreased concern and improved accessibility for startups and businesses |
| GPU stockpiles are growing | Increased stockpiling of hardware | Historic levels of investments in GPUs by major tech companies | Anticipation of future demand and potential decrease in demand |
| OpenAI dominates AI revenue | OpenAI has the majority share of AI revenue | Gap between OpenAI and other startups remains large | Need for AI companies to deliver significant value to consumers |
| The $125B hole is now a $500B hole | Increased revenue gap in AI industry | Growing disparity between revenue expectations and actual revenue growth | Need for additional companies to generate revenue in the AI industry |
| Introduction of B100 chip | Improved performance and increased demand for NVDA chips | Surge in demand for B100 chips | Expectation of better cost vs. performance improvement |
| Lack of pricing power for GPU data centers | Decreased pricing power for GPU computing | Increasing commoditization of GPU computing | Competition from new entrants and absence of monopoly or oligopoly |
| Potential for high rates of capital incineration | Speculative investment frenzies leading to capital loss | Possibility of high rates of capital loss in AI industry | Historical trends in technology investments |
| Rapid depreciation of last-gen chips | Faster depreciation of older chips | Overestimation of value retention for H100s | Continuous improvement in next-gen chips |
| Winners and losers in AI industry | Some companies benefit from infrastructure building | Long-term innovation and success for founders and company builders | Lower costs and learning opportunities for startups |
| Potential generation-defining technology wave | AI as a transformative technology | Creation of significant economic value through AI | Positive impact on the overall ecosystem |
| The road ahead will be a long one | Challenges and obstacles in the AI industry | Anticipation of ups and downs in the AI industry | Long-term commitment and perseverance required |
| Continued communication and collaboration in the AI industry | Seeking input and feedback from those building in the AI space | Opportunity for collaboration and knowledge-sharing in the AI industry | Community engagement and support for AI ecosystem development |
| Challenging assumptions on non-GPU data center costs | Revisiting energy costs in non-GPU data centers | Alignment between assumptions and Nvidia’s metrics | Need for accuracy and consensus in cost assessment |