Summary of Key Points
The recent release of Kimi3, a large-scale AI model representing the “dark side of the moon,” has caused a seismic shift in the AI industry chain. For the past seven months, the market has been profiting from “bottleneck transactions” (shortages of chips and hardware). However, Kimi3’s more efficient open-source approach has raised concerns: will cloud providers cut back on capital expenditures (CAPEX) and lead to a collapse in hardware demand? This article focuses on this core disagreement, comparing the narratives of various parties—those who see benefits for hardware and those who fear insufficient ultimate demand. It also uses examples like Apple’s “passive profit” strategy and the Jevons Paradox to emphasize the importance of doing your own research (DYOR) before investing, as well as paying attention to “narrative perception” (the degree of market consensus divergence).
1. What Exactly Has Kimi3 Triggered in the Market?
In simple terms, the market is worried that the trend of increased investment in AI infrastructure might come to an end.
Over the past six months, AI models have been seen as a lucrative opportunity, and the demand for chips and memory (the “stoves” and “cutting boards” needed to build these models) has driven prices up. Both chip manufacturers (like Nvidia) and cloud providers have benefited significantly from this situation. But with Kimi3’s emergence, some have realized that cheaper, open-source models can achieve similar results with fewer resources. If this new approach becomes widespread, cloud providers won’t need to invest heavily in new infrastructure, which could lead to a decline in hardware demand. This is the direct cause of the market downturn on Friday. The main question remains: will cloud providers indeed reduce their capital expenditures?
2. Multiple Parties Argue That Kimi3 Is Actually Beneficial for Hardware
There are several arguments from proponents of hardware:
1. Lower profits, but greater demand leverage: Partners at Paradigm suggest that while open-source models may squeeze the margins of leading AI companies like OpenAI, they will drive increased demand for hardware as more businesses adopt them. For example, 100 companies using these models would require more chips than one large company.
2. Cloud providers are reluctant to cut back on CAPEX: In a competitive market, any reduction in spending by one provider could give others an advantage, so they are unlikely to make such cuts.
3. Open-source does not mean free usage: Companies will still need to rely on cloud services to run large models like Kimi3 (with 2.8 trillion parameters), meaning cloud data centers will continue to be fully utilized.
4. Kimi3 actually consumes more high-end chips: According to SemiAnalysis, Kimi3’s large number of parameters (1.5TB of memory) requires expensive Nvidia NVL72 systems (at least 64 chips). While it reduces bandwidth costs, it also doubles the demand for chips.
3. The Opponents: Ultimate Demand Is the Critical Factor
Opponents argue that end-users (not AI companies) are less willing to spend heavily on hardware:
- Ordinary businesses are more cautious with their spending; they will only buy what they need, not invest in unnecessary infrastructure.
- Large-scale adoption of AI is still a long way off; even if it becomes widespread, models will likely become as common as other technologies, and demand for hardware won’t be as intense.
- A domino effect in the industry chain: Reduced demand at the end-user level could affect all seven layers of the supply chain, and the Jevons Paradox (more efficiency leads to more usage) might not hold up under such pressure.
4. The Jevons Paradox: More Usage Doesn’t Necessarily Mean Higher Profits
The article cites the British Industrial Revolution as an example: although coal consumption increased, investments in railways and canals had low returns (4%-5%), reflecting the cost of social progress borne by investors.
In the AI industry, more chip usage doesn’t necessarily mean higher profits for hardware companies, as competition may dilute margins or demand growth might not keep up with investment costs. Investors need to consider both demand and potential returns.
5. Apple’s “Passive Profit” Strategy
Apple’s strategy illustrates how one can avoid the risks of AI investment:
- It didn’t rush into developing large models, avoiding costly hardware and data center investments.
- Its Mac Mini became popular as a platform for AI models during OpenAI’s rise.
- Despite China’s competitive AI efforts, Apple avoided the competition and saw its stock price soar, even surpassing Nvidia’s.
This shows that sometimes “missing out” on a trend can be a good strategy; patience and strategic decision-making are more important than following the crowd.
6. Investing Requires Research and Narrative Perception
The author emphasizes the need for careful analysis before investing, avoiding blind speculation. You should do your own research (DYOR) and pay attention to market consensus. When there is significant disagreement (high CFI, or “conflict of interest”), prices can be volatile, and you should base your decisions on logical analysis rather than panic.
In summary, while the AI industry is exciting, making money depends on understanding the underlying demand, the real state of the industry chain, and your ability to remain rational in times of market turmoil.