NVIDIA is now the first More than $4 trillion surge In market cap, it rebounded earlier this year from the downturn caused by its deep index. Other AI chip manufacturers, including AMD and China’s Huawei, reported good financial results. Now, almost every major chip manufacturer is now centering its strategy on AI.
But what if AI can’t solve it?
This is not only a matter of assumption. Some signs that AI growth is Stallor at least slow down. The new model no longer shows significant improvements in scale-up or training data volume. Nobel Prize winner Demis Hassabis Recently pointed out In terms of AI development, “we are no longer making the same progress.” Anreessen Horowitz, one of the most outstanding investors in AI, also Focus on The AI model functionality seems to be smooth.
One of the reasons for the slowdown in AI’s performance is that the model has consumed most of the digital data available, and there is little left to improve further. Instead, developers are turning to synthetic data, but may be less efficient and may even make The model is worse.
Artificial intelligence development is also a large amount of capital-intensive. Training the most advanced models requires computing clusters, which cost billions of dollars. Even a training session can cost tens of millions of dollars. However, despite the rising cost of development, monetary rewards are limited. In addition to AI coding assistant, there is Several examples AI generates returns, justifying these huge capital investments.
Some companies are already expanding their investment in AI infrastructure due to costs. For example, Microsoft isSlow down or pause Some early stage projects” and canceled equipment orders for several global data center projects. It is said that Cut off their GPU orders. Chip bottlenecks, power shortages and public attention are also obstacles to Volkswagen’s AI adoption.
If AI BOOM is emitting, this is bad news for the Chip industry, which has used this new technology to avoid a serious downturn.
The chips are becoming more and more expensive. Developing new manufacturing processes costs billions of dollars; building a new factory can cost tens of billions of dollars. All of these fees are passed to consumers, but outside of AI, customers are not keen on buying more expensive chips. The exquisite technology in today’s AI processors is not that useful for other purposes.
Artificial intelligence has delayed industry estimates: manufacturing is becoming more and more expensive, while performance improvements are shrinking. The economic commitment of AI is reasonable, it is reasonable chip prices, but if that goes away, the chip industry needs to find something else to convince people to maintain investment in advanced chip manufacturing. Otherwise, advanced chip manufacturing will become unsustainable: new technologies will cost more and more, while delivering less and less.
The chip industry downturn will upend several geopolitical and economic goals. The government poured Billions of dollars Establish a domestic chip industry. US President Donald Trump often threatens Use tariff Bring semiconductor manufacturing home.
The so-called chip development field in the United States may prove to be Haishi Rage Building, especially China rules chip production. AI reversals will shake the world’s technology sector, forcing Big Tech to reconsider its bets.
Given these bets, decision makers need to encourage further innovation in AI by facilitating easier access to data, chips, power supplies and cooling. This includes pragmatic policies on copyright and data protection, a balanced approach to chip manufacturing onshore and offshore, and the elimination of regulatory barriers to energy use and power generation. Governments should not necessarily apply the principle of prevention to AI; at least in these early stages, this benefit is too great to be disabled. Large-scale AI applications such as autonomous cars or home robots should also not face unreasonably high demands for implementation.
Investors should also explore alternative AI approaches that do not require much data and infrastructure, which may unleash new AI growth. If it is just to manage its risks, the industry must also explore applications for non-AI chips.
To ensure the chip industry can slow down, it must reduce advanced debris costs. The company should work together in research and development and cooperation with universities to reduce development costs. More investment is needed in chips, advanced packaging and reconfigurable hardware. The industry must support interoperable standard, open source tools and agile hardware development. Shared, subsidized design and manufacturing infrastructure can help smaller companies finalize their ideas before they are made. But it is important that the momentum to head to onshore manufacturing can backfire: doing so carefully will greatly increase chip costs.
The future of chips and artificial intelligence is now deeply intertwined. If the chips are to reproduce, AI must grow. If not, the entire chip field could be in danger now.
Comments expressed on fortune.com are entirely the opinions of its author and do not necessarily reflect opinions and beliefs wealth.
