AI helped save the chip industry. What if it turns out to be bust?

What will happen if AI does not work?

This is not just a hypothetical question. Some signs indicate AI growth indecentor at least slowing down. The new models are no longer show significant developments from the size of the size or the amount of training information. Nobel Laureate Demis Hassabis recently celebrated This is over the development of the EU “we don’t get the same progress. Andreessen Horowitz is one of the most prominent investors in the EU, in the same way Shared concerns It turned out that the AI model opportunities were published.

The models that can already have a reason for the slowdown performance of the EU are already consuming the most remaining digital data for the most developing. The developers become synthetic data instead, but can be more effective and can even do Models are worse.

The development of the AI is also great capital intensively. The most advanced models of training require billions of dollars. Even a training run may cost ten million dollars. Again, while the development costs continue, cash prizes are limited. In addition to AI coding assistants, there are Several examples These are the Creator of the AI, which justifies large capital investments.

Some companies are already back in infrastructure investments due to cost. For example, Microsoft “to slow down or stop Some early-stage projects “and canceled equipment for several global data center projects. There are all meta, AWS and Google It was reported to have something Cut GPU orders. Chip gardens, power deficiency and public concerns are also a massive AI adoption barriers.

AI Boom Peters is bad news for the chip industry that uses this new technology to avoid a serious slip.

It becomes more expensive to make chips. The preparation of new production processes is billions of dollars; The establishment of new plants may cost ten billions of dollars. These costs are all transferred to consumers, but customers outside AI are eager to get more expensive chips. Fancy technologies are not useful for other purposes in today’s AI processors.

AI has postponed the calculation of an industry: production becomes more expensive, and performance gains are shrinking. EU economic promises justifies high chip prices, but if it moves away, the chip industry needs to find something else to convince people to invest in developed chip production. Otherwise, developed chips, unstable: New technologies will cost more and less while delivering less.

A chip industry shift will have several geopolitical and economic goals. Governments spilled billions of dollars to the establishment of the internal chip industry. US President Donald Trump threatens regularly Use tariffs To produce semiconductor to return home.

The probable lead in which the United States has taken to the development of chip can be a miracle as especially China is dominated by the manufacturing of heritage chip. And the AI reverse would shake the world’s technical sector, and would force great technology to re-thinking their bets.

Taking into account these stakes, politicians should encourage another innovation in AI by facilitating easy access to information, plugs, power and cooling. This includes a balanced approach to chip production on land and sea and pragmatic policies to overcome adjustment barriers for energy use and generation. Governments should not apply the reserve measures to AI; Benefits are very large to solve the development of at least in this early stages. Like autonomous vehicles or home robotots, large-scale AI applications should face baseless high demands for execution.

Investors should also investigate the new AI growth, potentially not require a new information and infrastructure. If industrial, you want to manage their risks, you should also investigate non-AI applications for chips.

To ensure the slowing of the chip industry, must reduce the advanced chip value. Companies must work together to work and reduce development costs, as well as working with universities. In chiplets, you need more investment in advanced packaging and reconstructed devices. It should support industry, mutual standards, open source tools and flexible device development. Shared, subsidized infrastructure for design and fabrication can help you complete the thoughts before the production of small companies. However, most importantly, the driver of the land, the driver may be counter-results: so much carelessness will significantly increase the cost of chip.

The future of chips and AI is now deeply combined. If the chips should develop, AI should grow. If not, the whole chip sector can now be in danger.

Reviews expressed in Fortune.com are the opinions of the authors and do not necessarily reflect the ideas and beliefs Fortune.