The AI compute arms race isn’t over yet! Could the next wealth opportunity be electricity and infrastructure?

Many people think competition in the AI industry is just a battle of models.
In reality, the real war may be happening in the data centers behind the models.
Training models requires compute, and running them requires compute too. The more users there are, the greater the inference demand, and compute consumption may continue to rise.
This creates an interesting business cycle: the more successful AI products are, the greater the demand for infrastructure; the more users there are, the more apparent the cost pressures become.
OpenAI’s revenue growth and accelerating enterprise business show that AI demand is being commercialized. But for the entire industry chain, this is just the beginning of the story.
GPUs, data centers, optical modules, networking equipment, energy storage, and power supply could all become important parts of AI expansion.
In the future, competition between AI companies may not be about whose model is the smartest, but about who can provide equally excellent services at a lower cost.
After all, models can be optimized through algorithms, but your electricity bill won’t automatically get discounted just because you wrote an excellent technical paper.
I believe AI will continue to pursue scale, but compute efficiency will become an important factor determining profits.
The next AI wealth opportunity may lie not only with model companies, but also with those who provide infrastructure for AI. $NAS100

‌#OpenAI第三季度营收升至100亿美元