The world of artificial intelligence is no longer just about developing new and more advanced models. High-quality infrastructure that ensures their operation is playing an increasingly important role. This infrastructure is responsible for responding to user queries in real time and meeting their demands. According to CNBC, Google must now double its AI infrastructure every six months just to keep up with the growing demand for its AI services.
This information was shared at a Google employee meeting held in early November. It was presented by Amin Vahdat, vice president at Google Cloud, who added that he believes Google has reached a point where demand for artificial intelligence is higher than ever before. Moreover, interest is growing not only among ordinary users but also among businesses and developers.
Serving takes center stage, not training
It is important to emphasize that the infrastructure capacity Amin Vahdat discussed during the presentation does not concern training new models. It refers to so-called “serving capacity,” meaning the operational capacity that handles AI queries and generates responses, images, or code according to user instructions. Its rate of growth is truly extreme.
CNBC stated that if Google is to keep pace with growing demand over the long term, it will have to increase its AI capacity by as much as a thousandfold by the end of this decade. That is a figure no one could have imagined just a few years ago.
Billions in investments and proprietary chips
It is therefore no surprise that such growth requires massive investment in expansion. Parent company Alphabet has therefore raised its capital expenditure forecast to more than 90 billion dollars. The funds will go toward expanding data centers and servers and developing hardware. According to CNBC, companies such as Meta, Microsoft, and Amazon face a similar scenario.
Moreover, Google does not want to rely exclusively on external suppliers in the future. It therefore recently unveiled a new generation of its proprietary AI chips, TPU Ironwood, which offer not only better energy efficiency but also substantially higher performance than its 2018 chips.
In his speech, Vahdat emphasized that he understands how much investment building the new infrastructure will require. However, he added that this is the only way to make the entire operation more reliable, more powerful, and less dependent on purchases from third parties. And that is precisely what should help Google handle the growing workload without continually increasing costs.
Better to overinvest than fall behind
Alphabet CEO Sundar Pichai now sees investment as crucial. In his view, it is better to invest more money than to underestimate overall spending. Otherwise, there is a risk that the company will be unable to keep up with capacity demands, creating a so-called “AI bubble”. And because of that, the company could very quickly lose its competitive advantage.
Google's actions and growing investments show that keeping pace with the demands placed on AI tools is not just about training ever-better models. Their actual operation and ability to handle a wide range of tasks in real time also play a significant role. However, through its decisions, Google is making it clear that it certainly has no plans to slow down in the AI race.
Source: CNBC
