Google Says AI Capabilities Need to Grow a Thousand-Fold by 2030

• Google executives have said they need to double their AI infrastructure every six months.
• The company is facing limitations in chips, energy, and storage networks.
• Google is relying on its own TPUs and optimization models to achieve scaling.

---

Google Corp. is facing an unprecedented increase in strain on its artificial intelligence infrastructure. According to the management, the company is forced to double computing power every six months.

This was announced by Google Cloud Vice President Amin Vahdat at a general meeting of employees. He emphasized that the company has to scale “another 1,000 times in 4–5 years,” CNBC reported.

According to Vahdat, the problem rests not only on the need to ramp up performance, but also on maintaining the same costs and energy consumption levels. He noted that Google is obliged to build an infrastructure that is “more reliable, performant and scalable” than the solutions available today.

At the same time, he said, the corporation must avoid a spike in the cost of operating data centers.

The growth in demand for AI remains a key limiting factor, the top executive noted. In this case, analysts are still unclear to what extent it is formed by real user activity, and to what extent — the integration of AI-functions in Google services.

OpenAI, which is building six data centers in the U.S. for its $400+ billion Stargate project, is facing similar infrastructure challenges.

An additional hurdle remains a global shortage of Nvidia GPUs needed to train and output AI models. In a recent quarterly report, the company said its chips are “sold out,” and the shortage is already affecting Google’s technology adoption schedule.

Google CEO Sundar Pichai cited the example of the Veo tool, which failed to expand to more users due to capacity limits.

To reduce its reliance on Nvidia, Google is betting on its own silicon designs. In November, the company unveiled its seventh-generation TPU processors, which it says are nearly 30 times more energy efficient than the first version.

In addition, Google plans to optimize model architectures and expand its physical data center infrastructure.

Pichai warned employees that 2026 will be a “busy year” due to competition and the need to meet demand for cloud services. He admitted that talk about the overheating of the AI market is “definitely relevant”, but the company is ready to continue to accelerate infrastructure development.