In the face of skepticism about the AI bubble, Google's AI infrastructure head admitted: the company must double its AI computing power every six months to meet demand

Amin Vahdat, head of Google Cloud AI infrastructure, revealed at the all-hands meeting on November 6 that in order to meet the demand for artificial intelligence services, the company must double its computing power every six months and expects to achieve a 1000-fold increase in computing power over the next 4-5 years.

According to media reports, Vahdat emphasized at the meeting that "the competition for AI infrastructure is the most critical and also the most expensive part of the AI race." This year, Google's parent company Alphabet has raised its capital expenditure expectations twice to between $91 billion and $93 billion, and has forecasted a "significant increase" in investment by 2026. The four major cloud service providers, including Microsoft, Amazon, and Meta, are expected to collectively spend over $380 billion this year.

At the same time, Alphabet CEO Sundar Pichai acknowledged concerns about an AI bubble but emphasized that the risk of underinvestment is greater. He pointed out that if there were more computing power support, Google Cloud's performance could be even better. Financial reports show that the business's revenue grew by 34% year-on-year in the third quarter, exceeding $15 billion.

Computing Power Bottleneck Restricts Product Promotion

During the all-hands meeting, Vahdat clearly stated the goal of doubling computing power every six months. He said, Google's goal is not to outspend competitors but to provide "more reliable, higher performance, and more scalable" infrastructure.

At the same time, Pichai used the video generation tool Veo as an example to illustrate that the supply of computing power is the bottleneck. He stated:

"Veo was very exciting when it was released; if we could provide this service to more people in the Gemini application, I think we would gain more users, but we can't do that due to computing power limitations."

In addition to infrastructure construction, Vahdat stated that Google is also enhancing processing capabilities through more efficient models and custom chips. Last week, Google officially announced the release of the seventh-generation Tensor Processing Unit (TPU) Ironwood, which the company claims has improved energy efficiency by nearly 30 times compared to the first cloud TPU launched in 2018.

Executives Acknowledge AI Bubble Risks

In response to employee questions about the AI bubble, Pichai acknowledged the validity of market concerns. Employee questions involved "in the context of significant AI investments and market discussions about a potential AI bubble burst, how to ensure long-term sustainability and profitability if the AI market does not mature as expected."

In this regard, Pichai reiterated his previous point, that the risk lies in insufficient investment. He stated:

“I acknowledge that it is always difficult at these moments because the risk of underinvestment is very high. In fact, I believe that although the data for cloud computing is excellent, if we had more computing resources, that data would be even better.”

At the same time, he highlighted Google's cloud computing business. This business recently reported a quarterly revenue growth of 34%, exceeding $15 billion, with backlogged orders reaching $155 billion. He stated that the company adheres to a rigorous business approach and pointed out that the company's core business strength is robust, with a solid balance sheet. He stated:

"Compared to other companies, we are more capable of withstanding failure."

Surge in Capital Expenditure Raises Concerns

Another employee pointedly asked a core financial question: "The pace of capital expenditure acceleration significantly outpaces our revenue growth," inquiring about the company's strategy to maintain "healthy free cash flow" in the next 18-24 months.

CFO Anat Ashkenazi responded that the company has multiple growth prospects, including migrating more customers from physical data centers to the cloud. She stated, "The opportunities in front of us are enormous, and we cannot miss this momentum."

This week, Alphabet also released its latest AI model, Gemini 3, claiming it can provide better answers to more complex problems, continuing to compete with AI companies like OpenAI and striving to promote advanced AI tools to more users