Google’s AI infrastructure chief told employees that the company must double its compute capacity every six months to keep up with rising demand for artificial intelligence services. The internal message came during a Nov. 6 all-hands meeting, where executives discussed the scale of required investments and the competition shaping the AI infrastructure landscape.
Internal Presentation Highlights Demand Growth
Amin Vahdat, a Google Cloud vice president, presented a deck titled “AI Infrastructure,” which CNBC reviewed. One slide on “AI compute demand” stated, “Now we must double every 6 months…. the next 1000x in 4-5 years.”
Vahdat said, “The competition in AI infrastructure is the most critical and also the most expensive part of the AI race.” Alphabet CEO Sundar Pichai and CFO Anat Ashkenazi joined the meeting to answer employee questions.
The presentation followed Alphabet’s stronger-than-expected third-quarter results and a raised capital expenditures outlook. The company now expects capex to reach between $91 billion and $93 billion this year, with a “significant increase” planned for 2026. Microsoft, Amazon and Meta also boosted their capex forecasts, and the four firms collectively project more than $380 billion in spending this year.
Balancing Investment and Performance Goals
Vahdat said Google’s responsibility is to build the needed infrastructure but not necessarily to outspend every rival. He said the objective is to provide systems that are more reliable, performant and scalable than anything else available.
To increase capacity, Google uses infrastructure expansion, more efficient AI models and its custom chips. Last week, the company announced the public launch of its seventh-generation Tensor Processing Unit, Ironwood, which it says is nearly 30 times more power efficient than the first Cloud TPU released in 2018.
Vahdat also said Google gains an advantage from DeepMind’s research into how AI models may evolve over coming years.
He told employees that Google must “be able to deliver 1,000 times more capability, compute, storage networking for essentially the same cost and increasingly, the same power, the same energy level,” adding that the company expects to advance through collaboration and co-design.
Pichai Addresses Market Concerns and AI Competition
Pichai said 2026 will be “intense,” citing competitive pressure and infrastructure demands. He answered an employee question about the possibility of an AI bubble, referencing growing skepticism among investors about whether projected industry spending can be sustained.
The employee’s question asked how Google plans to maintain long-term profitability if the AI market does not mature as expected. Pichai said the concern has been part of recent discussions and that underinvesting carries its own risks. He pointed to Google Cloud, which recorded 34 percent annual revenue growth to more than $15 billion in the quarter, with a backlog of $155 billion.
Pichai said the company uses a disciplined approach and is positioned to handle market volatility. He added that next year will include “ups and downs” and described the competitive pressure as substantial.
Google declined to comment on the meeting.
Market Activity and Industry Reactions
The discussion around a possible AI bubble intensified in the days leading up to Nvidia’s earnings report. Shares of companies that benefited from earlier AI momentum, including CoreWeave and Oracle, dropped sharply during the month.
Earlier this week, Pichai told the BBC that the market shows “elements of irrationality” and said that if a bubble were to burst, “no company is going to be immune, including us.”
Nvidia CEO Jensen Huang pushed back on the bubble narrative during Wednesday’s earnings call, saying, “We see something very different.” Nvidia, which supplies chips to Google, reported 62 percent revenue growth, beating expectations, and issued strong guidance for the fourth quarter.
Despite the results, markets declined on Thursday. Nvidia fell 3.2 percent, the Nasdaq dropped 2.2 percent, and Alphabet shares slid 1.2 percent.
AI Model Launches and Capacity Constraints
Google released its newest AI model, Gemini 3, earlier this week. The company said the model provides improved answers to more complex queries. Google is competing with firms such as OpenAI to get advanced tools to the largest possible user base.
Pichai said compute capacity limits remain a bottleneck. He referenced the company’s upgraded video generation tool, Veo, which launched last month. He said Google could not offer Veo to more users through the Gemini app because of the company’s infrastructure limits.
Questions on Capex Acceleration and Financial Strategy
Another employee asked why capital expenditures are growing faster than operating income and what the strategy is for maintaining healthy free cash flow over the next 18 to 24 months.
Ashkenazi said the company has several opportunities ahead, including migrating more customers from physical data centers to cloud services. She said the company sees momentum and does not want to miss it.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.