Google says company must double AI serving capacity every 6 months to meet demand

During a recent all-hands meeting, Google’s AI infrastructure chief emphasized the urgent need for the company to double its AI serving capacity every six months. Amin Vahdat, the vice president at Google Cloud, shared insights on the escalating demand for AI services, revealing a vision for exponential growth over the next few years. Vahdat presented a slide that stated, "Now we must double every 6 months.... the next 1000x in 4-5 years." He underscored the fierce competition in AI infrastructure, highlighting it as both a critical and costly aspect of the ongoing AI race. This meeting took place shortly after Alphabet reported impressive third-quarter earnings, prompting an increase in its capital expenditures forecast to between $91 billion and $93 billion for the year. Google is not alone in this endeavor; competitors such as Microsoft, Amazon, and Meta have also raised their capital expenditure projections, collectively aiming to invest over $380 billion in AI this year. Vahdat clarified that while Google intends to invest significantly, the goal is to develop infrastructure that surpasses competitors in reliability, performance, and scalability, rather than merely outspending them. To enhance capacity, Vahdat noted that Google is focusing on building more efficient models and utilizing custom silicon. Recently, the company unveiled its seventh-generation Tensor Processing Unit, named Ironwood, which boasts nearly 30 times the power efficiency of its initial Cloud TPU from 2018. The company is also leveraging its DeepMind research to anticipate future AI models. Vahdat stressed the importance of delivering a thousand times more capability in computing, storage, and networking while maintaining cost and energy efficiency. He acknowledged the challenges ahead but expressed confidence in achieving these ambitious goals through collaboration and innovative design. Alphabet’s CEO, Sundar Pichai, also addressed employees during the meeting, indicating that 2026 will be a critical year amid rising AI competition and increased demand for cloud services. He responded to concerns about a potential AI bubble, a topic gaining traction in Silicon Valley. Pichai highlighted the importance of aggressive investment, pointing out that underinvestment carries significant risks. He revealed that Google’s cloud business experienced a remarkable 34% annual revenue growth, surpassing $15 billion in the last quarter, with a backlog reaching $155 billion. Despite the pressures, Pichai noted that Google is well-positioned to navigate challenges and maintain operational resilience. The discussion surrounding a potential AI bubble gained momentum ahead of Nvidia’s quarterly earnings report. Nvidia, a key player in the AI sector and a major Google customer, reported a 62% revenue increase, exceeding expectations. However, concerns lingered in the market, with shares of several AI companies, including Google, experiencing declines. Pichai mentioned that capacity limitations are hampering the rollout of new AI tools, such as the upgraded video generation tool Veo and the recently launched AI model, Gemini 3, which promises improved responses to complex queries. He reiterated that overcoming compute constraints remains a priority for Google. In response to inquiries about capital expenditures and free cash flow, CFO Anat Ashkenazi highlighted the vast opportunities ahead, particularly in migrating customers from physical data centers to the cloud. She expressed optimism about the significant prospects available to the company moving forward.