Google reserves TPUs for AGI as spending surges

Alphabet is putting its own artificial general intelligence ambitions ahead of selling every available tensor processing unit to customers. CEO Sundar Pichai said the company’s first priority is reserving enough TPU capacity to support frontier AGI development, even as demand for Google’s accelerators and cloud services accelerates.

Pichai outlined the policy during Alphabet’s second-quarter earnings call, after Goldman Sachs analyst Eric Sheridan asked how the company balances customer demand for TPUs with its own need for computing power.

“In terms of allocating our TPUs … our first priority is making sure we are allocating what we need to compete at the frontier in terms of AGI development.”

“That is the foundation for everything we do,” Pichai added, referring to artificial general intelligence as AI with human-like intelligence.

How Google is allocating compute

Bernstein analyst Mark Shmulik later asked how Google divides capacity among Search, Cloud, and model training. Pichai said the baseline remains the compute required to continue AGI development at the frontier. Alphabet is also prioritizing core products including Search and YouTube, along with Cloud.

Within Google Cloud, capacity is being directed toward serving models through Vertex, Gemini Enterprise, data analytics, cybersecurity, and other core solutions. Pichai said computing power is primarily going to the company’s consumer and enterprise services.

Recommended reading

Aronofsky’s AI studio raises $11 million

The demand is reflected in Google Cloud’s results. Revenue jumped 82 percent year over year to $24.75 billion, while operating profit reached $8.8 billion, up 214 percent. Alphabet attributed the growth to strong demand for AI infrastructure and AI solutions. Google Cloud’s contracted backlog also reached $514 billion, representing services customers have signed up for but have not yet consumed.

Search revenue increased 17 percent, while YouTube advertising revenue rose 13 percent. Pichai said the company’s AI Mode for Search is driving an incremental increase in overall Search queries, countering predictions that generative AI would undermine Google’s search advertising business.

Google turns to third-party capacity

AI-powered Search requires specialized hardware that is expensive to acquire and operate. Pichai said engineering and hardware optimizations have reduced the cost of AI Mode responses to their lowest level since launch, despite the addition of more advanced capabilities.

Alphabet CFO Anat Ashkenazi said the company now expects to spend between $195 billion and $205 billion during the financial year, up from an earlier forecast of $180 billion to $190 billion. But Google remains unable to obtain all the hardware it wants because of what Ashkenazi called a “supply-constrained environment.”

The company plans to expand its use of third-party capacity in the third quarter as a temporary bridge while building more internal capacity. That spending could help Google secure very large cloud customers, Pichai said, even if the short-term cost is high.

“The incremental opportunities they are bringing to us, while a short-term cost over a few months may be very high, in the lifetime of the deal, as we bring more capacity on, is highly ROI positive.”

Alphabet reported quarterly revenue of $119.8 billion, up 24 percent, and operating income of $40.8 billion, up 34 percent. Still, free cash flow fell to -$5.9 billion—the company’s first negative result since 2004—and shares dropped four percent in after-hours trading.