The company states that its new experimental models surpass those of China’s DeepSeek

The tech giant Google announced that the company plans to invest 75 billion dollars in capital expenditures in 2025, aimed at developing technical infrastructures related to artificial intelligence (AI), such as servers and data centers, as it needs more “capacity” to meet demand.

“We’ve had very strong demand for our AI products in the fourth quarter and ended the year with more demand than available capacity, so we’re in a very tight supply-demand situation, working hard to activate more capacity,” said Alphabet CEO Sundar Pichai on Tuesday.

The executive stated that their new experimental artificial intelligence (AI) models outperform those of the Chinese company DeepSeek and that they will increase capital expenditures to absorb the “strong demand.”

In the conference to analyze results with analysts, the executive explained that Alphabet’s AI model, called Gemini, stands out in “cost, performance, and latency” and “leads the Pareto frontier,” an economic concept that refers to the efficiency equilibrium point.

“I’d say our Gemini 2.0 Flash and 2.0 Flash Thinking models are some of the most efficient out there, including when compared with DeepSeek’s V3 and R1,” said Pichai, who nonetheless congratulated the “tremendous team” at DeepSeek, which has done a “very good job.”

In the last quarter of 2024, Google Cloud, the cloud service that incorporates AI, posted revenues of 11.955 billion, a 30% year-over-year increase. However, this figure represents slower growth compared to other quarters, which executives attributed to limitations in meeting demand.

Pichai also predicted that tech companies are increasing spending on inference for their AI models and said he expects “reasoning models,” which can use logic to reach conclusions, to further accelerate development.

“That’s part of the reason why we’re so excited about the AI opportunity. We know we can drive extraordinary use cases because the actual cost of using AI will continue to drop, making more use cases feasible. The opportunity space is enormous,” he added.