Google CEO downplays DeepSeek threat as the company plans to spend $75 billion on AI this year
Feb 05, 2025 - businessinsider.com
Google CEO Sundar Pichai stated that the company's AI models are competitive with those of the Chinese startup DeepSeek, emphasizing their cost efficiency and performance. During Google's Q4 earnings call, Pichai highlighted the efficiency of Google's Gemini models, which excel in cost, performance, and latency. He downplayed DeepSeek's impact on Alphabet's business, despite the startup's AI model being trained at a lower cost compared to others like OpenAI's. Pichai noted that reasoning models are driving increased spending on inference, which is the process of making predictions or conclusions with a trained AI model.
Google announced a planned $75 billion in capital expenditure for the year, surpassing analysts' expectations of $57.9 billion, which led to an 8% drop in shares during postmarket trading. Pichai expressed excitement about the potential of AI, citing the decreasing cost of usage as a factor that will enable more feasible use cases. He praised DeepSeek's team but maintained that Google's models lead on the Pareto frontier, thanks to their full-stack development and cost-per-query optimization.
Key takeaways:
Google's AI models are comparable to DeepSeek's in terms of cost efficiency, according to CEO Sundar Pichai.
Google plans to spend $75 billion in capital expenditure this year, exceeding analyst expectations and impacting share prices negatively.
Pichai highlighted Google's Gemini models as leading on the Pareto frontier for cost, performance, and latency.
Reasoning models are driving increased spending on inference, with Pichai expressing optimism about the expanding AI opportunity space.
