Google CEO Pichai says Gemini's next leap depends on building "much larger base models"

Editor's take

Sundar Pichai indicated that Google's Gemini model's future advancements hinge on developing significantly larger foundational models. This statement, coupled with Alphabet's increased 2026 capital expenditure forecast to $205 billion, signals a continued massive investment in AI infrastructure and research, particularly within Google Cloud, which saw 82% growth in Q2.

This focus on "much larger base models" suggests a belief that current model sizes are approaching a plateau for certain capabilities, necessitating a scaling-up of parameters and training data to unlock new levels of performance. This approach directly impacts the competitive landscape, forcing rivals like OpenAI and Meta to consider similar, resource-intensive scaling strategies to keep pace.

The key question moving forward is whether simply increasing model size will yield diminishing returns or unlock novel emergent abilities. Investors and competitors will be watching to see if these larger models translate into tangible product improvements beyond incremental gains, and what the associated computational and environmental costs will be.