Details behind the Gemini delay revealed: computing power constraints and internal disagreements over development priorities and resource allocation.

Reports indicate a delay for the next-generation Gemini model. The launch of this flagship model has been postponed by two months due to internal disagreements regarding development priorities and resource allocation, compounded by tight computing power supplies and cumbersome approval processes.

Regarding computing power, despite possessing proprietary TPU chips, the company faces strained resources as it attempts to meet the simultaneous demands of model training, cloud services, and a vast array of consumer and enterprise AI products. Senior leadership has recently initiated organizational restructuring to alleviate resource conflicts between departments.

Meanwhile, shifts in senior leadership and management authority are underway. Co-founder Sergey Brin has recently become deeply involved in core model training, steering resources toward "recursive self-improvement." Additionally, a management reorganization has curtailed the autonomy of the research labs, with key leader Koray Kavukcuoglu assuming full control over Gemini's development trajectory. Amidst the departure of some key technical staff, the company is reconsolidating scattered resources to go all-out in catching up with the pace of cutting-edge AI development.