Google Gemini Pro delayed by months as coding falls short

Alphabet shares slipped more than three percent after reports emerged that Google is months behind schedule on delivering the next version of its flagship AI model, Gemini Pro. The delay stems from the technology's failure to meet internal coding benchmarks, a significant weakness in a market where competing models from Anthropic, OpenAI, and Meta are rapidly advancing.

Google had been widely expected to release the upgrade at its May developer conference. Instead, an attempt late last month to update the model's training data yielded disappointing results. A Google spokesperson defended the company's pace, stating it is “shipping quickly across a wide range of models” and testing the upgraded Pro alongside a new Flash model with partners.

The root of the problem appears to be internal fragmentation. Google Cloud, DeepMind, and the Android team are all developing their own AI coding tools, creating overlapping efforts that have slowed progress. Co-founder Sergey Brin has intervened to push for faster development, but former employees say he has been hampered by competing factions and by engineers who argue critical code must still be written by humans.

To address the chaos, Chief AI Architect Koray Kavukcuoglu is attempting to unite the company's internal coding tools. A new DeepMind team led by Sebastian Borgeaud has also been formed to focus on the issue. Google claims 75 percent of its internal code is now AI-generated and has consolidated its developer tooling under a platform called Antigravity. However, internal competition for computing power means engineers often hit capacity constraints, a bottleneck that external cloud customers also face.

For European businesses building on Google's infrastructure, the delays carry real costs. Customer experiences with the current Flash model vary wildly. Rodrigo Davies, a product manager at Figma, said the model hit “a sweet spot of speed and quality.” Conversely, Freddy Vega, CEO of education platform Platzi, found the Flash model “more expensive and slower than its predecessor while remaining far less capable than competitors,” prompting his team to switch to Anthropic.

The technical struggles have triggered an exodus of senior engineers to rival labs. As AI becomes foundational to European digital infrastructure, Google's inability to ship competitive coding models risks ceding ground to OpenAI and Anthropic at a critical juncture.