Research can be highly beneficial to Google, but ultimately, it needs to contribute to the company’s profitability. Currently, customers are not generally inclined to pay directly for AI features, leading Google to consider selling ads within the Gemini app. This approach aligns with Google’s longstanding strategy, which involves collecting user data, time, and attention, and offering tools at no direct cost to users in exchange for compliance with their terms of service.
According to Sensor Tower data, OpenAI’s ChatGPT has amassed approximately 600 million global app installs, significantly outpacing Google’s 140 million installs for the Gemini app. The AI landscape features several competitors including Claude, Copilot, Grok, DeepSeek, Llama, and Perplexity, many of which are supported by Google’s well-established rivals. These systems demand considerable investment and energy resources, yet have not yielded substantial financial returns or advanced eco-friendly practices. Companies in the AI sector focus on enhancing system efficiencies and reducing errors to attract more users, but a sustainable revenue model remains elusive.
Google contends with an additional challenge that its competitors do not face: potential losses in search ad revenue due to antitrust judgments, which could amount to up to a quarter of this income, as noted by JP Morgan analyst Doug Anmuth. This creates pressure within the company to secure alternative revenue streams. Google’s staff, including those working on Gemini, have reportedly worked through holidays for consecutive years to maintain progress. Co-founder Sergey Brin has indicated that a 60-hour workweek is optimal for maintaining productivity amid intensified competition in AI development. Concerns regarding layoffs, burnout, and legal issues linger among employees, as reported by WIRED.
Within Google, a prevailing sense of unease exists despite the benefits of generative AI. Regulatory bodies, like that of France, are gradually embracing the potential of such technologies. DeepMind, headed by Demis Hassabis, persists in its ambition to achieve artificial general intelligence (AGI), a comprehensive system akin to human cognition across diverse tasks. Hassabis is sometimes seen exploring London with an Astra prototype, envisioning a future where the physical realm becomes entirely searchable. AGI development demands advancements in reasoning, planning, and autonomy.
In January, OpenAI introduced another experimental project, the Operator service, an agentic AI capable of interacting with websites to perform tasks such as booking trips or completing forms. Although it operates more slowly than a human and carries a high cost due to reliability issues (offered at $200 per month), it signifies a step toward more autonomous AI systems. Google is concurrently working on incorporating agentic features in future models, with the goal of evolving capabilities from generating meal plans to assisting with online shopping, and eventually providing feedback on simple tasks.
Moving swiftly might lead to occasional errors, as evidenced by a mistake made by Google’s Gemini in estimating global cheese consumption ahead of the Super Bowl. As Gemini transitions from a factual tool to a more integrated life assistant, Google CEO Sundar Pichai emphasizes caution in development. After regaining a competitive position, Pichai and other executives aim to maintain their lead in the AI race as it continues.
