Musk says Grok Bot’s costs are ‘unsustainably high,’ while OpenAI admits Dots are ‘compute heavy.’ Is the personal AI agent push dead on arrival?

The push for personal AI agents has gained momentum, similar to Meta's Muse, but its sustainability is being questioned. Elon Musk revealed that running Grok Bot, introduced by SpaceX, costs are "unsustainably high relative to revenue," leading to compute allocation cuts. OpenAI's head of consumer business, Adam Fry, noted that Dots assistants, which execute tasks for users, are "on the compute heavy side."

Each Dot assistant has its own cloud computer, offering capabilities like email management, event ticket booking, and more. Fry also hinted at enabling Dots to operate in the "background mode," making them more persistent. OpenAI is investing in voice AI models and developing devices by Jony Ive for a powerful consumer experience, but the success of these initiatives depends on finding enough compute.

Today, Dots are restricted to Pro subscribers at $100 per month; wider availability demands a breakthrough in compute efficiency. As personal AI agents grow more powerful and constantly running, they will consume more compute. OpenAI has announced a system to route tasks to less compute-intensive models, such as smaller GPT-6.1 Terra, which will replace the original GPT-6 Astra.

However, securing more compute remains a significant challenge, as highlighted by Anthropic's reported $895.4 billion investment in compute infrastructure. OpenAI, like Anthropic, faces similar compute constraints, despite its reliance on Nvidia GPUs and efforts to develop its own chip, Jalapeno, by 2027. Private funding may be insufficient, with OpenAI's latest round raising $11 billion, falling short of the $30 billion target. Public funding could be the only solution, but success remains uncertain.