OpenAI executives discuss the arrival of artificial general intelligence while Jensen Huang claims the era has already begun. Simultaneously, researchers who helped build current models are departing OpenAI, Google DeepMind, and Meta to launch independent laboratories. These new groups raise billions of dollars without releasing commercial products. The founders argue that the standard training recipe cannot solve the remaining technical hurdles. We have catalogued 45 of these new labs and mapped their specific research directions. The analysis identifies the current limitations each group aims to overcome. Nvidia consistently secures the necessary hardware for these experiments. This shift indicates a belief that scaling alone will not produce the next breakthrough.
The movement suggests a strategic pivot toward specialised research rather than general product rollout. Investors fund these labs to test hypotheses about model efficiency and reasoning capabilities. Existing infrastructure at major tech companies may no longer support the specific needs of these new ventures. The departure of key talent signals a desire to control the development trajectory directly.
* Labs focus on overcoming current model limitations rather than immediate product features.
* Funding sources prioritise long-term research over short-term commercial returns.
* Nvidia hardware remains central to enabling these independent research efforts.
