OpenAI Unveils Astra, Sparking AGI Debate and Renewed Conviction in Hardware
OpenAI's full rollout of its new model, GPT-6 Astra, prompted the market to begin interpreting it as an early model of the AGI (Artificial General Intelligence) era. AGI refers to artificial intelligence capable of learning and reasoning to solve a variety of problems on its own, much like a human, and unlike prior AI that performed only a single specific task, Astra was noted for demonstrating a more advanced level of interaction, such as proactively suggesting next steps to the person asking questions.
Nvidia CEO Jensen Huang stated on his Twitter account that Astra was trained on more than 100,000 Nvidia Grace Blackwell GPUs and that it took just four years to go from GPT-5.1 to Astra, remarking that AGI had already arrived. He also noted that an additional 400,000 GPUs would be deployed to OpenAI, reinforcing conviction around expanded hardware investment.
On this news, SK Hynix ADRs surged more than 8% on U.S. markets, and semiconductor-led gains also emerged in the domestic market. Doubts had persistently been raised over the lack of tangible results from massive AI investment, but the Astra unveiling was seen as evidence proving the correlation between expanded investment and actual performance gains.
Anthropic's initial public offering (IPO) was also discussed. The timeline, originally expected in August, is reportedly being adjusted to a mid-October roadshow with listing completion just before the U.S. midterm elections in November. Concerns were cited that a Democratic victory in the midterms could fuel anti-data-center sentiment, which would be unfavorable for IPO demand, as a factor behind the schedule adjustment, alongside concerns that the large capital raise could act as a black hole absorbing market liquidity.