- GPT-6 Astra reportedly trained on a huge NVIDIA computing infrastructure.
- NVIDIA connected 72 chips in a single system, creating a massive AI computing platform.
- Jensen Huang suggested another 400,000 chips are preparing to go online.
AGI, Artificial General Intelligence, is possibly getting closer, at least according to NVIDIA Chief Jensen Huang. The new insights he provided concerning OpenAI’s GPT-6 Astra brought up how fast AI models are developing and how much computing power is now needed to create them.
OpenAI GPT-6 Astra pushes AI hardware to a new scale
Perhaps even more crucial than the model itself is the hardware infrastructure supporting GPT-6 Astra. As the company continues to push the envelope past its classic architecture, which usually comprises a relatively smaller number of GPUs, it now offers an architecture that can link 72 chips into one system, enabling artificial intelligence companies to process greater amounts of work.
In his original claim, Huang said that 300,000 systems like the one NVIDIA offers were utilized for the training of GPT-6 Astra. Huang later had to change that to 100,000 systems. Why this reduction was made is unclear since NVIDIA never explained it to the public. This demonstrates how hard it is to estimate the scale of modern AI training initiatives.
NVIDIA revenues are being boosted by growing AI adoption
The vast appetite for AI computing resources is also being reflected in the massive revenue growth of NVIDIA. According to some reports, NVIDIA made revenues totaling $89 billion from computer sales for the purpose of AI in just three months, which is significantly higher than last year’s numbers.
For NVIDIA, the story is relatively straightforward—the need for better AI systems leads to a vast demand for computing resources.
GPT-6 Astra—Is this AGI?
Here lies the question mark. While Huang stated that AI made a huge breakthrough, OpenAI did not officially state that GPT-6 Astra is an example of AGI. Moreover, Greg Brockman from OpenAI also did not give any official comment regarding this statement.
Therefore, it can be regarded as a very important statement from the industry point of view but not an officially confirmed breakthrough. Moreover, Huang noted that another 400,000 chips will appear soon. Whether GPT-6 Astra really means the emergence of human-like AI or just another step towards this achievement, the scale of equipment deployment demonstrated how fast the AI race takes place.
Overall, not only has GPT-6 Astra attracted attention due to its alleged abilities, but also due to the enormous computer resources necessary to train such an AI system. The coming months will show whether all these enormous expenses will bring breakthroughs in the development of AGI that so many expect.
Keep up with the tech that actually matters.
Add us as a preferred source on Google Search for quicker access to our coverage.
