Nvidia CEO declares 'AGI has arrived' via GPT-6 Astra as Zitron reports Stargate Abilene still roughly half-built
2026-09-06
On September 6, 2026, Nvidia CEO Jensen Huang posted on X that "AGI has arrived," crediting GPT-6 Astra and calling Abilene "the birthplace of AGI," adding "400K GPUs coming online next." Per Ben Thompson's Stratechery, this echoes earlier statements attributed to OpenAI Chief Research Officer Mark Chen (told TIME OpenAI was "80% of the way to AGI") and OpenAI COO Greg Brockman (said OpenAI had entered "the AGI era, whether you view it as this model, the last one or the next one"); Thompson notes none of them define AGI, and states his own view that AGI requires continuous learning, which current LLMs don't do. Separately, Ed Zitron reports — citing unnamed sources with direct knowledge of the Stargate Abilene data center site being built by Crusoe for Oracle to lease to OpenAI — that at most four of the site's eight planned buildings are functional and that the site has capacity for a total of 400,000 Blackwell GPUs, of which Zitron estimates at best half are actually installed and functional. Simon Willison separately quotes OpenAI Chief Scientist Jakub Pachocki arguing continued frontier-model training is needed to build "aligned AI for defense" against "rogue agents," while cautioning that "racing forward at all costs" would be "absurd."
Significance 3: The AGI-arrival claims from Huang, Chen, and Brockman are corroborated across multiple outlets per Stratechery (TIME, WSJ), but the specific Stargate Abilene construction-status claim rests solely on Zitron's unnamed sourcing without independent confirmation, so significance is held at 3 pending corroboration.
Implications · machine-drafted, not owner judgment
The gap between public AGI-arrival rhetoric and Zitron's single-sourced reporting that Stargate Abilene is roughly half-built is exactly the kind of stated-vs-revealed gap the owner tracks in lab leadership; if the infrastructure claim holds up, it undercuts near-term AGI timelines tied to this compute buildout and adds weight to the reading that "AGI" is being used as an undefined marketing term ahead of anticipated public offerings for OpenAI and Anthropic. Pachocki's defensive-scaling rationale is a specific, checkable justification for continued frontier investment that can be weighed against actual safety-incident rates going forward.
- Independent verification (other outlets, satellite imagery) of Stargate Abilene's actual building/GPU count
- OpenAI or Nvidia offering a concrete, testable AGI definition or benchmark
- Whether OpenAI's stated defense/safety framing is matched by any observable defensive-AI deployment
Sources
- curator Write Things Down retrieved 2026-09-08
- curator Concentration Risk retrieved 2026-09-08
- curator Quoting Jakub Pachocki retrieved 2026-09-08