workflow run · 20261025 · /opt/data/workflows/runs/20261025-065046-595b
<!-- 00-gather.md -->
collected 2026-10-25T06:50:49Z · ddgs + HN + GitHub + arXiv + radar.db
<!-- 01-claims.md -->
| # | statement | would be FALSE if |
|---|---|---|
| 1 | Jensen Huang said a 1 GW AI factory costs about $50B | no primary/verbatim source carries the number |
| 2 | The $50B figure was attached to DSX POD as its price | the number is an industry benchmark for any 1GW factory |
| 3 | NVIDIA announced DSX POD at GTC Berlin (21 Oct 2026) | DSX POD predates the keynote or was announced elsewhere |
| 4 | DSX POD deploys in 14 days vs ~6 months | NVIDIA newsroom does not say 14 days / says different baseline |
| 5 | DSX POD has a list price | NVIDIA announced a price |
<!-- 02-verify.md -->
| # | statement | verdict | source | note |
|---|---|---|---|---|
| 1 | 1 GW AI factory ≈ $50B (Huang, GTC Berlin) | confirmed | https://uk.investing.com/news/stock-market-news/nvidia-at-gtc-berlin-2026-keynote-europes-ai-buildout-93CH-4922188 | keynote transcript: "a 1 gigawatt AI factory can cost about $50 billion"; a 3-month delay ≈ $10B loss |
| 2 | $50B is DSX POD's price | contradicted | https://aimagazine.blog/nvidia-dsx-pod-vera-rubin-racks-14-day-setup-price/ | NVIDIA announced no DSX POD price; analysts put a Vera Rubin rack >$7M, 7 racks >$49M before integration |
| 3 | DSX POD announced at GTC Berlin | confirmed | https://nvidianews.nvidia.com/news/nvidia-introduces-dsx-pod-ai-factories-to-accelerate-europe | NVIDIA newsroom release tied to GTC Berlin |
| 4 | 14 days vs 6 months | confirmed | https://qz.com/nvidia-dsx-pod-containerized-ai-data-centers-102126 | container ~28,000 kg, truck-shippable, deployment cut to 14 days |
| 5 | DSX POD list price exists | contradicted | https://www.futuriom.com/articles/news/nvidia-bets-costly-dsx-pod-against-ai-factory-time | no price announced; only analyst estimates per POD |
<!-- 03-verdict.md -->
The number is real and it is Huang's: he said a 1 GW AI factory costs about $50B, and that a
3-month delay costs roughly $10B in lost revenue. But the sentence attaches that benchmark to
DSX POD as if it were the product's price — NVIDIA announced no DSX POD price at all.
Before: "DSX POD — ประกาศกรอบราคา ~$50B ต่อ 1GW"
After: "DSX POD — Huang ใช้ตัวเลข ~$50B ต่อ 1GW เป็นต้นทุนของ AI factory ขนาดนั้น (ไม่ใช่ราคา DSX POD;
DSX POD ยังไม่ประกาศราคา — นักวิเคราะห์ประเมิน >$49M ต่อ POD จาก 7 racks)"
An NVIDIA press release or on-slide price for DSX POD, or a keynote quote that names DSX POD and
$50B in the same sentence.