Wistron's D1 plant in Dallas, Texas, held its opening ceremony on July 22, 2026, according to a TechNews report. The facility represents Wistron's first manufacturing site in the United States, built at a cost of nearly US$700 million (approximately NT$22.6 billion), with a floor area of about 324,000 square feet, TechNews reported.
According to TechNews, the D1 plant is also where the first NVIDIA GB300 compute substrate produced in the United States was born and successfully mass-produced. This distinction marks Dallas D1 as the origin point for domestic GB300 substrate output, per the report.
The opening ceremony drew a lineup of officials and executives, TechNews reported: Taiwan's Representative to the U.S., Alexander Yui (俞大㵢), NVIDIA CEO Jensen Huang and his daughter Michelle Huang, NVIDIA Executive Vice President of Operations Debora Shoquist, and Fort Worth Economic Development Director Jessica Rogers all attended in person.
At the event, Wistron Chairman Simon Lin said, "NVIDIA is driving an entirely new AI era," adding that the pace of technological change is extremely fast and "if you can't keep up, you'll be left behind," according to TechNews.
According to TechNews, citing Wistron's Lin Jian-Xun (林建勳), Wistron purchased the D1 building in April last year and began moving in and preparing the site in May. Mass production started in July this year, timed to align with customer schedules, the report said.
TechNews reported that the D1 plant is currently responsible for AI server-related system assembly and testing at the L6 level, and that its first production line is already running smoothly. Lin Jian-Xun also said the Dallas plant's automation level currently stands at about 60%, which he said has improved production efficiency and capacity stability, per the report.
Wistron has already planned a D2 plant, with a future scale expected to be at least twice the size of D1, TechNews reported.
| Metric | Value | Evidence |
|---|---|---|
| Investment | ~US$700 million (NT$22.6 billion) | E2 |
| Floor area | ~324,000 sq ft | E3 |
| Automation level | ~60% | E10 |
| Planned D2 scale | At least 2x D1's size | E7 |
| Date | Milestone | Evidence |
|---|---|---|
| April (last year) | Wistron purchased the D1 building | E8 |
| May (last year) | Move-in and site preparation began | E8 |
| July (this year) | Mass production began | E8 |
| July 22, 2026 | Official opening ceremony | E1 |
Taken together, the figures TechNews reported outline a compressed timeline: Wistron moved from purchasing the D1 building in April last year to mass-producing NVIDIA's GB300 substrate within roughly 15 months, reaching a stated 60% automation level along the way. The plant's current scope — L6-level AI server assembly and testing on one operating production line — sits alongside Wistron's disclosed plan for a D2 facility at least twice D1's size, indicating the July 22 opening was framed by Wistron and the attending NVIDIA executives as an early step rather than a finished buildout.
According to TechNews, D1 currently handles AI server-related system assembly and testing at the L6 level, and it is also where the first NVIDIA GB300 compute substrate made in the United States was produced. Its first production line is already running.
Per TechNews, citing Wistron's Lin Jian-Xun, Wistron purchased the D1 building in April last year, began moving in and preparing the facility in May, and started mass production in July this year to align with customer schedules.
TechNews reported that Wistron has already planned a D2 plant, with a future scale expected to be at least twice the size of the D1 facility.
OpenAI admits its pre-release models breached Hugging Face in a cyber test, per TechCrunch; Hugging Face's AI agents stopped it, The Verge adds.
According to Central News Agency (CNA) and UDN Money reports, TSMC (台積電) announced on July 22, 2026 the issuance of NT$18.5 billion in unsecured ordinary corporate bonds, designated as green bonds, with proceeds earmarked for green building and environmental spending. The offering is split into a 5-year Class A tranche and a 10-year Class B tranche, underwritten by Taishin Securities.
Google announced three new Gemini models on July 21, 2026 — 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber — with 3.6 Flash cutting token use by up to 65% on long-horizon engineering benchmarks like DeepSWE, according to ithome.com.tw and VentureBeat. Gemini 3.5 Pro is now testing with partners, per Google's Logan Kilpatrick.