Nvidia’s Infrastructure Push Goes Global: A New Campus in Taipei and a Blackwell Cloud Deal Take Shape

Nvidia announces $1.27B Taipei HQ and $1.6B Blackwell order, reports record $81.6B Q1 revenue with 85% YoY growth, and guides $91B Q2.

The chipmaker is pouring capital into both physical bricks and cloud capacity at a pace that underscores just how fast the AI buildout needs to move. Two announcements this week — a $1.27-billion headquarters in Taipei and a $1.6-billion order for Blackwell systems at a Texas data centre — show Nvidia is no longer content to just design chips; it is actively shaping the infrastructure that runs them.

Jensen Huang confirmed the Taipei project, dubbed "Nvidia Constellation," will break ground in June or July in the Beitou-Shilin Technology Park. The company secured the land under a 50-year lease in February 2026, with an option to extend. The campus is slated to open in 2030 and will house up to 10,000 employees focused on chip design, AI research and software development. The location puts Nvidia right next to TSMC, its most critical manufacturing partner, which alone accounts for roughly 22 percent of the foundry’s revenue. Huang also disclosed that Nvidia’s total annual spending in Taiwan has ballooned to around $150 billion — including chips, advanced packaging and services — up from just $10-15 billion five years ago and already past the $100-billion mark today.

On the cloud side, IREN has placed a $1.6-billion order with Dell for air-cooled Blackwell systems destined for its Childress, Texas facility. The hardware — GPUs, servers, storage, networking, integration and warranties — will power a five-year cloud-services contract IREN signed with Nvidia on May 7 valued at roughly $3.4 billion. That deal tasks IREN with providing managed GPU cloud services for Nvidia’s internal AI and research workloads, and it includes a strategic partnership that could eventually support up to 5 gigawatts of near-DSX AI infrastructure. Nvidia also secured the right to buy up to 30 million IREN shares at $70 apiece, a potential $2.1-billion equity stake. The order turns a paper commitment into a live deployment, with initial operations targeted for early 2027.

These infrastructure moves come on the heels of a record quarter. In the first quarter of fiscal 2027, Nvidia reported revenue of $81.6 billion, up 85 percent year over year. The data-centre segment contributed $75.2 billion, with compute accounting for $60.4 billion and networking adding $14.8 billion under the old segmentation. The board authorised an additional $80 billion in share buybacks and raised the quarterly dividend to $0.25 a share. The next-generation Vera Rubin AI platform is in full production, with partner shipments starting in the second half of the year; Nvidia promises a tenfold reduction in inferencing costs compared with Blackwell. Volume ramp for Rubin is set to begin in the fourth quarter of 2026.

The outlook remains robust. Nvidia guided for second-quarter fiscal 2027 revenue of around $91 billion, plus or minus 2 percent, with gross margins at 75 percent. But the stock has struggled to break higher, closing at €184.46 — roughly 8 percent below its 52-week high of €201.05. The relative strength index of 39.5 hints at mild bearish momentum despite the positive news flow. On a weekly basis the shares are down 4 percent, though the year-to-date gain of 14.5 percent keeps the long-term trend intact.

Geopolitical headwinds are not easing. Nvidia continues to expect zero data-centre compute revenue from China in the current quarter, and Bloomberg reported this week that US authorities are investigating alleged chip smuggling into China via Japanese intermediaries. Nvidia has urged partners to tighten compliance, a reminder that even the strongest supply chain is vulnerable. Execution is now the single measure that matters: can Nvidia deliver Vera Rubin on time, convert pipeline like IREN’s into capacity, and keep its Taiwanese — and American — supply lines running smoothly?