[News] NVIDIA Dismisses AI Bubble Concerns, Reportedly Projects $500B in GPU Sales from Blackwell and Rubin

At NVIDIA’s first-ever Washington GTC from October 27-29, CEO Jensen Huang pushed back against AI bubble concerns while unveiling the next-generation Vera Rubin superchip. According to CNBC, he projected combined GPU revenue of $500 billion from the Blackwell lineup and next year’s Rubin release, noting that demand remains strong with 6 million Blackwell units shipped over the past four quarters.

Meanwhile, gurufocus notes that Huang expects 20 million units of the new generation to ship, up from 4 million of the prior Hopper series. He reportedly emphasized that current AI models are driving substantial infrastructure investments, helping to calm market fears of an AI bubble.

As CNBC highlights, AI hyperscalers are ramping up plans to build massive supercomputing data centers, largely powered by NVIDIA AI processors, to handle soaring workloads. Citing analysts, the report suggests total hyperscaler capital expenditures are expected to climb 24% next year, reaching nearly $550 billion.

Huang: No NVIDIA Export Licenses Filed Amid China Curbs

At GTC, Huang also raised concerns about the Chinese market. According to Reuters, he said China has effectively blocked NVIDIA, and the company has not sought U.S. export licenses for its latest processors because of Beijing’s stance. “They’ve made it very clear they don’t want NVIDIA there right now,” Huang told reporters, adding that he hopes that changes in the future, as China remains a very important market.

Back in September, China accused NVIDIA of breaking its anti-monopoly law after the Trump administration arranged a revenue-sharing deal allowing the U.S. AI chip maker to export H20 chips. Since then, Huang reportedly said NVIDIA’s share of the Chinese market has dropped from 95% all the way to zero.

At GTC, Jensen Huang highlighted that NVIDIA’s fastest AI chips, the Blackwell GPUs, are now rolling off production lines in Arizona, reinforcing the company’s commitment to U.S. manufacturing. Reuters notes that the chips are produced at TSMC facilities in Arizona, with servers assembled in Texas and networking equipment built in California.

Vera Rubin Unveiled

As per TechNews, at the GTC October 2025 event held in Washington, Jensen Huang unveiled the company’s next-generation Vera Rubin superchip for the first time. He revealed that the motherboard will integrate one Vera CPU with two Rubin GPUs, support up to 32 LPDDR memory channels, and feature HBM4 high-bandwidth memory on the GPUs.

Huang noted that the Rubin GPUs have returned to the lab for testing as the first samples manufactured by TSMC, according to TechNews. Each GPU reportedly includes eight HBM4 interfaces and two reticle-sized GPU dies. The Vera CPU features 88 custom Arm-based cores, supporting up to 176 threads.

According to NVIDIA’s plan, Rubin GPUs are expected to enter mass production in the third or fourth quarter of 2026, roughly in line with—or potentially earlier than—the full-scale production of the current Blackwell Ultra GB300 Superchip platform, the report notes.

Read more

• [News] NVIDIA May Be TSMC’s Only A16 Customer So Far While Apple Reportedly Yet to Start Talks

(Photo credit: NVIDIA)