ADVERTISEMENT

On Tuesday, Nvidia announced its new generation of artificial intelligence (AI) chips, designed to power future advancements in robotics and self-driving cars at the annual (GTC) in San Jose, California.

However, its shares fell more than 3% following the announcement as investors continued offloading technology shares amid economic uncertainties linked to Trump’s tariffs. The other stocks in the Magnificent Seven group also ended the session lower.

The AI powerhouse’s stocks declined 15% this year, weighed down by global risk-off sentiment and the launch of Chinese DeepSeek’s cheaper AI model.

Additionally, the company’s latest quarterly earnings report failed to impress investors as sales growth showed signs of slowing. However, Josh Gilbert, a market analyst at eToro Australia, said: “Investors may see that as an opportunity, particularly with its valuation remaining attractive on the backdrop of ongoing growth.”

The GTC, a key gathering of global AI developers and investors, is a pivotal event for Nvidia. During the week-long conference, the chipmaker must persuade hyperscalers to sustain their heavy investment in its next-generation chips.

This is particularly important following DeepSeek’s launch of a cost-effective generative AI model, which poses a competitive challenge to leading US tech firms.

New AI chips to boost market dominance

At the event, a key update of Nvidia’s chip line is the successor of Nvidia’s Blackwell supercomputing chip, called Blackwell Ultra, which is set to begin shipments in the second half of 2025. Blackwell series chips, Nvidia’s flagship products powering its future growth, commenced shipments in the last quarter of 2024.

The upgraded version can handle more tasks during the same amount of time as its predecessor, allowing cloud providers to generate 50 times the revenue as the last generation Hopper GPUs. “We designed Blackwell Ultra for this moment — it’s a single versatile platform that can easily and efficiently do pretraining, post-training, and reasoning AI inference,” said CEO Jensen Huang.

Another major development was the announcement of Vera Rubin, a next-generation system combining CPU and GPU capabilities, expected to launch in the second half of 2026. Named after the astronomer who helped discover dark matter, Vera Rubin is designed as a custom-built supercomputing system that can manage 50 petaflops while doing inference, more than double the current Blackwell chips.

Nvidia also announced collaborations with Walt Disney and Google DeepMind on a project called Isaac GR00T N1, aimed at accelerating robotics development. It has also partnered with General Motors for next-generation automotive AI and will work with T-Mobile US and Cisco Systems to develop AI-powered 6G network hardware.

AI investments are attracting hundreds of billions in the next decade

Nvidia’s major customers are top cloud providers including Amazon, Microsoft, Alphabet, and Oracle Corp., which have collectively purchased 3.6 million Blackwell AI chips so far in 2025, Huang said. Bloomberg reported that Hyperscalers, including Microsoft, Amazon, and Meta Platforms are expected to spend $371bn (€340bn) on AI infrastructure in 2025, with investment expected to rise to $525bn (€480bn)by 2032.

Future AI expenditure is projected to focus more on processing power and inference capabilities, inspired by DeepSeek’s innovations, rather than the development of entirely new AI models.

Huang said: “In the last 2 to 3 years, a major breakthrough happened, a fundamental advance in artificial intelligence happened. We call it agentic AI,” and “It can reason about how to answer or how to solve a problem.”