Nvidia Corporation has encountered engineering challenges in the development of its new advanced chips, leading to delays. The setbacks primarily involve the company's highly anticipated Blackwell line-up and related products, which were initially set to strengthen Nvidia’s lead in AI computing.
According to sources familiar with the matter, the delays are affecting two key areas. First, a version of the Blackwell chip, designed as an AI accelerator, is undergoing rework to improve compatibility with existing data centre infrastructure. This infrastructure was originally tailored for Nvidia’s earlier Hopper H100 chip. Second, the company faces issues with a combined processor and graphics chip, which will not be available in large quantities as soon as expected due to problems with supporting technology.
These obstacles reflect the broader challenges Nvidia faces as it accelerates its innovation pace under CEO Jensen Huang. The company has been pushing out new chip designs rapidly to maintain its market dominance in AI computing, a sector where it has seen substantial growth in sales and market valuation over the past two years.
Nvidia declined to comment specifically on the reported engineering problems, labelling them as "rumours." However, the company reassured stakeholders that Blackwell production is progressing and is expected to ramp up in the latter half of the year. Nvidia has already started sending out samples of the Blackwell chip to customers, and demand for its Hopper generation remains robust.
Despite the company’s optimistic outlook, the delays have impacted Nvidia’s stock performance. Shares fell 6.4 per cent on Monday, exacerbated by a broader tech sector downturn. In contrast, rival Advanced Micro Devices saw a 1.8 per cent increase, as investors speculate that AMD might capitalise on Nvidia’s setbacks.
Nvidia’s chips are integral to major data centres being built by tech giants like Microsoft and Alphabet’s Google, which are investing heavily in AI services. While the delays might affect the supply flow of components, which are manufactured by Taiwan Semiconductor Manufacturing Company, analysts largely view the issues as temporary. TD Cowen analyst Matt Ramsay noted that while such bumps are part of the rapid innovation cycle, they are unlikely to significantly derail Nvidia’s revenue growth or long-term prospects.
In a May earnings call, Jensen Huang had projected that Blackwell chips would be fully in production and available to cloud-computing providers by later this year. Huang expressed confidence that demand for Blackwell would surpass supply, maintaining a positive outlook on revenue for the new line-up. Nvidia is scheduled to provide its next quarterly update on 28 August, where it will likely offer further details on the status of its chip development and market strategy.