• Nvidia's new AI chip launch is delayed, raising doubts about its annual release promise.
  • The delay is due to manufacturing challenges with Taiwan Semiconductor Manufacturing Company.
  • Analysts remain optimistic, citing strong demand from major customers like Meta and Google.

News that Nvidia's next chip launch will be delayed is calling into question whether the company can deliver on its promise to launch a new chip every year.

The rollout of Nvidia's next-generation of AI chips, Blackwell, is going to be delayed according to media and industry analyst reports. Though a few Blackwell chips are already in customers' hands, according to the company, the ramp-up to full-scale production was supposed to come this fall. It will now be delayed two to three months.

The chip experts at research firm Semianalysis wrote Sunday that Nvidia is facing "major issues in reaching high production volume," adding that, in some cases, Hopper chips would likely be sent to customers instead of the delayed Blackwell chips.

The Information reported Friday that some Blackwell shipments would now be pushed to the first quarter of next year when CEO Jensen Huang had promised revenue from Blackwell by later this year.

Nvidia's largest customers are Meta, Microsoft, and Google. Smaller cloud firms that have built their businesses on Nvidia chips could be affected by the delay too.

Why a new product every year?

Huang announced on the company's first-quarter earnings call in May that it would move to a one-year product cadence.

"I can announce that after Blackwell, there's another chip and we are on a one-year rhythm," Huang said May 22. Each new generation of AI chips promises step changes in computing capacity, speed and efficiency, allowing for faster computations and larger models.

For cloud providers, Nvidia's customers, the new chips also provide an opportunity to tier-out their services further, creating a higher ceiling on the most elite AI computing capabilities and building a base of older chips that could make the more entry-level computing capacity more affordable.

Analysts have also noted that the annual cadence of innovation makes Nvidia much harder to catch for competitors like AMD and Cerebras.

The delay stems from Nvidia's Blackwell design, according to Semianalysis. Taiwan Semiconductor Manufacturing Company, Nvidia's manufacturing partner, is using a relatively new process to manufacture the Blackwell chips. The more complex design is intended to yield more memory capacity, but ramping the extremely precise manufacturing process for this next-generation chip at the scale and pace Nvidia originally intended has been challenging. Semianalysis reported that several parts need to be redesigned.

Bernstein analysts wrote in a note to investors Monday, "While potential delays are never great, we aren't panicking just yet." Demand for Nvidia's products continues to rise and signals from some of Nvidia's biggest customers reporting earnings last week were encouraging, according to the analysts.

Meta, Microsoft, and Amazon all said their spending on AI computing infrastructure was nowhere near winding down in earnings calls last week.

"NVIDIA's competitive window is so large right now that we don't think a three-month delay will cause significant share shifts," Bernstein analysts wrote.

What happens now?

Blackwell is still rolling out this fall, but not at the volume the company originally projected.

Nvidia declined to comment on how this news impacts the plan to launch new products every year. The company told Reuters over the weekend that demand for Hopper, the latest generation of widely available AI chips on offer, is strong. Some Blackwell chips have begun shipping to customers and production will still increase in the second half of the year.

In a research note Sunday, Bank of America projected a limited impact on Nvidia's revenue in the third quarter, suggesting that the firm can ship more Hopper chips and more simplistic versions of the Blackwell chip in light of the manufacturing challenge.

Given the optimism of big tech firms, who are all-in on AI, Bernstein analysts concluded, "We find it difficult to believe that any roadmap timing changes, if occurring, are not already contemplated in current outlooks from these customers."

Got a tip or an insight to share? Contact Senior Reporter Emma Cosgrove at [email protected] or use the secure messaging app Signal: 443-333-9088

Read the original article on Business Insider