Unlock Editor's Digest at no cost
FT editor Roula Khalaf selects her favourite stories on this weekly newsletter.
Delays to Nvidia's next-generation AI processor won’t derail the chipmaker's plans to provide a new edition of its flagship product yearly, CEO Jensen Huang told the Financial Times.
A production issue at Blackwell, Nvidia's highly anticipated latest chip platform, “doesn't matter” to the plan to speed up the discharge schedule from two years to a “one-year cadence,” Huang said after announcing the ambitious latest schedule last October.
Nvidia's accelerated pace of innovation is seen as key to maintaining its dominance in AI chips in a market estimated to be value tons of of billions of dollars in the approaching years.
Earlier this month, last-minute complications forced Nvidia and its partner Taiwan Semiconductor Manufacturing Company to make changes to Blackwell's production, forcing Nvidia to delay the chip's full release until later this 12 months.
The delays contributed to a drop in Nvidia's share price on Thursday as investors apprehensive about slower growth and tighter profit margins after the chipmaker rode the wave of the AI boom to a valuation of over $3 trillion.
The US technology giant's ability to fulfill ambitious production and profit targets has assumed disproportionate importance within the US stock market. A rapid rally in the corporate's shares is answerable for 1 / 4 of the S&P 500's annual gains up to now.
Despite the increasing complexity of producing the advanced chips, Blackwell's successors are still on schedule for annual upgrades, Huang said. The company's engineers are working on the subsequent two generations, he said. “They will likely be ready on their annual cycle.”
Nvidia is accelerating product improvements to keep up its lead over competitors similar to AMD and Intel within the graphics processing unit market.
Companies like Meta and OpenAI depend on Nvidia's GPUs to coach their large language models, the AI systems behind chatbots like ChatGPT. For nearly two years, demand for the corporate's strongest chips has far exceeded supply.
Huang was bombarded with questions from analysts on Wednesday about how Nvidia's customers are earning profits from their huge investments in AI infrastructure, as investors debate how long the massive tech firms can proceed to spend tens of billions of dollars each quarter to equip their data centers.
Huang told the FT that Nvidia itself is achieving significant productivity gains by utilizing AI to hurry up data processing and assist within the design and construct of its products.
“We can't develop (our chips) without generative AI,” he said. “We couldn't develop Blackwell, we are able to't write our software without AI anymore.”
He said that because of the AI push, Nvidia's 20,000 engineers are “actually about 60,000.”
“I suppose Nvidia is the smallest large company on the earth,” he said. “We can do things that should not reasonably possible with hand-written software… Nvidia isn’t a big contracting company.”
Huang told analysts on Wednesday that Blackwell's production problems are primarily related to the “mask” – the template that guides the design of a chip through a Process called photolithographyThe adjustments were needed to enhance Blackwell's manufacturing yield, which is critical to Nvidia's profit margins.
“The mask modification is complete,” Huang said. “No functional changes were needed.” Blackwell will ramp up mass production and start shipping to customers in Nvidia's fourth fiscal quarter, which ends in January 2025, he added.
When asked in regards to the impact on Nvidia's profit margins, which the corporate told investors will decline barely in the approaching months, CFO Colette Kress said the corporate “might even see slight differences within the fourth quarter, again as a consequence of our transitions and the various cost structures as we launch our latest products.”
Some Wall Street analysts have already begun to revise their estimates for Nvidia's gross margins downward, contributing to the decline within the share price on Thursday.
TD Cowen analyst Matthew Ramsay said in a note to clients Thursday that the challenges facing Nvidia and its competitors in attempting to emulate the iPhone's annual release cycle are enormous.
“The entire server industry is bringing to market some of the complex products in computing history,” Ramsay said. “This will inevitably create tensions in design and provide chains, and we proceed to imagine that a slight shift in production timing doesn’t change the thesis or reflect a change in overall demand for AI computing solutions.”
