Facepalm: Nvidia’s extremely anticipated Blackwell sequence of AI chips has encountered a big setback. Newly found design flaws will delay shipments by at the least three months. This delay will doubtless trigger appreciable disappointment amongst prospects who’ve positioned billions of {dollars} in orders.
Nvidia’s extremely anticipated Blackwell sequence of AI chips is going through vital delays because of design flaws found late in manufacturing. The Data cites two nameless sources concerned with Nvidia’s chip and server {hardware} manufacturing, stating that the difficulty might take at the least three months to resolve.
The crux of the difficulty lies within the processor die connecting two Blackwell GPUs on a GB200 chip – an issue recognized by producer TSMC. In response, Nvidia is revising the design and might want to conduct new manufacturing checks with TSMC earlier than mass manufacturing begins. As a stopgap measure, the corporate is contemplating producing a single GPU model of the Blackwell chip to expedite supply.
The delay has far-reaching implications for Large Tech gamers who’ve invested closely in Nvidia’s expertise. As an illustration, Google has ordered over 400,000 GB200 chips in a deal exceeding $10 billion. Equally, Meta has positioned a $10 billion order, whereas Microsoft had plans to have 55,000 to 65,000 GB200 GPUs prepared for OpenAI by the primary quarter of 2025 – a timeline now in jeopardy.
Nvidia has reportedly knowledgeable Microsoft and one other cloud supplier in regards to the delay affecting essentially the most superior AI chip fashions within the Blackwell sequence. Consequently, vital shipments of those chips aren’t anticipated till the primary quarter of 2025, probably disrupting the AI methods of those tech giants.
Regardless of these experiences, Nvidia’s official stance stays optimistic. An organization spokesperson said that “manufacturing is on observe to ramp” later this yr with out immediately addressing the reported delay. In the meantime, the affected corporations, together with Microsoft, Google, Amazon Net Providers, and Meta, have declined to remark.
The setback might permit Nvidia’s rivals to achieve floor within the AI chip market. Intel and AMD have struggled to impression Nvidia’s market share for the reason that outset of the AI growth. Nevertheless, the delay may allow them to reposition their merchandise as viable alternate options for purchasers needing fast options.
As an illustration, AMD designed its open-source ROCm framework to compete immediately with Nvidia’s CUDA, providing builders a substitute for construct AI functions with out being locked into Nvidia’s ecosystem. Likewise, Intel is creating AI accelerator chips, together with the Gaudi line, as extra reasonably priced alternate options. In keeping with Intel, its AI accelerators are one-third to two-thirds the worth of competing manufacturers.
As prospects transfer previous the preliminary disappointment of the delay, they could query Nvidia’s capacity to keep up its dominant 80-percent market share within the face of manufacturing challenges and growing competitors. Because the AI arms race intensifies, the trade will carefully watch how Nvidia navigates this hurdle and whether or not it may well ship on its guarantees to its high-profile prospects.