The Blackwell Series Of Chips Exposed To Waterloo Before Mass Production, And Shipments Delayed For 3 Months

The Blackwell Series of Chips Exposed to "Waterloo" before Mass Production, and Shipments Delayed for 3 Months

0021-12887 8"CLAMP RING

0020-21361 CLAMPING RING 6" AL JAPAN MAJOR

Customers who spend billions of dollars on orders are certainly not satisfied. According to the United States technology website The Information, due to design flaws, the shipment of Nvidia's next-generation Blackwell architecture series AI flagship chip GB200 will be delayed by at least three months. An unidentified Microsoft employee said Nvidia had notified Microsoft this week that shipments of its most advanced Blackwell series AI chip models would be delayed.

 

Design flaws are found before mass production

Nvidia released the Blackwell series in March, and CEO Jensen Huang confidently said in May that the company plans to start mass shipping of the Blackwell series chips later this year. The GB200 chip contains two connected Blackwell GPUs and a Grace central processing unit. However, in recent weeks, when TSMC engineers were preparing for mass production, they found design flaws in the bare die that connected the two Blackwell GPUs. This defect can lead to reduced chip yield or yield, and is typically done by stopping mass production. As a result, Nvidia had to make adjustments to the chip design and work with TSMC for a new trial production before starting mass production. Delays in chip shipments are not unheard of, but it is still very rare to find a major design flaw just before mass production. TSMC originally planned to start mass production of the Blackwell series of chips in the third quarter and start bulk shipments to Nvidia customers in the fourth quarter. However, due to the discovery of design flaws, mass production had to be postponed to the fourth quarter, and mass shipments are expected to be postponed to the first quarter of next year. TSMC reserved production capacity for the mass production of the GB200, but had to leave the production line idle until the problem was resolved. The delay has had a profound impact on big tech companies that have invested heavily in NVIDIA technology. Google, for example, has ordered more than 400,000 GB200 chips in transactions worth more than $10 billion. Similarly, Meta has placed a $10 billion order, while Microsoft originally planned to have between 55,000 and 65,000 GB200 GPUs ready for OpenAI in the first quarter of 2025 – but that timeline is now in jeopardy. Nvidia has reportedly notified Microsoft and another cloud provider that delays in the most advanced AI chip models in the Blackwell family will be impacted. As a result, these chips are not expected to ship in large quantities until the first quarter of 2025, which could disrupt the tech giants' AI strategies. At present, the market demand for B200 is very strong, and customers have switched from B100 to B200 (more than 450,000 units have been demanded). The Meta model trainer also said that they also need to add orders for B200, at least 150,000 B200s, which were expected to be placed next month. In addition, Nvidia will launch the H200 in the third quarter, and the quarterly performance is expected to be more than $33 billion. From October to November, some H200 will be added to the customer for emergency, and all of them will be normal by December at the latest.

Nvidia reaches the eye of the storm

Nvidia is once again involved in the center of the vortex of the AI wars. GPU leader Nvidia has been on a rollercoaster ride this week: it fell 7% on Tuesday, rose 13% on Wednesday, and fell nearly 7% on Thursday, with its stock price even more volatile than Bitcoin. According to the data, Nvidia's 30-day implied volatility soared to 71%, while Bitcoin's DVOL index, a measure of 30-day implied volatility, fell to 49%. Behind the sharp swings in Nvidia's stock price is the cost shadow looming over the tech giants, and the market is increasingly worried about the return on huge AI investments. Wall Street investment banks and hedge funds, represented by the bears, believe that technology giants have invested heavily in AI, but they cannot generate sufficient returns, and the current application scenarios of AI are limited, so such a huge investment is very unwise and dangerous. And many parties, represented by technology giants such as Mag 7, believe that it is necessary to increase capital expenditure in the field of AI now, otherwise they will miss the upcoming AI era.

Wall Street, where money is paramount, pays more attention to the input/output ratio.

Wall Street, which is gradually losing patience, believes that the capital expenditure of the tech giants in the field of AI is so high, but it does not bring a corresponding return and more efficient applications. In the past two years, only two phenomenal AI products, ChatGPT and Github Copilot, have emerged. A report by Goldman Sachs back in late June noted that AI technology may not be developing as quickly as expected, and its cost-benefit ratio may not be as attractive as it seems. AI is predicted to increase United States productivity by only 0.5% over the next decade, and its cumulative contribution to GDP growth is only 0.9%. Barclays also pointed out in an August report that tech giants are crazy about throwing money at AI out of "FOMO" and are afraid of missing out on AI development opportunities. Although AI technology is still in its infancy, the capital expenditure of large players on AI has shown an irrational boom, and FOMO sentiment is dominant, and as this sentiment subsides, major manufacturers will gradually cut back on AI investment next year. Analysts expect capital spending in the AI sector to reach a cumulative $167 billion from 2023 to 2026, and this figure is based on optimistic expectations for demand for AI products. However, in stark contrast, the revenue increment of AI cloud services is expected to be only $20 billion by 2026. Elliott Management, a top hedge fund on Wall Street, is reportedly even more aggressive, pointing out that big tech giants, especially Nvidia, are in a bubble and the artificial intelligence technology that drives their stock prices skyrocketing is overhyped.