laitimes

Nvidia's latest AI chip faces delayed delivery, and Huang Jenxun cashed out more than $400 million in two months

Nvidia's latest AI chip faces delayed delivery, and Huang Jenxun cashed out more than $400 million in two months

Lei Jianping on August 3

Nvidia (NVIDIA) has revealed that the upcoming AI chip will be delayed by three months or more due to design flaws, and the mass production of Blackwell may be delayed until the first quarter of next year.

This could affect customers such as Meta Platforms, Google, and Microsoft, who together have ordered billions of dollars worth of chips.

Morgan Stanley said that the production of Blackwell chips may be suspended for about two weeks, but it can be caught up by TSMC's efforts in the fourth quarter of 2024.

NVIDIA declined to comment on the delayed statement, but said customers are testing samples of Blackwell chips and expect production to ramp up later this year.

Engineers discover chip defects in preparation for mass production

It's not uncommon to find major design flaws before mass production. This is because multiple production tests, runs, and simulations are required at an early stage to ensure the viability of the product and the smooth running of the manufacturing process.

Blackwell's design problems have emerged in recent weeks, with TSMC engineers discovering flaws in preparation for mass production, according to the source.

The GB200 chip contains two connected Blackwell GPUs and a Grace CPU. The defect involved the connection of two Blackwell GPUs' processor chips (the silicon wafers used to house the chip's circuitry), a hurdle that reduced TSMC's chip production capacity for NVIDIA and could even lead to the company's shutdown.

Nvidia is in the midst of a new trial production with its chipmaker TSMC, and to avoid bottlenecks in the machine, TSMC has restarted production of another high-end product that is close to mass production to solve the problem. This is also rare.

According to the original plan, TSMC will start mass production of Blackwell chips in the third quarter of 2024 and deliver them to NVIDIA from the fourth quarter. Nvidia founder Jensen Huang said in May that Nvidia plans to ship Blackwell in large quantities later this year.

Nvidia's customers are worried

Blackwell can be described as the "white moonlight" in the hearts of technology companies, carrying the high hopes of giants.

If the upcoming release of AI chips such as the B100, B200, and GB200 is delayed for three months or more, Nvidia's customers will really have to worry.

These customers include Microsoft, Meta, OpenAI and others, who have high hopes for NVIDIA's AI chips and plan to use NVIDIA's "supercomputer" to produce future generations of large language models, Meta AI assistants and other automation capabilities.

Meta has placed orders of at least $10 billion, and Microsoft has increased its order size by 20% in recent weeks. Microsoft plans to prepare 55,000-65,000 GB200 chips for OpenAI by Q1 2025.

And with the quarterly report, more than 40% of Nvidia's revenue comes from some of the familiar names in the "Big Seven" stocks — Microsoft, Meta, Alphabet and Amazon.

Nvidia's largest customer comes from Microsoft, which spends 19% of Nvidia's revenue. is Nvidia's largest customer to date, almost twice as much as Meta spends, and three times as much as Alphabet and Amazon.

Huang cashed out more than $400 million in less than February

Since 2024, NVIDIA's stock price has continued to rise, once exceeding $3 trillion in market value, even surpassing Microsoft and Apple to become the world's most valuable company.

In the process of Nvidia's market value continuing to rise, Nvidia founder and CEO Jensen Huang has recently continued to sell the company's equity.

According to the documents, Huang was found on June 13, 14, 17, 18, 20, 21, 24, 25, 26, 27, 28, 2024, July 1, 2, 3, 5, 8, 9, 10, 11, 12, 15, 16, 17, 18, 19, 22, 23, 24, 25, 26, 29, 30, 31, July 1, 2024, On the 2nd, the holdings were reduced by 120,000 shares each time.

Nvidia's latest AI chip faces delayed delivery, and Huang Jenxun cashed out more than $400 million in two months

Huang has reduced his holdings for 35 consecutive trading days, reducing his holdings by 4.2 million shares in less than two months.

As of Friday's close, Nvidia shares were $107.27 and had a market capitalization of $2.6387 trillion.

Nvidia's latest AI chip faces delayed delivery, and Huang Jenxun cashed out more than $400 million in two months

Nvidia's stock price is currently at a relative low level in several months, and the $12.63 million cashed out on August 2 means that Huang has cashed out at least $440 million in less than two months.

Huang isn't the only insider to sell stocks.

In the first half of 2024, company executives and directors sold more than $700 million worth of shares, more than at any other time in the company's history.

Some analysts believe that the sell-off by Huang and other executives shows a lack of confidence in the company's recent stock performance and could signal a continued decline in the company's stock price. There are also analysts who believe that the sale of shares by insiders should not be seen as the only indicator for making investment or trading decisions.

In the AI jungle, the data reminds investors that chipmakers eat first, and hyperscalers are far from being able to give investors more hints about when AI investments will translate into AI revenue streams.

For example, in a recent earnings call, Microsoft said it plans to continue to expand infrastructure investment, build and lease data centers, and expects capital expenditures in fiscal 2025 to exceed the current fiscal year. This expenditure is necessary to support the demand for AI services, and land and data centers are long-term assets that may take 15 years or more to pay off.

Investors may be disappointed by the extended payback time. A growing number of investors are concerned that AI technology is being "overhyped".

One fund manager said, "Wall Street doesn't have a lot of patience." They see that you're spending billions of dollars and want to see an increase in revenue as well. If the performance of these companies does not meet expectations or does not far exceed expectations, they will be hit. ”

Other investors say that many of the supposed uses of AI "will never be cost-effective, will never really work, will only consume too much energy, or will prove to be untrustworthy."

This has also exacerbated volatility in Nvidia stock, and more and more investors are now turning to value stocks.

———————————————

Lei Di was founded by Lei Jianping, a media person, if it is reprinted, please indicate the source.

Read on