As AI giant NVIDIA is said to delay its upcoming Blackwell series chips for months, which are now expected to hit the market around early 2025, the related semiconductor supply chain is experiencing a reshuffle. According to a report by the Korea Economic Daily, Samsung Electronics, which is eager to expand its market share for HBM3 and HBM3e, is likely to emerge as a major beneficiary in addition to AMD.

In March, NVIDIA introduced the Blackwell series, claiming it could enable customers to build and operate real-time generative AI on trillion-parameter large language models at up to 25 times less cost and energy consumption compared to its predecessor.

However, according to The Information, last week, NVIDIA informed major customers, including Google and Microsoft, that the shipments of its Blackwell AI accelerator would be delayed by at least three months due to design flaws.

Blackwell Delayed Potentially due to Design Flaws and TSMC’s Capacity Constraints

Tech media Overclocking points out that the defect is related to the part connecting the two GPUs, and creates problems for NVIDIA’s dual GPU versions, including the B200 and the GB200.

The delay has prompted tech companies to look for alternatives from NVIDIA’s competitors, such as AMD, according to the Korea Economic Daily. Microsoft and Google have already been working on next-generation products with AMD. For instance, Microsoft has purchased the MI300X, an AI accelerator from the US fabless semiconductor designer, the report says.

Samsung to Benefit thanks to the Collaboration with AMD

Samsung, as its HBM3 received AMD MI300 series certification in 1Q24, and is likely to provide HBM3e chips to AMD afterwards, is expected to benefit. Citing a semiconductor industry source, the Korea Economic Daily notes that as it is very risky for a single company to dominate the AI chip supply chain, the situation will create opportunities for Samsung and AMD.

It is also worth noting that Samsung’s HBM3 has passed NVIDIA’s qualification earlier, and would be used in the AI giant’s H20, which has been developed for the Chinese market in compliance with U.S. export controls.

According to TrendForce’s forecast in mid-July, the shipment share of AI servers equipped with self-developed ASIC chips in 2024 is expected to exceed 25%, while NVIDIA owning the lion’s share of 63.6%. AMD’s market share, on the other hand, is projected to reach 8.1% in 2024.

Read more

• [News] Foxconn Remains Unfazed Despite NVIDIA’s GB200 Delay, with Customers Aim to Secure Their Orders
• [News] Samsung’s 8-layer HBM3e to Start Mass Production in Q3, Driving HBM Sales to Soar 3-5 Times in 2H24

(Photo credit: NVIDIA)

According to a report from Commercial Times, despite ongoing turbulence in the semiconductor industry, including Intel’s capital expenditure cuts and reported bottlenecks in NVIDIA’s B-series GPU, TSMC’s leading position in the industry may remain unshaken.

The sources cited in the report note that the issues with the B-series GPU, stemming from mask replacements to enhance chip stability, have been quickly resolved by the foundry.

The sources cited in the report believe that NVIDIA’s Blackwell started production at the end of the second quarter. To improve stability, NVIDIA replaced some masks, causing about a two-week production delay. The redesign has been completed, and large-scale production will proceed in the fourth quarter.

The same source do not believe it will affect TSMC’s CoWoS revenue, as the idle two-week capacity will be filled by the equally strong demand for H100.

On the other hand, Intel’s CPUs are reportedly facing issues as well. As per the company’s statement, the 13th and 14th generation Intel Core desktop systems are experiencing instability due to a microcode algorithm resulting in incorrect voltage requests to the processor.

Although the company has provided a two-year warranty extension and real-time updates to fix the errors, concerns about design flaws and manufacturing process issues still exist.

In 2024, Intel’s new platforms, Arrow Lake and Lunar Lake, will have their CPU tiles produced using TSMC’s 3nm process, accelerating the production schedule. Lunar Lake and Arrow Lake are expected to ship officially by the end of the third and fourth quarters of this year, respectively.

With the support of the 3nm technology, these measures are expected to alleviate market concerns.

The sources cited by Commercial Times estimate that TSMC’s competitor Intel has begun to strictly cut costs, reducing capital expenditures by 20%. This could affect key capabilities in mass production and defect resolution in wafer manufacturing.

Therefore, sources cited by the report believe that TSMC’s leading position remains difficult to challenge in the short term.

Read more

• [News] TSMC Pushes for FOPLP Mass Production by 2027, Reportedly Eyeing on Innolux’s Plant
• [News] TSMC Reportedly Forms a Team on FOPLP Development, with Mini Line on the Road

(Photo credit: TSMC)

According to a report from The Information, NVIDIA’s “world’s most powerful AI chip,” the GB200, is said to be experiencing yield issues, leading to a one-quarter delay in mass shipments.

As per sources cited by a report from the Economic Daily News, it’s suggested that the problem likely lies in the yield rates of advanced packaging, mainly affecting the non-reference-designed GB200 chips.

The supply of the reference-designed GB200 chips remains stable, with Foxconn being the sole contract manufacturer receiving an adequate supply of these chips. Foxconn is set to ship according to the original schedule in the fourth quarter.

Furthermore, the sources cited by the same report point out that Foxconn is currently the only manufacturer able to meet the scheduled shipment of the GB200 in the fourth quarter. This is primarily due to Foxconn securing NVIDIA’s reference-designed GB200 chips orders, which are prioritized for shipment amid the supply shortage.

The term “reference-designed” refers to the GB200 AI servers ordered by NVIDIA for production at Foxconn and other manufacturers. These products are made according to NVIDIA’s reference designs and are not customized. Once produced, they can be sold to cloud service providers (CSPs) and other clients.

In contrast, “non-reference-designed” refers to customized versions of the GB200, which are tailored to specific customer requirements. The current yield issues are affecting the production of these non-reference-designed items, with the priority given to shipping the reference-designed products first.

Following the reports addressing the tight supply of GB200, customers are said to be scrambling to secure their orders from Foxconn due to its ample chip supply. Foxconn, traditionally silent on customer and order details, will reveal the latest status of its product lines during the press conference on August 14th.

The GB200 was originally scheduled for mass shipments starting in the fourth quarter of this year. However, over the weekend, reports emerged about yield issues, pushing the mass shipment timeline to the first quarter of next year, causing a stir among the market.

Read more

• [News] GB200 AI Servers Generate USD 210 Billion Annual Revenue, Boosting Supply Chain Growth
• [News] Foxconn Rumored to Secure Exclusive NVIDIA GB200 AI Deal

(Photo credit: NVIDIA)

As the demand for AI GPUs increases, TSMC’s advanced packaging capacity for CoWoS is struggling to keep up. Recently, according to a report from Commercial Times, NVIDIA has reportedly turned to Intel for advanced packaging solutions.

According to industry sources cited by the same report, TSMC’s CoWoS-S and Intel’s Foveros packaging technologies are similar, allowing clients to turn to Intel and secures the capacity needed quickly.

Despite its current struggling on transformation, Intel has been gradually developing its ‘s foundry services. In addition to clients like Qualcomm and Microsoft, Intel’s advanced packaging has also attracted interest from companies like Cisco and AWS.

Under the IDM 2.0 strategy, Intel has opened up its wafer outsourcing and foundry services to customers, establishing an the independent IFS foundry service. Earlier this year, Intel secured a major USD 15 billon foundry order from Microsoft for the first system-level AI foundry service, which is expected to use the Intel 18A process.

The report from Commercial Times further suggested that Microsoft’s move is anticipated to reduce its heavy reliance on TSMC. The report also indicates that chip customers, including NVIDIA, have engaged with Intel. Intel’s flexible foundry strategy, which can provide advanced packaging, software, and chiplet services tailored to customer needs, has been well-received by chipmakers.

Sources cited by the same report reveal that the U.S. has begun allocating specialized funds to increase investments in the advanced packaging sector as well. This move could highlight the importance of advanced packaging as the next key area for global competition in production capacity.

In November last year, the U.S. Department of Commerce’s National Institute of Standards and Technology (NIST) released a report titled “National Advanced Packaging Manufacturing Program,” highlighting that advanced packaging technology is one of the key technologies in semiconductor manufacturing.

Additionally, the U.S. Department of Commerce plans to invest approximately USD 3 billion to advance the National Advanced Packaging Manufacturing Program. Intel, alongside Amkor, is another giant in local advanced packaging in the U.S.

The main focus of advanced packaging is on interconnect density, power efficiency, and scaling. From Foveros to hybrid bonding technology, Intel is gradually scaling down bumping pitch sizes, which allows for higher current loads and better thermal performance.

Furthermore, in May last year, Intel’s advanced packaging technology roadmap outlined plans to transition from traditional substrates to more advanced glass substrates.

Read more

• [News] Intel’s Earnings Fall Short, Cutting Over 15% of Workforce
• [News] Intel Hires Micron Executive to Lead Its Foundry Business

(Photo credit: Intel)

Amid the heating tech war between the U.S. and China, and the stringent sanctions imposed to prevent China from obtaining cutting-edge chips, it appears that China is still able to find its way out. According to a report by Tom’s Hardware, citing the New York Times, the latest tactic of China would be setting up new companies to trade advanced hardware and operating them until they are shut down.

Before that, it is understood that Chinese firms have been smuggling NVIDIA chips through some underground networks, which involve buyers, sellers and dispatchers, according to a previous report by the Wall Street Journals.

Now the country seems to find another option in order to evade the sanctions. According to the New York Times and Tom’s Hardware, buyers that include state-owned or affiliated companies, even sanctioned companies, are reportedly collaborating with the Chinese defense industry, as transactions ranging from a few hundreds of GPUs to a deal worth USD 103 million have been observed lately.

The new tactic, it is reported, would be to establish new companies to acquire advanced chips before facing U.S. sanctions. For instance, after Sugon, established under the strong promotion of the Chinese Academy of Sciences and focuses on fields like computing, storage, security and data center, was banned from obtaining NVIDIA chips due to its ties with the Chinese military, some former executives created a new company named Nettrix.

The reports further note that within six months, Nettrix became one of the largest Chinese manufacturers of AI servers, as tech giants including NVIDIA, Intel, and Microsoft have already begun doing business with it, all without violating any American laws. Given the company’s recent establishment, the U.S. likely hasn’t had the opportunity to thoroughly vet its background.

The reports suggest that the White House might significantly reduce Chinese backdoors in trade by ensuring that only licensed, white-listed buyers can legally procure these chips. However, many in the industry oppose increasingly stringent bans, arguing that they harm American companies more than they help.

Before the upcoming U.S. presidential election in November, the Biden authority is said to be considering a series of actions targeting semiconductors. The latest one includes new measures that might unilaterally impose restrictions on China as early as late August, preventing major memory manufacturers like Micron, SK hynix, and Samsung Electronics from selling high-bandwidth memory (HBM) to China.

• [News] US Reportedly Weighs Stricter Limits on AI Memory Access for China
• [News] China Reportedly Acquires Nvidia AI Chips Indirectly; Nvidia, Gigabyte Emphasize Compliance with U.S. Regulations

(Photo credit: Nettrix)