The chip engineer is handling a wafer. Compared to NVIDIa’s export-stricken chips, the performance difference between Huawei and H20 is lower than a full generation, “said Dialen Patel, Founder, CEO and the chief analyst of Semyalisis.
Sinology | Moment | Getty images
Along with restricting China from purchasing advanced semiconductors used in Artificial Intelligence Development, Beijing is giving expectations on domestic options like Huawei.
This task has been made more challenging than the fact that US Carb not only prevents China’s access to the most advanced chips in the world, but also takes advantage of important technology to create AI chip ecosystem.
Those barriers expand the entire semiconductor value chain, in which design and manufacturing devices are used for supportive elements such as memory chips.
Beijing has collected Tens billions According to experts, to try to try to fill those intervals, but when it has been able to “cruel force” in some successes, it still has to go a long way.
Paler Triolo, a partner and senior vice-president for China at the Advisory firm DGA-Ablight Stonebridge Group, said, “US export control on advanced NVidia AI Chips has encouraged Chinese industry to develop options, while it is also more difficult for domestic firms to do so.”
It is reported here that China is required to create AI chips in four major segments against the rest of the world.
Ae chip design
Nvidia is considered as the world’s leading AI chip company, but it is important to understand that it does not actually build physical chips used for AI training and computing.
Rather, the company designs AI chips, or more accurately, graphics processing units. Orders for the company’s patented GPU designs are then sent to the chip founder -manufacturers who are specialized in the mass production of semiconductor products of other companies.
While American competitors such as AMD and Broadcom provide different options, GPU designs from NVidia are widely recognized as industry standard. The demand for Nvidia chips is so strong that Chinese customers have continued to buy any chips of the company that they can get their hands.
But Nvidia is struggling with strict sanctions from Washington. The company revealed in April Additional curb It was prevented from selling its H20 processor to Chinese customers.
NVIDIA’s H20 was a low sophisticated version of its H100 processor, especially designed to skirt previous export controls. Nevertheless, experts say, it was still more advanced than anything domestic. But China hopes to change it.
In response to restrictions, more Chinese semiconductor players are entering AI processor Arena. He has included a wide array of upstarts, such as Enflem Technology and Berene Technology demands absorbing. Billions of dollars in demand for GPU left by Nvidia,
But no Chinese firm appears closer to providing a true option for Nvidia compared to Hisilicon, a Chinese firm Huawei’s chip design.
Huawei’s most advanced GPU in large production is its Ascend 910B. The next generation of ascending was 910c Allegedly It is expected to start a large -scale shipment in early May, although no updates have been revealed.
Dylan Patel, CEO of Seminalis and founder of the chief analyst, told CNBC that when ascending chips stay behind Nvidia, they show that Huawei is making significant progress.
“Compared to Nvidia’s export-stricken chips, the performance difference between Huawei and H20 is less than an entire generation. Huawei is not far behind products that are allowed to sell NVidia in China,” Patel said.
He said that 910B was two years behind NVIDIA like last year, while Ascend 910C is only one year behind.
But when it suggests that China’s GPU design capabilities have made great progress, the design is just one aspect that stands in the way a competitive AI chip ecosystem is made.
Ai chip fabrication
To build its GPU, NVIDIA depends on the world’s largest contract chip foundry TSMC, which produces most of the world’s advanced chips.
TSMC complies with US chip control and the US trade blacklist is also prohibited from taking any chip order from companies. Huawei was placed on the list in 2019.
This has inspired Chinese chip designers such as Huawei to list local chip founders, the largest of which is SMIC.
Smic TSMC is far behind – it is known to be able to produce officially 7 nanometers Chips require less advance techniques than 3-nenometer production of TSMC. Chip processing power and efficiency more than the size of the small nanometer.
There are indications that SMIC has progressed. The company suspects to be behind the 5-nenometer 5G chip for Huawei’s Mate 60 Pro, which was, which was. Self-confidence Us chip control in 2023. The company, however, can go a long way, before it can produce the advanced GPU in a cost-skilled manner.
According to Independent Chip and Technology Analyst Ray Wang, SMIC’s known operation capacity is dwarf by TSMC.
“Huawei is a very good chip design company, but they are still without good domestic chipmakers,” Wang said, seeing that Hua is Huawei Allegedly Working on your own construction capabilities.
But the lack of major manufacturing equipment is in the way of both companies.
Advanced chip tool
The capacity of SMIC to meet the GPU requirements of Huawei is limited by the familiar problem of export control, but in this case, from the Netherlands.
While there can be no major semiconductor designers or manufacturer in the Netherlands, it is home ASMLThe world’s leading suppliers of advanced chipmaking devices – machines that use light or electron beams to transfer complex patterns on silicon wafers, which form the basis of microchips.
According to US export controls, the country has agreed to block the sale of ASML’s most advanced ultraviolet (EUV) lithography machines. Equipment is important for creating a scale and cost-effectively advanced GPU.
According to Jeff Coach, an analyst of semi -neckis, EUV is the most important obstacle for Chinese advanced chip production. “Most of the other tooling are available with them, but lithography is limiting their capacity of 3Nm and the process nodes below,” he explained to CNBC.
SMIC has found Ways to work around Lithography restrictions using ASML’s less advanced deep ultraviolet lithography systems, which have observed comparatively low sanctions.
Through this “brutal forceing”, chips are produced at 7 Nm, but the yield is not good, and the strategy is likely to reach its limit, the coach said, “The current yield appears that SMIC appears that SMIC cannot produce enough domestic accelerator to meet the demand.”
Lithography is a Chinese company working on technology Allegedly Is connected to Huawei.
The coach said that copying the existing lithography tool may take years in decades. Instead, China is likely to pursue other techniques and various lithography techniques to carry forward innovation rather than copying.
AI memory component
While GPU is often identified as the most important components in AI computing, they are only away from those. To operate AI training and computing, GPU must work with memory chips, which are capable of store data within a broad “chipset”.
In AI applications, a specific type of memory industry has become a standard standard, known as HBM. South Korean SK Hinix HBM has led the industry. Other sector companies include Samsung and US-based Micron,
Analyst Wang said, “This phase of AI progress has become necessary to run high bandwidth memory training and AI model.”
With the Netherlands, South Korea is collaborating with US -led chip sanctions and has started complying with it Fresh karb On the sale of some HBM memory chips to China in December.
In response, Chinese memory chip maker Changxin Memory Technologies, or CXMT, chip-packaging and test company in partnership with Tongfu microelectronics, in the early stages of production of HBM, according to A Report by Reuters,
According to Wang, CXMT is expected to be behind global leaders in HBM development three to four years, although it faces major obstacles, including export control over chipmaking devices.
Semi -linked Estimated in April This CXMT remained one year away from ramping any appropriate amount.
Chinese Foundry Wuhan Xinxin Semiconductor is manufacturing Allegedly Construction of a factory to produce HBM wafers. A Report SCMP stated that Huawei Technologies had participated with a firm in the production of HBM chips, although companies did not confirm the partnership.
Huawei has bent over HBM StockPiles from suppliers like Samsung for use in its Ascend 910C AI processor, Semiyalis said in April ReportGiven that even when the chip was designed domestic, it still depends on the foreign products despite the restrictions or earlier.
“Whether it is HBM from Samsung, wafers from TSMC, or tools from America, Netherlands and Japan, is a major dependence on foreign industry,” Semianlisis said.
