The Race Is Heating Up: Inside The Intel, TSMC, and Samsung's High-Stakes '2nm Competition'
AInvestThursday, Sep 5, 2024 11:09 am ET
4min read
INTC --
TSM --

The semiconductor foundry sector once proclaimed a big three at the beginning of the year, with TSMC, Samsung Electronics, and Intel all confident in their 2-nanometer chip manufacturing processes, and many were waiting for a reshuffling in the industry.

However, before the third quarter of 2024 has ended, Intel may be the first to be "struck out". The company is about to hold a board meeting, and speculations are circulating in the market that Intel may sell its chip foundry business, which also fills the future of Intel's chip business with unknowns.

What's more worrying is that the 18A process, the most significant part of Intel's chip dream, seems to have encountered significant problems. It is likely to become the latest setback for Intel's chip foundry department, further intensifying market doubts about its future.

Is the story of the "big three" about to end?

Intel's Dream

In 2021, Intel launched its chip foundry business, and the company's CEO, Pat Gelsinger, called it an important part of Intel's revival strategy. At that time, he announced that Intel had formulated a four-year, five-process nodes plan, and 18A, which is the production of 1.8-nanometer chips, is the last node.

According to the plan, the 18A process will be put into production in the second half of 2024, which will be the moment when Intel overtakes its competitors, TSMC and Samsung. At that time, both TSMC and Samsung Electronics conservatively estimated that their 2-nanometer process mass production time would have to wait until 2025.

However, according to three informed sources, chip manufacturer Broadcom conducted a trial production of the 18A process and recovered the tested silicon wafers last month. However, after engineers and executives studied it, Broadcom believed that Intel's 18A was not yet suitable for mass production.

In response to this test, an Intel spokesperson said that 18A has been put into use, is running well, and has good output, expected to be mass-produced next year. The entire industry is very interested in Intel's 18A. Broadcom said that the company is evaluating Intel's foundry products and services and has not yet reached a conclusion.

Sources pointed out that Broadcom engineers seem concerned about the yield rate of Intel's 18A, which means that the number of defective products on each wafer exceeds expectations.

Samsung's Desire to Overtake

Intel's 18A process uses RibbonFET transistors and backside power supply technology. Among them, RibbonFET adds backside power transmission technology, which is better than Samsung and TSMC's GAA technology in 2-nanometer processes in terms of performance improvement and voltage reduction.

In theory, a perfectly made Intel 1.8-nanometer chip will be more competitive than TSMC and Samsung's 2-nanometer chips. But theory and reality are often not consistent.

And Intel's difficulties also mean that Samsung or TSMC is more likely to win the 2-nanometer technology race. Unfortunately, Samsung, which is also trying to overtake, has a rough road ahead.

According to data from market research firm CounterpointResearch, in the second quarter of 2024, Samsung's foundry business market share was only 13%, far behind TSMC's 62%, and this gap remained consistent with the previous quarter.

The industry estimates that if this share gap remains unchanged, Samsung's chip foundry business will lose more than 1 trillion won, about $750 million, this year. The main reason for the gap is that Samsung Electronics finds it difficult to attract major customers from TSMC.

The Monopolized 3-Nanometer Market

In the market for the most advanced 3-nanometer chips that have been put into production, TSMC has almost achieved a winner-takes-all situation. TSMC's 3-nanometer production line has been running at full capacity this year, thanks to large orders from Intel, Apple, Qualcomm, and MediaTek.

It is understood that TSMC's 3-nanometer process capacity has been booked until 2026, and a group of major manufacturers planning to update consumer electronics in the next one or two years are competing for TSMC's capacity.

In contrast, TSMC's 3-nanometer chip process has been continuously increasing in price in the market because it is hard to get, with its wafer price now above $20,000; while Samsung's 3-nanometer process is still being sold at a low price.

The reason why the situation is completely leaning towards TSMC is that the yield rate has played a key role.

The so-called yield rate is the ratio of the actual output of chips to the total input. Ensuring a stable output is one of the most important indicators for chip foundries. The more usable chips on each silicon wafer mean more cost-effectiveness and output benefits.

Samsung's attempt to overtake was an important reliance on updating the traditional FinFET transistor technology to GAA technology at the 3-nanometer process. The advantage of this technology is that it can more accurately control the current, improving the power efficiency and performance of the chip, but too advanced technology also means a straight decline in yield rate.

Yield Rate Defense War

According to Notebookcheck, Samsung's 3-nanometer process yield rate is hovering around 50%; and according to a South Korean media report in February, Samsung's new 3-nanometer process has a major problem.

The most direct consequence of this problem is that Google has given the Tensor processor order back to TSMC. Google once intended to have Samsung produce all processors before the fourth generation of Tensor, but after seeing the effect of Samsung's 3-nanometer process, it quickly abandoned this idea.

In June of this year, the famous analyst Ming-Chi Kuo also warned that Samsung's self-developed Exynos 2500 processor 3-nanometer chip yield rate is lower than expected and cannot be shipped.

TSMC, which still retains the FinFET technology in the 3-nanometer process, is beginning to reap the rewards. Although the 3-nanometer process also poses a yield rate challenge for TSMC, compared to its only competitor, Samsung, TSMC's performance is more recognized by major manufacturers.

This also makes Samsung go all out in the 2-nanometer competition. Analysts say that Samsung's future in chip foundry depends on whether its 2-nanometer process can take the lead, and secondly, on its ability to transform into artificial intelligence, high-performance computing, and automotive electronics manufacturing.

But Samsung doesn't have much time left.

The 2-Nanometer Competition Enters the Home Stretch

Reports say that TSMC began trial production of 2-nanometer process chips in mid-July this year, earlier than the market estimate for the fourth quarter.

In May of this year, TSMC's Senior Vice President of Business Development and Co-Chief Operating Officer, Xiaoqiang Zhang, revealed that TSMC's 2-nanometer process is progressing very smoothly, and its nanosheet transition performance has reached 90% of the target, that is, the yield rate is over 80%.

But Samsung is not without hope. On July 9, Samsung announced that it will cooperate with artificial intelligence startup Preferred Networks to manufacture AI chips for the latter based on 2-nanometer process technology and 2.5D packaging technology I-Cube S. This cooperation is also seen as a milestone breakthrough for Samsung in the foundry: it finally has a serious big customer!

Intel is also struggling. Last month, Intel CEO Gelsinger said that this summer, Intel has begun to release its 18A technique manufacturing tool kit to chip manufacturers, ready to mass-produce for customers. At an investor meeting last week, Intel also said that more than a dozen customers are actively using the tool kit.

This seems to mean that Gelsinger still wants to fight a way out for the foundry business, which has been losing money for years, but it is unknown whether the 18A technique can bear such heavy expectations.

Disclaimer: the above is a summary showing certain market information. AInvest is not responsible for any data errors, omissions or other information that may be displayed incorrectly as the data is derived from a third party source. Communications displaying market prices, data and other information available in this post are meant for informational purposes only and are not intended as an offer or solicitation for the purchase or sale of any security. Please do your own research when investing. All investments involve risk and the past performance of a security, or financial product does not guarantee future results or returns. Keep in mind that while diversification may help spread risk, it does not assure a profit, or protect against loss in a down market.