As the last glory of the US semiconductor manufacturing industry, Intel has been entrenched in the semiconductor industry for half a century and has created countless global firsts.
However, Intel, which was once at the forefront of the industry, is now a hero. In addition to the inability to keep up with the advanced manufacturing process, AMD has made a strong breakthrough, and recently it has been confirmed that Intel is about to outsource the production of high-end chips to old rivals.
According to Bloomberg, US chip giant Intel has negotiated with TSMC and Samsung Electronics to outsource high-end chip production to these foundries. Following the “breakup” between Intel and Apple, Intel’s advanced process technology also failed to break through.
Samsung and TSMC are eyeing, who will Intel choose?
According to sources, the decision was made because of continuous delays in the development of Intel’s chip manufacturing process, so Intel intends to outsource the manufacturing business, but no final decision has been made yet. And, Intel CEO Bob Swan (Bob Swan) has also promised investors that it will develop outsourcing plans and get its production technology back on track when Intel reports earnings on January 21. Swan also said at an investor conference that he made the decision because of the need to ensure Intel has enough factory capacity.
While Intel has previously outsourced the production of low-end chips, it has kept some of its best semiconductor manufacturing operations in-house as its core strength. The company’s engineers have traditionally custom designed designs based on the company’s manufacturing processes, and outsourcing flagship products was unthinkable in the past. However, due to the constraints of advanced chip manufacturing process, Intel has not yet had enough capacity to digest some orders. Therefore, we can only pin our hopes on TSMC and Samsung, and the outside world is also curious about how Intel will choose?
If it is only from the previous case, it should be TSMC with a high possibility. After all, the two cooperated in 2018, when Intel outsourced some 14nm chip production to TSMC. According to industry sources, Intel has recently purchased a large number of chips and other components from TSMC, but this batch of goods will not be put into the market until 2023 at the earliest. In addition, TSMC also plans to tailor a chip manufacturing capability based on the 4nm process for Intel, but will use the 5nm process in preliminary tests.
Not only that, according to market expectations, as long as TSMC can take three-quarters of Intel’s orders, it will bring in an additional $10 billion in annual revenue.
Of course, in order to achieve the above conjectures, TSMC’s efforts are needed to cooperate. After all, TSMC’s 7nm chip production capacity is already at full capacity, and even Qualcomm, Nvidia, etc. are queuing up. In this state, if TSMC wants to accept a large number of orders from Intel, it needs to expand additional production, or even build a new factory.
But from the perspective of Intel itself, it is only because the process is one step behind that it has to seek foundry. According to Intel’s overall development strategy, after the technology is brought up, it will eventually have to produce 7nm chips by itself. How will TSMC’s extra production capacity be vented? Therefore, TSMC may not take orders from Intel.
In contrast, Samsung is more likely to be the big winner.
In terms of market share, Samsung is far behind TSMC, and it was previously reported that Nvidia and Qualcomm would give up their cooperation with Samsung and hand over orders to TSMC. As a result, Samsung’s production capacity will have a large surplus, which can just be used for Intel’s foundry production, and Samsung’s wafers are cheaper in terms of price, and TSMC’s foundry price has always been higher than the market average.
In fact, Intel and Samsung seem to have a good personal relationship. Intel Chief Architect Raja Koduri (Raja Koduri) had previously visited Samsung’s Kiheung factory in South Korea and also participated in Samsung’s Advanced Manufacturing Forum last year.
Most importantly, Samsung’s technology can also fully meet Intel’s production standards. Currently in the foundry industry, only Samsung and TSMC provide EUV technology for chip manufacturing. Soaring customer demand for EUV processes will provide excellent opportunities for Samsung’s foundry business. And currently only Samsung and TSMC can mass-produce 7nm chips.
Based on the above reasons, some industry insiders also speculate that Intel is more likely to choose Samsung.
Intel in Internal and External Troubles
Some commentators believe that Intel’s current embarrassing situation is entirely caused by previous wrong decisions.
In previous years, other semiconductor giants in the United States were closing or selling component production, outsourcing low-margin, low-tech operations. In contrast, Intel still adheres to the road of the overall main business. This strategy has led to the dispersion of Intel’s main business resources, the erosion of its advantages a little bit, and finally led to its market value being surpassed by its old rival AMD.
First, is the weakness of the core business. The data center business, which sells server chips, is Intel’s main source of profit, and its main customers are enterprises and governments, and this part of the revenue plummeted in the third quarter of last year. Data center revenue fell 7 percent, or 4 percent on an annualized basis, and revenue from government and enterprise customers plunged 47 percent.
“Demand is shifting from desktop PCs and high-end enterprise PCs to entry-level consumer and educational PCs,” said George Davis, Intel’s chief financial officer. “Although sales are good, average selling prices are falling, so that’s going to be a bit of a drag on gross margins.” impact.” While revenue from cloud computing customers and network operators has acted as a buffer, the low prices of these chips and limited margins do not materially help.
Because Intel has always insisted on self-management of the chip business, the price of Intel’s products is much higher than that of its peers. Of course, the advantages of the self-operated model are also obvious. The technology and product quality have won a good reputation in the market. Intel can also gain a good market share. It’s a pity that the good times don’t last long. This year’s old rival AMD’s technology has also been improved, and it has also given the market a lower price by outsourcing, which has won the highest market share since 2003 in one fell swoop.
In fact, as early as 2018, when there was a problem with Intel’s 14nm chip inventory, AMD seized a large number of markets with sufficient supply from outsourcing to TSMC, narrowing the gap with Intel.
As the saying goes, good fortune comes, and misfortune does not come singly. Intel not only has problems with its core business, but its future growth engine, cloud computing processors, has also slowed. Its sales to cloud service operators rose only 15% in the third quarter, compared with a 47% increase in the second quarter, and cloud-related sales are expected to slow further in the fourth quarter as customers enter a “digestion period” .
After the 10nm process was put into production, Intel hopes to accelerate the development of the 7nm process in order to shorten the gap with TSMC. It originally planned to put the 7nm process into production this year. However, in the conference call after the earnings report, Swan once again confirmed that the 7nm process will be delayed. He said that the self-produced 7nm chips will not be unveiled until the second half of 2022 to early 2023. TSMC has already put into production the 5nm process and is expected to put into production the 3nm process in 2022. So far, Intel has completely fallen behind in the chip manufacturing process.
In contrast, Intel’s data center rivals AMD and Nvidia are expected to post full-year revenue growth of 32% and 44%, respectively, and also surpass Intel’s current market cap.
Analyst firm Barron said Intel is now in its biggest crisis. Although from the data point of view, Intel’s revenue has not reached the point of “hopeless”, but “lagging behind” can cause a company to collapse very quickly. More seriously, Intel has fallen behind on the new generation of chips. To develop the next-generation advanced process technology, it still needs a lot of capital investment, but now Intel has been unable to convince investors that they can solve this problem.
Declining US semiconductor industry
The hollowing out of manufacturing in the United States has long been put on the table. Due to the rapid development of the financial market in the United States, the asset-heavy industries have been gradually abandoned. Enterprises only focus on repurchasing stocks, distributing red envelopes, raising stock prices, and pleasing shareholders and Wall Street capital. Outsourced the core manufacturing industry that depends on survival, and Intel is also obsessed with it.
In 2019, Intel launched a $20 billion repurchase program, repurchasing $7.6 billion in stock. In August last year, the acquisition plan accelerated again. Throughout 2020, Intel’s planned capital expenditures are only $15 billion, even lower than the $16.2 billion in 2019, and there is absolutely no sign of increasing investment in R&D, nor does it intend to strive for excellence in manufacturing, instead focusing on outsourcing.
Not only Intel, but many IDM companies in the United States have begun to outsource chips to foundries. For example, TI has outsourced logic and embedded IC production to foundries, Broadcom also intends to sell its wireless chip business, and AMD has also separated chip manufacturing and sold it to Middle Eastern companies.
Xilinx, AMD, Nvidia, Qualcomm, Broadcom and other companies have also handed over manufacturing to Asian foundries, especially TSMC and Samsung. The chip manufacturing capabilities of TSMC and Samsung have also surpassed that of domestic companies in the United States, and Japan dominates at least 52% of the world’s share of semiconductor raw materials.
Moreover, while the chip is outsourcing, some semiconductor companies in the United States are gradually closing their factories. According to incomplete statistics, in the past two decades, Intel has closed three factories, SK Hynix has closed one, and Texas Instruments has closed two factories, and said that two 6-inch fabs will be closed in the next five years.
In my opinion, the US semiconductor industry currently has two major problems.
The first is the shortage of semiconductor talents. Because the United States vigorously develops the financial industry, there are fewer and fewer scientific research talents engaged in semiconductor research and development, and it is gradually developing in the direction of short supply.
The second is that manufacturing is deteriorating. Most US companies pursue high returns and high profits, so they have abandoned the relatively low-profit manufacturing industry and transformed to light assets. At the same time, these companies also outsource a large number of manufacturing links and a small number of design links, resulting in a gradual shortage of semiconductor manufacturing in the United States.
The reason why Intel chose foundry outsourcing this time was mainly from investors, so Swan had to make a choice in advance. And from the perspective of industrial development, Intel, which has a backward manufacturing process, does need foundry assistance from TSMC or Samsung.
But what cannot be ignored is that it will be even more difficult for Intel to return to the hegemony with TSMC and Samsung if it follows the development of Moore’s Law, and the human and financial resources to be paid have also increased exponentially. In the context of the “hollow” manufacturing industry in the United States, whether Intel’s shareholders and investors are willing to give up the investment direction of higher profits, and take a step back to choose advanced process technology research and development with lower returns, this is still the main reason for Intel’s dilemma .