news

"core matters" is tencent technology's semiconductor industry research project. this issue is the 59th, focusing on the interpretation and analysis of the chip giant intel's predicament of being besieged on all sides. 

author: xie hao, silicon research lab, kiki

editor | zheng kejun and su yang from tencent technology

“the past few weeks have been tough and we decided to clearly show the current state of the company, but the market has not responded positively.”

on august 29th local time, at a california technology conference hosted by deutsche bank, intel ceo pat gelsinger rarely expressed negative emotions.

huge losses, obstacles in factory construction, plummeting stock prices, layoff crisis... the negative news that followed one after another has put this star company in the semiconductor field in a dilemma.

the turn of events was very rapid.

for the past decade, intel has been regarded as a "giant" in the global chip industry. gordon moore, one of its founders, proposed "moore's law" which has influenced the chip industry for more than half a century. even in 2023, intel, which has received the largest subsidy from the "chip act" to date, is still regarded as the mainstay of the united states' ambition to "reshape the chip manufacturing industry."

unfortunately, thisthe story of restoring the glory of american chips is on the verge of collapse as intel's performance becomes increasingly poor.

1. intel’s favored position: accident and necessity

two years ago, the chip act officially came into effect with the ambition of "bringing high-end manufacturing back to the united states."

this 1,054-page document was commented by the new york times (article published on july 27, 2022) as "the most significant intervention in industrial policy by the u.s. government in decades."

"the united states invented semiconductors. more than 30 years ago, the united states accounted for 40% of global chip production. later,the manufacturing industry, the backbone of our economy, was hollowed out, and semiconductor manufacturing went overseas. now this law brings semiconductors back to the united states... and we will lead the world again in the decades to come."when signing the $52 billion chips act, biden described the u.s.'s "chip dream" in his speech.

today, the chip act has been in effect for two years.

according to the report card issued by the white house on august 9, the u.s. department of commerce has announced that it has reached a preliminary agreement with 15 companies to includemore than $30 billion in total subsidies have been pledged as part of a $39 billion direct incentive package for the semiconductor manufacturing industry.

according to data from sia (semiconductor industry association), since the chip act was first introduced in 2020, major semiconductor manufacturers have launched more than 80 new projects in 25 states in the united states, with a total private investment of nearly us$450 billion.

receiving subsidies and building factories in the united states have therefore become a concrete practice of the american "chip dream" in the past two years.

among them, intel is not only the semiconductor giant that directly obtains the most funds, but is also strongly bound to the "chip act". an article from the cato institute, an independent think tank in the united states, believes that:“for many in washington, the success or failure of the chips act will depend on intel.”

the reason why intel is favored and becomes the most dazzling star company in the "chip act" is both a reality and historical necessity, but also an accident in the overall dream-making of the us government.

"the united states can at least rely on intel. without intel, no american company would manufacture cutting-edge processors." miller, associate professor of international history at tufts university in the united states, once described intel's lofty position in the us chip industry in "the chip wars."

u.s. commerce secretary raimondo also called intel "america's champion semiconductor company," describing it as "the only american company that can make cutting-edge chips."

this glorious historical position made intel thethe "chip act" focuses on the hot topics.

after that, as the bill moved forward, the biden administration soon discovered that they did not have many options left except betting on intel.

in contrast, u.s. domestic companies such as micron and microchip technology and overseas giants such as tsmc, samsung, and sk hynix have all officially announced investments in the united states. the huge investments and high-profile factory construction ceremonies also added color to the "chips act" for a time, but the factory construction path of the above-mentioned companies has either been a lot of noise but little results, or they have faced the dilemma of not being able to adapt to the local environment.

according to a previous report by the new york times, tsmc's chip factory in phoenix, arizona, usa, has run into "difficulties in starting production".faced with challenges such as cultural conflicts, labor competition and political pressure.

wu zihao, a former tsmc plant construction expert, told silicon labs that delayed production is actually very common in cross-country and cross-regional operations. in the short term, overseas companies building plants in the united states will mainly face two types of problems:first, on the construction side, there is a lack of professional construction talents.

"you can't get everything you want in the united states like you can in your home country," he said, citing an example. "for example, there are fewer people in the united states who can build semiconductor clean rooms."

the second is on the operational side, including supply chain costs of raw materials and manpower issues. supply chain issues mainly bring about freight costs, but they do not account for a large part of the overall cost of chip manufacturing. "what is more troublesome is actually people," wu zihao said.

tsmc is well-known for its strict work pace and has its own employment criteria on the operational side. "we can send engineers from taiwan, but the united states has unions and we need to protect local employment."

at the same time, judging from the amount of subsidies, although the policy is attractive, asian semiconductor manufacturers such as tsmc and samsung have limited production capacity in the united states, coupled with relatively high factory construction and operating costs.

under the inevitability of history and reality, there are also accidental factors in intel becoming the "child of destiny", and this accident stems from intel's own need for transformation.

before the chips act officially came into effect, intel, which had been resting on its laurels, missed the first-mover wave in the mobile market and ai because it relied too much on its past leading position. at the same time, in the latecomer competition with manufacturers such as nvidia and amd, it encountered difficulties in technology, market, capital and other aspects.

in 2021, intel's "veteran" kissinger was given high hopes when he took office. soon, he proposed the "idm 2.0" strategy, hoping to build intel into the world's second largest wafer foundry by 2030.

corresponding to the ambitious plan is a radical reform - while intel is investing heavily in building wafer fabs, it will also establish a separate wafer foundry service division in 2021 to open its wafer manufacturing capacity to external customers.

kissinger and intel's ideas undoubtedly coincide with the "repatriation of american manufacturing" advocated by the biden administration. intel, which is in a period of transition, is a representative of the former glory of american manufacturing. if it bottoms out and rebounds, this will undoubtedly be the most moving story script.

as a result, the "chip act" and intel entered a honeymoon period. intel not only received $8.5 billion in subsidies, but also ushered in its own factory construction boom.

so far, intel has invested in and built factories in arizona, new mexico, ohio, oregon and other states in the united states.biden personally supported intel's new factory and called the new wafer factory in ohio a "land of dreams."

in january this year, intel also officially announced that it had obtained asml's high numerical aperture (high-na) extreme ultraviolet (euv) lithography machine, and was determined to be the first to adopt the high numerical aperture extreme ultraviolet lithography machine to surpass its competitors in key technologies.

standing in the c position and with all the ammunition, intel does not seem to have got the script for a comeback in the past two years. the second quarter financial report was lower than expected, the large-scale layoffs and the sale of arm shares have plunged intel and the american chip dream into a new fog.

2. shadows under the glory: resistance and challenges

the biggest challenge hindering intel's efforts to revive its chip dream is undoubtedly its weak operating performance.

if you were to choose the "most disappointing" award among silicon valley giants in the past few years, intel would be one of the strongest contenders.

while the share prices of companies such as nvidia and amd have soared due to the ai ​​wave and the enthusiasm for us stock investment, intel has become an embarrassing outlier among the few technology stocks that are still falling. in the second quarter of this year, they handed in a disastrous financial report.

on august 1, local time, intel released its second-quarter financial report. its revenue for the quarter was us$12.8 billion, down 1% from the same period last year; its net income plummeted 85% to only us$83 million; and its loss reached us$1.61 billion.

for reference, intel's net profit attributable to shareholders for the full year 2023 is us$1.689 billion. in other words,this company lost all its entire year's profits in just one quarter.

even kissinger, who is always confident, could not embellish this financial report which was so difficult to find any bright spots even with a lantern:

"second quarter financial results remain 'disappointing'. our revenues did not grow as expected and we have not yet been able to benefit from ai."

affected by this news, intel's stock price fell 5.5% to $29.05 per share on the day the financial report was released. since then, intel's stock price has fallen to $18.99 at its lowest, corresponding to a market value of less than $100 billion, which is almost the lowest point of the company's market value since 2013. in comparison, nvidia's market value is 31 times that of intel, and amd's market value is more than twice that of intel.

larry raimer, an american stock analyst, first used the word "lifeless" to describe intel in an article on an investment research website: the company's technological backwardness is "causing them to lose huge market share."

the result of the above negative effects is that not only is intel unable to become the "lifeline" of the american chip dream - even the company itself needs a lifeline.

the "us$10 billion cost savings plan" released along with the financial results is a symbol of this.

in this plan, intel officially announced that "it will lay off 15% of its employees in the future", which means that 15,000 employees are expected to leave intel. the peak period of intel's employee size will be around 2022, and the number of employees will once exceed 130,000.

although the layoff plan is marked as "global", considering that data centers, artificial intelligence and foundry businesses have made "significant contributions" to the performance losses, the final focus may not be difficult to guess.if poor performance eventually leads to layoffs of u.s. employees in the foundry business, it would be a black humor in this beautiful dream.

after all, just on august 9, the white house released a summary document on the second anniversary of the launch of the chip act. the title of this document is so long and exaggerated that readers don’t need to read the text at all.

"it's proven: two years after the enactment of the chips and science act, the biden-harris administration's policy of bringing the semiconductor supply chain back to the united states has achieved historic achievements worthy of celebration. it has created jobs, supported innovation, and ensured national security."

if the situation is allowed to continue to develop, then before long, how the american chip dream will piece together the impossible triangle of "relying on intel", "intel layoffs", and "chip bill creating jobs" may become a hot puzzle.

putting aside its operating performance, the low competitiveness of intel's own chip foundry business has also greatly dragged down the company.

as tsmc founder morris chang summarized,the wafer foundry industry is essentially a "winner takes all" industry. the market will only choose the most advanced equipment, and falling behind means losing everything.

for quite a long time, intel's chip foundry business has been based on a "tick-tock" (pendulum mode). this mode is based on two years, one year focusing on chip manufacturing and process improvement, and one year focusing on chip design and architecture update.

while intel maintains its technological leadership, this model completely blocks the path for latecomers to overtake, almost making it "difficult for competitors to breathe."

however, with the continuous delays in the 10nm node and tsmc taking the lead, the drawbacks of this model were fully exposed. unable to focus on process technology upgrades, the struggling intel almost never recovered in the chip foundry business, and it can even be said to be "lingering on."

in terms of market share, according to trendforce data,in the third quarter of 2023, intel's market share in the top ten global foundry revenue rankings was only 1%.the market share of tsmc, the first place, is as high as 58%. in the first quarter of this year, intel failed to even maintain its top ten position in this ranking, while tsmc has won many championships.

in terms of customer scale, in stark contrast to the management’s bold statement that “intel has won customer favor at every process node”,there are few customers willing to pay for intel's foundry business.

since 2023, investors and the market have been waiting for a "flagship customer" that believes in intel.

however, various potential customers retreated due to a public opinion storm over the "13th and 14th generation core process issues" that occurred in july this year.

although intel repeatedly emphasized in subsequent announcements that the above-mentioned problems can be solved through microcode patches, considering that intel's own 3nm new products lunar lake/arrow lake are still on the production line of tsmc's factory, the market's doubts about intel's foundry capabilities are getting louder and louder.

chris caso, an analyst at wolfe research, said it is easier for intel to get customers like amazon, google or microsoft because they do not compete with intel. "but no matter which big customer, they need intel's guarantee that there will be no mistakes in manufacturing."

at a time when the above problems have not been effectively solved, even the most confident investors in intel may find it difficult to hide their doubts after listening to the rumbling sounds of drilling rigs in arizona, new mexico, ohio and oregon.does a wafer foundry department that lacks technology, customers and markets really need so many factories?

ironically, intel management's attitude towards the above-mentioned factory construction plan is very ambiguous.

recently, a reuters report stated that kissinger and other senior intel executives are brewing a plan to be submitted to management at the end of this month, which may include further cuts in the company's capital expenditures on factory expansion.

at the same time, speculation about intel splitting and selling its chip foundry business has been rampant.

when intel officially announced a five-year, $100 billion investment to build factories, some industry insiders said that the production progress of these factories would be very worrying.

in this regard, wu zihao, a former tsmc plant construction expert, also expressed a similar view to silicon labs:"in the chip production process, the biggest cost comes from equipment depreciation, and some of intel's factories have been slow to purchase this equipment."

this shows that intel's ultimate purpose of building a factory may not be business needs. "they may just use the project to occupy the carrot hole of the chip bill first, and then put it into production if the development is good. if the development is not good, building a few buildings will not bring too much loss."

another person who is not optimistic about intel's chip foundry business is lip-bu chen, who recently resigned from intel's board of directors. people familiar with the matter said that the core reason why this semiconductor industry veteran chose to leave intel was that he was frustrated with the company's contract manufacturing (chip foundry) methods and intel's risk aversion and bureaucratic culture.

based on the above news,intel's ambitious chip manufacturing plan may end up with a "tiger head but a weak tail".the company's loss of competitiveness in the chip foundry field is becoming a well-known fact.

from any perspective, intel is no longer the semiconductor legend that used to be surrounded by halo. with the emergence of many negative news, the lofty ideals and prospects woven by its predecessors have been repeatedly "demystified" by the market.

3. the american chip dream cannot be awakened

engineers are afraid of losing money. the us chip manufacturing industry is "short of people". it is difficult to recruit people with $100,000

andy grove, the most legendary ceo in intel's history, once wrote in "only the paranoid survive": "the word 'point' in strategic turning point is a misuse. it is not a point but a long and arduous struggle." this sentence also applies to intel, which is now besieged on all sides.

the reason why the story of the elephant's difficulty in turning around is repeated on intel is that there is a core theme:intel failed to wait for its own "jen-hsun huang" and "lisa su" like nvidia and amd did.

since the eighth ceo kissinger took office, the management's overly aggressive strategic moves have caused intel to continue to decline.

jim osman, founder of research firm the edge group, once commented: "kissinger's record is average at best, and intel still needs a new ceo."

however, it would be unfair to attribute all of intel's failures to a single ceo. rome did not fall in a day.intel's decline was not caused by the failed decisions of one or two people, but by the poor performance of four consecutive ceos in decision-making and execution.they include:

• fifth ceo: paul otellini (2005-2013) ——although he was not an engineer by training, he led intel to achieve unprecedented growth in revenue and profit. the pendulum strategy he proposed was once intel's key weapon to squeeze out its competitors. however, otellini failed to accurately foresee the rise of smartphones, and even rejected jobs' request to provide chips for the iphone because of price reasons.

• sixth ceo: brian krzanich (2013-2018) ——after otellini, krzanich, the chief operating officer, took over. he led intel's overseas investment and expanded the company's business from cpu to 5g, artificial intelligence, autonomous driving, cloud computing, drones and wearables. however, as moore's law slowed down, intel's pendulum model gradually failed. in the process of evolving from 14 nanometers to 10 nanometers, intel's chip iteration cycle was continuously lengthened. it was ridiculed by the outside world as a "toothpaste factory" and was surpassed by tsmc around 2018.

• seventh ceo: bob swan (2018-2021) ——when swan, who has a cfo background, became intel's new number one, the challenges he faced were a mixture of old and new: the old problem was that in terms of process technology, intel had clearly fallen behind tsmc, samsung and other companies, especially in 2020 when swan told investors that intel's new 7-nanometer chip technology was 6 months behind schedule and might outsource part of the chip manufacturing; the new problem was that due to lack of experience in the semiconductor industry, swan missed the opportunity to prepare for the ai ​​era faster. the most typical case was his refusal to invest in openai in 2021.

after going through the troubles of the first three,the eighth ceo, gelsinger, was once highly expected because he was both a technical background and a senior employee of intel. he once attributed intel's problems to the failed smartphone business, the inability to build a "great" manufacturing plant, and the cancellation of the gpu business in 2010.

in response to these old problems, kissinger formulated his own plan - in terms of technology, he increased investment in manufacturing and proposed a radical goal of "5 nodes in 4 years"; in the field of ai, in addition to benchmarking nvidia and accelerating the iteration of gaudi chips, intel has formulated a more open ai ecosystem strategy. in kissinger's words: "make ai ubiquitous."

but as analyzed above in the challenges facing intel, when a grander plan meets a ship that is difficult to turn around quickly in the short term, the outcome of kissinger's gamble seems to have been quietly written into fate.

in the years to come, intel, whose pace is slowing down, and the u.s. high-end manufacturing industry, which is struggling to regain its glory, are like two mirrors reflecting each other, always accompanying and following each other.

looking at the global chip industry chain, it has undergone three industrial transfers - the first stage was from the united states to japan; the second stage was from the united states and japan to south korea and taiwan, china; the third was from taiwan, china, south korea to mainland china. each industrial transfer was based on cost considerations to transfer lower value links abroad.

but industrial relocation is a dynamic competitive process.with the help of demand transfer and local innovation, different countries or regions have formed today's highly specialized chip industry division of labor. a core issue behind this is that behind the flow of industry is the flow of people.

"chip manufacturing is still manufacturing. it cannot be separated from the competition for engineers. if you don't compete for engineers, you will not be competitive. it's that simple. advanced technology is only a basic condition." wu zihao said, "this is a business model issue. asian engineers treat customers in a completely different way. it's not that i am ten or five times smarter than you. if you don't run the business 24 hours a day, you won't be able to do it."

take tsmc as an example. when it was first founded, its motto was "to be a partner of the customer." its work culture also advocates diligence and strictness, but this is contrary to the american workplace culture.

with an additional investment of nt$5 billion and the construction of three new factories, tsmc is investing heavily in the united states and transforming itself into "tsmc usa"?

according to foreign media "rest of world", when discussing competitiveness, morris chang said: "if (a machine) breaks down at one in the morning, it may not be fixed until the next morning in the united states. but in taiwan, it will be fixed at two in the morning."

in other words,the talent shortage may be the biggest problem hindering the revival of chip manufacturing in the united states.recently, mckinsey calculated from the perspective of talent that if the production capacity targets set out in the chip act are to be achieved, 160,000 new engineering positions will be needed. however, currently only about 1,500 engineers join the chip industry in the united states each year. as for the crucial chip production line technicians, 75,000 new positions will be needed in the next five years. however, currently only 1,000 new technicians enter the chip production line each year.

ultimately, no matter how high-end chip manufacturing is, it is still a manufacturing industry that requires human input, effort, dedication and hard work.

unfortunately, intel and the us chip manufacturing industry are at their lowest point and it is difficult to recover.

many stories cannot be changed by the power of a single company or a single bill. as for intel, after being deeply tied to the government and its chip dream, the company has in fact no way out.

no matter how far-fetched the ideal of reviving the u.s. chip manufacturing industry is, intel has only one choice:

blindfold your eyes and keep running forward.

related reading: