news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

intel ceo kissinger and qualcomm ceo amon

phoenix.com technology news: on september 21, beijing time, the new york times reported on friday that qualcomm has tentatively approached intel to see if it is possible to acquire part of intel's business. subsequently, business insider published an article saying that even if qualcomm and intel "marry", they will not be able to solve the chip manufacturing dilemma in the united states.

here is the article excerpt:

"real men need to own their own fabs." this is a famous quote once said by amd co-founder jerry sanders, which is intended to satirize semiconductor companies that cannot manufacture chips themselves.

the quote dates back to the late 1980s and is sexist, so it needs to be updated. but in 2024, the core idea of ​​sanders' quote is still applicable, except now it would be better to change it to "real countries need to own their own wafer fabs."

fabs are factories where semiconductors are made. these massive facilities take billions of dollars and years to build, and they are difficult to run efficiently.

for decades, intel dominated the chipmaking world. however, around 2018, intel’s lead began to crumble due to a series of missteps. taiwan’s tsmc came from behind, grew steadily, and is now undoubtedly the best chipmaker in the world.

since 2018, intel's stock price has plummeted, while tsmc's stock price has soared. intel's current market value is less than $100 billion, not even in the top 150 in the world. in contrast, tsmc's market value is close to $1 trillion, ranking in the top ten in the world.

why does the united states need wafer fabs?

intel's fall from grace has created a major strategic and geopolitical problem for the united states.

if the u.s. wants the best chips, it has to go to taiwan or south korea, where samsung electronics has also built a strong foundry business making semiconductors for other companies.

in the united states, many well-known "chip manufacturers" do not actually produce their own chips. in fact, nvidia, qualcomm, amd and other chip companies design chips and then have them manufactured by tsmc. apple and many other large technology giants do the same, designing their own chips and having tsmc produce them.

intel

however, it is extremely difficult to produce these complex chips flawlessly on a large scale. if there are any accidents in tsmc's production, it will be disastrous for the united states and even europe. without more advanced chips, the performance upgrade of iphone will be delayed. nvidia will also need to find other suppliers to produce its gpus, and the development of artificial intelligence may stagnate.

therefore, "a real country should have its own fab" is now a more appropriate slogan. chips drive the modern economy. if a country must rely on overseas to obtain these key components, then you will be passive.

that’s why intel’s decline is so worrisome. it’s the only american company that knows how to mass-produce high-performance chips. globalfoundries also makes semiconductors, but not cutting-edge ones.

about two weeks ago, reuters reported that qualcomm was considering acquiring some of intel's businesses. on friday, the wall street journal followed up with a report saying that qualcomm had approached intel in recent days to explore the possibility of acquiring intel.

manufacturing dilemma

even if the deal is reached, their cooperation will not solve the chip manufacturing problem in the united states. one analyst even used the word "strange" to describe the potential cooperation.

qualcomm may not be interested in intel's manufacturing business. according to reuters, the company is interested in some of intel's chip design business. intel has two main businesses: one is designing semiconductors for personal computers, data center servers and other devices; the other is manufacturing these designs.

intel's design and manufacturing operations have been tightly integrated for decades, and that model has worked well for a long time. the company can build factories to the precise specifications given by its in-house chip designers.

however, the world is beginning to move toward an alternative model pioneered by tsmc: instead of designing and manufacturing chips itself, why not just operate fabs and produce chips for other companies?

tsmc

when tsmc was just getting started in the late 1980s, the idea was ridiculed, prompting amd co-founder sanders to make a sexist comment that "real men need to own fabs." but tsmc's model has gradually caught on, driven by intel's missteps and modern technological advances.

intel's biggest mistake was missing out on making chips for the iphone, a revolutionary phone that had just been released. apple eventually chose tsmc. qualcomm, another major smartphone chip designer, had tsmc make most of its chips. other chip designers, such as amd, also began turning to tsmc.

this gave tsmc access to huge chip manufacturing volumes from many different customers, allowing it to learn how to make chips better than any other company.

while tsmc learns from a variety of large customers, intel's manufacturing business has only one customer: itself. as smartphone chips become mainstream, intel does not have enough production to compete with tsmc in chip manufacturing. to make matters worse, the development of artificial intelligence has made this situation even more serious. nvidia is far ahead in this field, and tsmc is responsible for manufacturing its gpus.

for intel, extricating itself from its manufacturing woes will be an expensive, risky and complicated endeavor. the company has even begun hiring tsmc to make some of its chips, but the road ahead is still long.

the u.s. company recently took a major step: further separating its foundry business from its chip design business, which will make it easier for outside customers to trust intel to make their chips without having to worry about competing with them.

but the next challenge is the biggest: regaining the ability to make quality chips. intel's foundry business can't really challenge tsmc until it gets a few big customers. again, to be good at chip manufacturing, you have to have a huge volume of production from a variety of customers so you can find defects, fix the process, and feed that knowledge back to your fab.

it's a chicken-and-egg problem. without large volumes, outside customers will be wary of letting intel make the valuable chips they design. but without customers, intel can't improve its technology.

does spin-off work?

since qualcomm is unlikely to acquire intel's chip manufacturing business, will intel be split up? this is something that has been speculated in recent months.

how would intel's foundry business perform if it were separated from its chip design business and became an independent company? the question is again about volume. currently, intel's foundry business produces chips designed in-house. if even these businesses are gone, the company will have almost no production base from which to learn and improve technology.

“we don’t know how to separate,” stacy rasgon, a chip analyst at bernstein, wrote in a report in early september. “given the huge losses and lack of scale in the manufacturing unit, it cannot operate independently now.”

he added that the spin-off would only make sense if it attracted a significant amount of third-party business, which seems to be years away, or may never happen.

for more first-hand news, please download the phoenix news client and subscribe to phoenix technology. if you want to read in-depth reports, please search "phoenix technology" on wechat.