You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Qualcomm Inc. Chief Executive Officer Cristiano Amon said Intel Corp.’s production technology isn’t currently good enough for the maker of mobile phone processors to use as a supplier. If Intel is able to advance its manufacturing techniques to produce more efficient chips, then Qualcomm would consider becoming a customer, Amon said Friday in an interview on Bloomberg Television’s Bloomberg Tech.
“Intel is not an option today,” Amon said. “We would like Intel to be an option.” Instead, the CEO said Qualcomm will stick with current producers Taiwan Semiconductor Manufacturing Co. and Samsung Electronics Co. Qualcomm is a designer of chips that, like most of the industry, relies on outsourced production. Intel, formerly the world’s largest chipmaker, is trying to turn around its fortunes by attracting outside customers, such as Qualcomm, in addition to manufacturing its own designs. The San Diego-based company, which gets the majority of its revenue from selling processors that are the heart of Android smartphones, is undergoing its own transformation but from a stronger base than Intel, which is facing shrinking sales. Under Amon, Qualcomm is pushing its mobile technology into new markets, such as automotive, seeking revenue growth that doesn’t depend on a smartphone boom. It has told investors it can secure $22 billion in revenue from automotive and connected devices by 2029.
I would bet 18A being targeted at PCs, laptops, and servers, and not mobile, is a problem for QCOM. Snapdragon laptops are probably not high volume enough to be a significant factor.
This is natural in a sense
Ultra-low power consumption applications will be fully supported from Intel 14A, so it is inevitable that 18a will not be considered. What Intel can do now is to prepare.
I would bet 18A being targeted at PCs, laptops, and servers, and not mobile, is a problem for QCOM. Snapdragon laptops are probably not high volume enough to be a significant factor.
Cristiano needs to be more transparent about suppliers. Samsung is a former supplier now making legacy chips. QCOM only went to Samsung for the price and they got burned on yield more than once. QCOM is now all in with TSMC at N4, N3 and N2. If you are exclusive to TSMC you get a much more intimate relationship. I compare it to flying first class versus coach. For those of you who have not done both let me tell you it is a HUGE difference. Once you go first class with TSMC you do not go back, absolutely.
All of these CEOs can claim they buy American made TSMC wafers but if you do the math on N4 capacity in AZ (about 250k wafers in 2025?) you will realize it is not possible. N3 will start high volume manufacturing in Arizona in 1H 2028? The top semiconductor companies will already be on N2 by then. N2 starts HVM in Taiwan in the 1H 2026.
I hope analysts start asking CEOs just how many US made wafers they are actually buying. Percentage wise it will be single digits which is why the Taiwan Silicon Shield gets stronger every year.
This is natural in a sense
Ultra-low power consumption applications will be fully supported from Intel 14A, so it is inevitable that 18a will not be considered. What Intel can do now is to prepare.
TSMC is the low power king thanks in part to Apple. Apple writes some very big ecosystem checks and low power is a big driver. Process is important but you have to have EDA tools to get you there and there is no company closer to EDA than TSMC. Intel Foundry has a good shot since they are a big EDA customer. Samsung not so much since they do not have the customer base to make it interesting for EDA or IP.
Cristiano needs to be more transparent about suppliers. Samsung is a former supplier now making legacy chips. QCOM only went to Samsung for the price and they got burned on yield more than once. QCOM is now all in with TSMC at N4, N3 and N2. If you are exclusive to TSMC you get a much more intimate relationship. I compare it to flying first class versus coach. For those of you who have not done both let me tell you it is a HUGE difference. Once you go first class with TSMC you do not go back, absolutely.
All of these CEOs can claim they buy American made TSMC wafers but if you do the math on N4 capacity in AZ (about 250k wafers in 2025?) you will realize it is not possible. N3 will start high volume manufacturing in Arizona in 1H 2028? The top semiconductor companies will already be on N2 by then. N2 starts HVM in Taiwan in the 1H 2026.
I hope analysts start asking CEOs just how many US made wafers they are actually buying. Percentage wise it will be single digits which is why the Taiwan Silicon Shield gets stronger every year.
Shush!!! Don't tell Trump, or TSMC will get in big trouble! Probably Apple too. I suspect no one senior in his administration understands the math regarding how small TSMC AZ's capacity is compared to the TW Gigafabs.
Cristiano needs to be more transparent about suppliers. Samsung is a former supplier now making legacy chips. QCOM only went to Samsung for the price and they got burned on yield more than once. QCOM is now all in with TSMC at N4, N3 and N2. If you are exclusive to TSMC you get a much more intimate relationship. I compare it to flying first class versus coach. For those of you who have not done both let me tell you it is a HUGE difference. Once you go first class with TSMC you do not go back, absolutely.
All of these CEOs can claim they buy American made TSMC wafers but if you do the math on N4 capacity in AZ (about 250k wafers in 2025?) you will realize it is not possible. N3 will start high volume manufacturing in Arizona in 1H 2028? The top semiconductor companies will already be on N2 by then. N2 starts HVM in Taiwan in the 1H 2026.
I hope analysts start asking CEOs just how many US made wafers they are actually buying. Percentage wise it will be single digits which is why the Taiwan Silicon Shield gets stronger every year.
700,000 wafers per year for both fabs combined is the last estimate I heard but that was before LBT became CEO. I did hear recently that 18A is doing well in Fab 52 but demand is the real question. Intel really needs to stop designing to TSMC. Lip-Bu will make sure of that, I do not think he wants to be fabless and neither does the US Government.
700,000 wafers per year for both fabs combined is the last estimate I heard but that was before LBT became CEO. I did hear recently that 18A is doing well in Fab 52 but demand is the real question. Intel really needs to stop designing to TSMC. Lip-Bu will make sure of that, I do not think he wants to be fabless and neither does the US Government.
Agreed. If they still have a monopoly over the products, that number makes sense. It's a constrained optimization problem — involving competition, capex, profit margins, and so on. I would think the first target should be for IFS to break even.
700,000 wafers per year for both fabs combined is the last estimate I heard but that was before LBT became CEO. I did hear recently that 18A is doing well in Fab 52 but demand is the real question. Intel really needs to stop designing to TSMC. Lip-Bu will make sure of that, I do not think he wants to be fabless and neither does the US Government.
If we divide Intel’s 700,000 yearly wafer capacity by two (for fabs 52 and 62) and then by 12 months, each fab can make about 30,000 12-inch wafers a month. After covering Intel’s own product needs, there isn’t much left for outside customers, especially those that need high volume.
Many people argue that Intel can’t split into two separate companies, Product and Foundry, because the Foundry side depends on the Product side for funding. By that logic, Intel’s Product division has to get stronger first.
Right now, most Intel’s revenue and profit come from its products, not from Foundry services. For Intel’s products to grow and sell more, the Product division should be free to use either Intel’s fabs or external foundries, depending on what makes the most money. Otherwise, how can the Intel Product have enough money to support their brothers and sisters at Intel Foundry? For Intel Foundry to survive, Intel’s products will need to use significant TSMC’s services.
I think Intel started to understand this “product first, revenue first, profit first” reality even when Pat Gelsinger was still CEO.
In practical terms - is it leakage power that makes N3 lower power than 18A? Or is it more actual power draw at "moderate" ( < 4GHz ) frequencies? Both?
The voltage curve even on Intel 7 tends to look pretty good at high frequencies ( > 4GHz) vs TSMC N4/N5, and 18A should be substantially better than 7.
Agreed. If they still have a monopoly over the products, that number makes sense. It's a constrained optimization problem — involving competition, capex, profit margins, and so on. I would think the first target should be for IFS to break even.
TSMC is the low power king thanks in part to Apple. Apple writes some very big ecosystem checks and low power is a big driver. Process is important but you have to have EDA tools to get you there and there is no company closer to EDA than TSMC. Intel Foundry has a good shot since they are a big EDA customer. Samsung not so much since they do not have the customer base to make it interesting for EDA or IP.
I know quite a few ex tsmc who jumped to intel foundry got the layoff inbthe latest round. I would wager that using tmg to cover for foundry pdk won't make the grade
If we divide Intel’s 700,000 yearly wafer capacity by two (for fabs 52 and 62) and then by 12 months, each fab can make about 30,000 12-inch wafers a month. After covering Intel’s own product needs, there isn’t much left for outside customers, especially those that need high volume.
Remember, it takes more time to design and tape-out a complex chip than it takes to build a fab. LBT was right on the money about not building fabs without customer orders. TSMC does not build fabs without customer orders. The problem TSMC has had in the past is competition not yielding and customers over ordering from TSMC. Now TSMC does not have competition so it is easier to plan capacity requirements. When you hear CC Wei constantly saying capacity is tight on investor calls that means he is doing his job.
