This week the pages of the Intel corporate blog featured a post by Keyvan Esfarjani, executive vice president in charge of manufacturing operations for the corporation around the world.He tried to reassure investors that deteriorating macroeconomic conditions would not force Intel to abandon its plans to build new facilities in the U.S.and Europe, although he acknowledged that funding for these projects should be carried out in proportion to projected demand.Source image: IntelThe main idea of the message of the Intel representative was that the company should already prepare for the future growth in demand for semiconductor products.It takes three to five years to build a plant and equip it with the necessary equipment, and the company simply cannot afford to ignore future market demands.Intel's management shares the experts' optimism, predicting an average semiconductor market turnover growth rate of 5 percent a year by the end of the decade.By the end of the forecast period, the market capacity should double to $1 trillion, so the company considers it necessary to invest in the construction of new enterprises, even now, when the current situation is not favorable to it.The second important idea, which broadcasts the executive vice president of Intel - the need for geographical diversification of chip production.With 80% of capacity concentrated in one tiny region (meaning Taiwan), semiconductor manufacturing is very vulnerable, according to Intel.Reportedly, this week the company acquired ownership of a piece of land in Magdeburg where a chip packaging and testing facility will be built, and eventually a second one will appear.The company has also signed a contract with Bechtel, a construction contractor that will start building new Intel facilities in Ohio.Here, the corporation expects to master mass production of chips using Intel 18A technology by 2025, with the U.S.Defense Department mentioned among the first customers.At the first stage, Intel expects to invest at least $20 billion at this site.A management representative made it clear that the company will work closely with German and Ohio state authorities to determine when and how much funding is needed for the relevant projects.Intel in this regard is going to be guided by the needs of the market and plan its major expenditures wisely.
The day before the Taiwanese media reported that TSMC, the largest contract manufacturer of semiconductor components in the world, had already decided on the location for its most advanced facility, which will master the 1N technology production in a few years.As explained by TSMC representatives, no decisions have been made yet.Image source: TSMC According to Focus Taiwan, the original publication of the Commercial Times, citing an anonymous source, mentioned the intention of TSMC to place the new venture in the technology park Xinzhu in northwestern Taiwan.According to official reports, TSMC is undecided on the site and is still in the process of finding a suitable location on the island.As soon as the choice is made, the company promises to make the appropriate announcement.Deputy director of the technology park Xinzhu not commented on rumors about the plans of TSMC, but explained that the resident companies can choose the site in the third area of the park for the construction of production facilities, because the first place is all taken, and the second reserved for recreational areas.It is expected that the company will begin mass production of 3-nm products in the current quarter, and next year revenue from their sales will be as follows Mass production of 2-nm products TSMC will begin in 2025.The company's management did not speak about terms of mastering more advanced lithography at the last quarterly report conference, but in May there was already information about TSMC's readiness to start development of 1.4nm process.
Describing the current state of adoption of new process technologies, Intel management avoided making any new ambitious statements at its quarterly report conference, merely mentioning that it continues to believe in its ability to regain technological leadership by 2025.It was also said that Intel 18A process will be more suitable for the needs of third-party customers, while the previous Intel 20A process will be mainly used by the company itself.Image source: IntelIn the next few years Intel will have to reduce costs by $8-10 billion a year, CFO David Zinsner was forced to clarify that the amount of investment in product development and adoption of new technologies will be affected to a minimum.In fact, over the next couple of years, the share of capital costs will reach 35% of revenues, before dropping to the more familiar 25%.The Intel 20A technology will use a new RibbonFET transistor structure, and in this regard, the company expects to maintain its status as an innovator in the semiconductor industry, as CEO Patrick Gelsinger made clear.The Intel 20A process is on track to be adopted, and the company's CEO believes it will be used primarily for its own needs.After transitioning to Intel 18A by the middle of the decade, the number of third-party customers interested in using this new lithography step will significantly increase.As Gelsinger stressed, a digital project on the Intel 18A process for a large external customer has already been completed, and the first workable prototypes are available.According to early statements from Intel, it is a product for the needs of the U.S.Department of Defense, which will be produced at the under-construction facility in Ohio.In addition, the company already has test samples for its own needs, manufactured using Intel 18A technology.In the future, several more Intel customers will receive test chips produced using this process.According to Gelsinger, based on the results of tests of these samples, they will be able to decide on the advisability of mass production by Intel.
Intel Corp.'s third-quarter revenue fell 15% to $15.3 billion, in full compliance with its management's projections.By the middle of the decade, Intel hopes to cut annual expenses by $8-10 billion, with savings reaching $3 billion as early as next year.According to the head of the company, staff cuts will not be a significant factor in savings, although they will start already in the current quarter.Image source: Intel Net income at Intel in the past quarter fell by 59% from $5.9 to $2.4 billion using Non-GAAP method and operating profit margin fell from 31.8 to 10.8 %.Profit margin fell from 58.3% to 45.9% year over year.At the same time, the company managed to increase spending on research and development by 12% to $5.4 billion.Intel was forced to end the quarter with operating losses, but due to the tax refunds it showed a net profit.The period ended with an operating loss of $175 million, but a net profit of $1.02 billion.In the client computing area, which is most dependent on PC market conditions, Intel's revenues fell 17% year over year to $8.1 billion, but grew sequentially by 6% due to an increase in average selling price.Intel expects PC sales to dip 15 to 19 percent this year, primarily in the consumer and education sectors, as well as inventory adjustments in the OEM segment, but year-over-year sales should remain above pre-pandemic levels for the foreseeable future.At the end of next year, PC sales should either remain at current levels or decline slightly.According to the Intel representatives, from 270 to 295 million computers will be sold next year.The head of the corporation even said that Intel was able to significantly strengthen its position in the PC segment in the third quarter.Image source: IntelIn the server segment last quarter, revenue at Intel dropped by an impressive 27% to $4.2 billion, but in the reporting documents the company proudly announced that it had started production of those models of processors Sapphire Rapids family, which were aimed at mass consumption, and also expects to slightly increase its market share in this segment by the end of the year.In the past quarter, Intel's server revenue was undermined by weak demand in the Chinese corporate sector.Image source: IntelThe immediate issue of the upcoming staff reductions at the quarterly event Intel management spoke reluctantly, only noting its inevitability.Intel's freelancers and contractors will be the first to suffer, they will feel Intel's desire to save money as early as this quarter.Marketing and advertising costs will be cut.Quantitative figures for staff reductions are not specified, but CFO Davis Zinsner let it slip that they will be significant.CEO Patrick Gelsinger added that production costs would be prioritized over payroll in the search for cost savings.By early October, the company had 131,500 employees.According to Intel's management, the company is on track to implement a plan to master five new process technologies in four years.There were no delays in learning Intel 4 and Intel 3 technologies, the first prototypes of chips made with the Intel 20A and Intel 18A technologies were already running in the company's labs, and this also applies to products for a large third-party customer - at least the digital design of the respective product is ready and the first working samples exist.In the customer segment, the company is going to introduce Meteor Lake processors (Intel 4) next year, at the same time Emerald Rapids servers will appear.On schedule, the company is preparing to launch server processors in 2024.The Meteor Lake series stepping design should be completed this quarter, and Intel's server segment managed to get the first Intel 3 Granite Rapids processor components last quarter, systems based on their samples are already working in Intel labs.For the first time, the company will use ultra-hard ultraviolet (EUV) lithography as part of its Intel 4 and Intel 3 processors.This will endow the respective products with breakthrough performance in terms of performance per watt of power consumption and density, according to Intel.Image source: IntelSome of Intel's business areas showed positive revenue growth in the third quarter.Among them were networking and edge computing solutions, which grew revenue 14% to $2.3 billion and kept operating income at $75 million, despite an 85% decline from the same period last year.
As it began building new facilities in Ohio, Intel made no secret that it would use them to serve outside customers in advanced 18A (18 angstroms - 18 tenths of a nanometer) technology, and some of the first customers connected with the U.S.defense sector will receive their orders.The company also plans to package and test the products in Arizona for those customers.Over the summer, the U.S.Defense Department complained that up to 98 percent of the microelectronics products it needs are manufactured or tested in Asia.Intel will try to solve this problem by creating a line at existing facilities in Arizona to test and package semiconductor components for the U.S.Defense Department in the foreseeable future.The Defense Department buys $3 billion worth of chips each year, Intel Foundry Service (IFS) President Randhir Thakur told EE Times.The company has already signed a $250 million contract with the Defense Department to design advanced semiconductor components with Intel.Image source: EE Times, Alan PattersonIntel intends to use the department's SHIP ecosystem, which allows for the integration of disparate computing components in one package.The U.S.Defense Department will be the first Intel customer to receive semiconductor chips made with 18A technology from the company in the second half of 2024.They will already be using a gate-around transistor (GAA) structure.IFS customers already include MediaTek, Amazon and Cisco, bringing in about $800 million in revenue in the first year of Intel's contract business.This is not so much against the background of $56.9 billion received by TSMC last year, but Intel is at the very beginning of this journey.Notably, for large customers like MediaTek, Intel is going to provide \"untouchable quotas\" that guarantee stable output for the customer even when demand is high.The purchase of Israel-based Tower Semiconductor will add to the IFS ecosystem with a network of facilities located in Israel, Europe, the United States and Japan, which last year together generated about $1.5 billion.These facilities use 500 to 65 nm processors, and especially valuable contributions to Intel business will come from these assets on the automotive and meth In the case of Intel's 18A technology, the toolkit has reached version 0.5, indicating that the research phase has passed, although the development of the process is still in full swing.According to Intel, the 18A process is comparable to the well-known 2nm process.To become a U.S.defense contractor, Intel must go through a complicated and lengthy process of security certification, but if it has such a certificate, it will automatically increase other customers' confidence in it.One of the confidence-building steps will be the establishment of a chip packaging and testing line in Arizona, since such operations with Intel products are currently carried out either in Malaysia, or in China or Vietnam.Intel now uses six different chip packaging companies, distributing orders to 18 companies.
Intel showed silicon wafers with mysterious 34-core Raptor Lake processors, but didn't say what they're for
At the Innovation 2022 event, Intel showed a silicon wafer on its booth that reveals 34-core Raptor Lake-S processors not officially introduced.As Paul Alcorn (Paul Alcorn) from Tom's Hardware writes, the company did not do it on purpose, because before that it has announced a consumer series of processors Raptor Lake, which have no more than 24 cores.Image source: Tom's HardwareJournalist points out that before that there were rumors that Intel is working on some 34-core processors.However, it turned out later that they were just MCC chipsets for Sapphire Rapids server processors.But the chips showed yesterday belong to the consumer Raptor Lake-S series.This is evidenced by the sticker available on the wafer, as shown in the photo below. The plate with the 34-core Raptor Lake chipsA closer look at the plate (photo above) makes it obvious that the uncut chips in it don't look like the standard Raptor Lake processor chips (photo below).The cores in the chips are not arranged horizontally in two rows like in standard Raptor Lake crystals.In addition, it seems that the 34 cores of the unknown Raptor Lake processors are connected by a kind of mesh structure, like in Ice Lake crystals (Xeon and 10th generation Core processors), as well as by a ring bus.Also, as part of the 34-core Raptor Lake crystals can distinguish eight DDR5 memory controllers and UPI blocks, which is unusual for standard desktop models.Crystal already presented desktop Raptor Lake- SThe staff of the stand, where they demonstrated the silicon wafers, also confirmed that these processors Raptor Lake-S with 34 cores, but did not specify whether they are server or consumer.The journalist notes that the size of crystals is much larger than the consumer model Raptor Lake-S for the LGA 1700 processor socket.It is very likely that Intel has shown crystals of some future powerful processors.Perhaps these crystals are for high performance desktop (HEDT) processors Raptor Lake-X, or on such crystals will be built processors for workstations Xeon W or other models.
Intel inaugurated construction of two plants in Ohio - they will start producing advanced chips by 2025
U.S.authorities' delay in approving a $52 billion package of legislation to subsidize the construction of semiconductor facilities in the country has delayed Intel's schedule for building two facilities in Ohio somewhat, but only on a ceremonial level.The president of the United States, the governor of the state and the head of Intel attended the ceremony this week.Image source: IntelAccording to a press release on Intel's website, other officials of various levels attended the grand opening ceremony for the Ohio site.While Intel's intentions to spend $20 billion to build facilities in Ohio have long been known, this week the company emphasized a related educational initiative.As clarified, it will allocate $50 million over ten years to core educational programs in Ohio that will cover more than 80 institutions in the state.Over the next three years alone, Intel will allocate $17.7 million to support research and educational programs, which will be distributed among eight academic institutions in Ohio.Over the next three years, this program will provide training for about 9,000 graduates, and 2,300 students will become recipients of targeted scholarships.About 7,000 people will be employed directly in the construction of plants in Ohio, and in the long term, they will provide jobs for about 3,000 skilled professionals.As Intel representatives once again stressed, the company's two new facilities in Ohio will focus not only on manufacturing products for its own needs, but will also serve the interests of third-party customers.Earlier it was explained, that it will be in Ohio, where advanced lithographic scanners ASML with high value of numerical aperture, which in the future will allow to produce products with Intel 18A technology.It will allow Intel to regain technological leadership in lithography by the middle of the decade.In addition to its own products, Intel is already considering making advanced products in Ohio for several customers.Most likely, the talk is about representatives of defense industry interested in getting American-made products that meet the latest criteria.Intel CEO Patrick Gelsinger thanked representatives from the U.S.Presidential Administration, Congress and state leadership for their assistance in the company's effort to \"restore the deserved U.S.position as a leader in the advanced chip manufacturing\".In the coming years, Intel will build two new facilities in Arizona, expand its production capacity in New Mexico, and improve its research center in Oregon, which will focus on developing advanced packaging techniques for semiconductor components and innovations in lithography.It will house an experimental production line that will allow advanced engineering ideas to be tested at prototype level.
As demand for electronics declines in the consumer sector, it is more common to hear of intentions to cut capital spending from memory manufacturers, which is only natural.Experts note that even caution, which is currently displayed by some market players will not prevent the semiconductor industry from ending the year with a record $185.5 billion in capex.Source image: IntelAccording to IC Insights, analysts expect the growth of CapEx in the industry to reach 21% by the end of 2022.This is less than last year's growth of 35%, and less than the forecasted value at the beginning of this year (24%), but the nuance is that for the third consecutive year, the capital expenditure of the semiconductor industry is growing consistently by double-digit percent.This was last seen from 1993 to 1995 inclusive.Image source: IC InsightsIc Insights analysts previously expected semiconductor capital spending to increase to $190.4 billion this year, but difficult macroeconomic conditions and signs of an overproduction crisis in some market segments suggest that growth will be limited to 21 percent and that spending will not exceed $185.5 billion.In any case, together with the last year market participants will spend an impressive $338.6 billion for the construction of new enterprises.At the same time, conveyors of vertically integrated producers of chips are still loaded more than 90%, and contract manufacturers load their capacities at 100%, still not being able to meet the entire demand.Most likely, as the authors of the forecast summarize, next year the industry will face a decrease in the amount of capital expenditures, as market participants will need to assess the situation after three years of active growth.
South Korean conglomerate SK Group said last month that it intends to allocate up to $22 billion over the coming years to U.S.science and industry, of which $15 billion will go directly to the U.S.semiconductor sector.Unofficially, SK hynix plans to build a chip packaging and testing facility in the country by 2025, with the location to be chosen by early 2023.Image source: SK hynixThis was reported by Reuters citing its own sources.According to reports, SK hynix is ready to spend several billion dollars to build the plant and create about 1,000 jobs, while choosing the site for the new company will be given preference to areas in the relative proximity of universities that can supply it with qualified personnel.The U.S.enterprise SK hynix will start mass testing and packaging of chips in the period from 2025 to 2026.Official representatives of the Korean company has no comment on these rumors, only referring to the desire of the parent structure SK Group to allocate $ 15 billion for the development of the semiconductor industry in the region.In addition to the enterprise for packaging and testing chips, the Korean giant is going to invest in creating a network of research centers in the United States.This type of activity will also allow the company to qualify for government subsidies as part of the recently passed package of laws.
According to some reports, in the generation of Galaxy S23 smartphones that will debut in early 2023, Samsung Electronics will completely abandon the use of internally developed Exynos processors in favor of Qualcomm 4nm components, which will be produced by TSMC.Only a year later, 3nm Samsung processors will have a chance to fit in smartphones of this brand.Image source: Samsung ElectricSuch as mentioned earlier, the \"first wave\" of 3nm chips produced by Samsung Electronics will be intended for Chinese developers of specialized gas pedals used in the mining of cryptocurrencies.It was their production in small quantities was established at the pilot line in the Korean Hwasong by the end of last quarter.For its own needs, as explained by Business Korea, Samsung Electronics will use a second-generation 3-nm process, which will be mastered only next year.The company will begin producing its own mobile processors using this processor by early 2024, at which point they will be able to fit into Samsung smartphones.It is possible that production of Samsung's 3-nm processors will be set up at facilities in Pyeongtaek because the production capacity at Hwaseong is limited in terms of scaling capacity.Rival TSMC will set up mass production of 3-nm chips only by the end of this year, but it is not certain that in its case the start will be as protracted, so it is premature to talk about a clear advantage of Samsung Electronics in terms of timing.
The decline in demand for chips in the second half of the year will hit mature processors the hardest
There are almost no market players left who can dispute the fact that demand for components for PCs and low-cost smartphones is declining.TrendForce experts expect this trend to be strongest in the mature lithography sector in the second half of the year, but even those using advanced manufacturing processes will be affected by the demand correction.Image source: TSMSAnalytics TrendForce operates equipment utilization rates - an indicator that shows the actual utilization of production lines relative to their design capacity.During the periods of peak demand for some types of components this value exceeded 100%, but now the market is going to normalize and even to work in the loading mode, which can not be considered optimal from the economic point of view.In the 200mm wafer and 110nm to 350nm process sectors at least, the second half of the year will see equipment utilization dip to 90-95%, with power electronics, image sensors and display drivers all affected.90-99% utilization will be typical of product lines producing a broader range of components in the 28nm to 90nm process, including those using 300mm silicon wafers.The general rule is as follows: the more products are tied to consumer electronics demand, the greater the risk of lower utilization rates in the second half of the year.Source image: TrendForce Chipmakers using 4nm to 19nm lithographic standards will fare relatively well, as their utilization rates will be in the optimal range of 95% to 100%.The longer production cycle, which smoothes out short-term fluctuations in demand, will also partly contribute to this.Migration to more advanced lithography will also help keep equipment utilization at a high level, if demand for 7nm products, for example, declines somewhat in the second half of the year, then 5nm and 4nm products will enjoy higher demand at least due to their novelty for consumers.It is important for every manufacturer to maintain optimal equipment utilization, as it helps reduce the unit cost of the products produced.Fixed costs, which are in any production, it is possible to extend to the maximum number of products.When the equipment is not utilized fully, the unit cost increases.
A report from Micron Technology has let industry analysts know that the company is ready to cut the amount of capital spending in the next fiscal year, which in the memory manufacturer's calendar will start in the first half of September.The exact amount of the reduction has not yet been named, as there are about two months left until the end of the fiscal year, but experts believe that Micron's example is ready to be followed by other semiconductor manufacturers.Image source: TSMSCA Analysts Citi, for example, expect the main market players to reduce the amount of spending on building new plants and production lines next year.If this year, according to their estimates, for the corresponding needs will be spent $95 billion, next year this amount can be reduced by 16% to $80 billion.On such measures can go not only memory manufacturers, but also contract manufacturers of chips, including the largest.Recall that only TSMC is ready to spend this year from $40 to $44 billion for construction of new plants and development of advanced lithography.Next year, according to preliminary estimates, the company's capital spending will exceed $40 billion, although so far its representatives are not ready to make clarifications on this parameter.The value of TSMC's core costs is proportional to its share of revenues in the global market of contract services.The rapid growth of costs, according to the Taiwanese media, will force the company to raise prices for its services by 6% by the beginning of next year.This unpopular step is pushed not only by capital costs, but also by an increase in electricity tariffs, which took place at the beginning of this month.
If U.S.government fails to provide subsidies, Intel will spend more on building facilities in Europe
Intel CEO Patrick Gelsinger's activism last month was aimed at encouraging U.S.lawmakers to pass a $52 billion package that would subsidize the national semiconductor industry.The company even said that it might focus on investments in Europe if it does not get support.It will spend 12 billion euros to expand facilities in Ireland, and a chip packaging and testing facility is set to appear in Italy, with 4.5 billion euros earmarked for the project.Research centers will appear and expand in France and Poland.However, in the case of European investments, Intel expects significant material support from the local authorities.Last Wednesday, as noted by CNBC, Patrick Gelsinger allowed himself to say that if U.S.lawmakers do not form the conditions for granting subsidies to the U.S.semiconductor industry, the company will have to focus more efforts in Europe.From the very beginning, as the head of Intel stressed, the company was counting on the support of the authorities in the construction of a new complex in the state of Ohio.It will either take a long time to build and reach a modest scale if the company has to finance it at its own expense, or reach a significant scale in a more modest time if the authorities provide subsidies.In the first case, the budget could be as low as $20 billion, while in the second it could be as high as $80 billion.According to Gelsinger, Intel canceled a groundbreaking ceremony for a manufacturing complex in Ohio scheduled for this summer because it wants to make sure it gets state subsidies first.Ohio Governor Mike DeWine has suggested that Gelsinger is using the delay in building the facilities in the U.S.to pressure lawmakers.On the other hand, the governor's own sources in both houses of the U.S.legislature make it clear that a $52 billion subsidy package will be passed soon.State officials are prepared to give Intel about $2 billion in incentives to build the manufacturing complex, of which $700 million will be used to build the engineering infrastructure.Intel representatives explained that Ohio attracted them not only by availability of developed base for qualified personnel training, but also by access to sufficient water resources.
U.S.Secretary of Commerce urged lawmakers to speed up $52 billion in subsidies for the semiconductor industry
President Joseph Biden already expressed concerns about the protracted consideration of the bill to allocate $52 billion in subsidies for the U.S.semiconductor industry.Now Commerce Secretary Gina Raimondo appealed to lawmakers to speed up work in this area - she thinks it's crucial to pass the law by the beginning of September.Source image: Intel \"Mark my words, if Labor Day passes and Congress does not approve the package of laws, these companies will not wait and will expand their presence in other countries,\" she thought it necessary to warn members of the American officials.This year, Labor Day in the U.S.is September 5, and in August, U.S.lawmakers are on vacation, so ideally the final version of the law should be approved in the summer.Apparently, in her speech, Gina Raimondo referred to companies like TSMC, Samsung, Intel and GlobalWafers, which confirmed the intention to build new semiconductor plants in the U.S.announced yesterday plans of Taiwan's GlobalWafers to build a silicon wafer fab in the U.S., which could cover all the needs of the local semiconductor industry by 2025, to some extent dependent on the willingness of U.S.authorities to subsidize the construction.At least, GlobalWafers head Doris Hsu shared exactly such considerations with Gina Raimondo, referring to the upcoming $5 billion cost of building the facility.According to the minister, the consequences of delaying the passage of this package of laws for national security would be most unfortunate.Intel CEO Patrick Gelsinger also visited Washington this month to consult with top officials.He said the company's project to build two $20 billion advanced facilities in Ohio will largely depend on subsidies from the authorities.If they do not, Intel will either have to stretch out construction over time or scale it back.
TSMC is the undisputed market leader in contract manufacturing services for semiconductor components, while Intel until recently successfully challenged their status as the largest supplier in terms of revenue, but then ceded the title to Samsung.In the second quarter, TSMC for the first time could surpass Intel in terms of revenue, based on preliminary calculations.Image source: TSMKak reports The Register with reference to analysts' forecasts, summarized Yahoo Finance, experts expect a consistent increase in revenue TSMC in the second quarter by 43% to $ 18.1 billion.At the same time earnings Intel should consistently fall by 2% to $ 17.98 billion, which will allow TSMC to bypass the processor giant on this criterion.This comparison is not the most usual, because most analyst agencies do not try to compare revenues Intel and TSMC, because the companies have worked in different market niches.The situation, however, is changing.It is known that Intel will not only use TSMC's services more actively to create its own components, but will also develop its contract business.This will allow Intel to become a competitor of TSMC.However, in the first quarter Intel's contract activity provided the company only $283 mln of income, while TSMC managed to bring in $16.7 bln, so the amounts differ by an order of magnitude.
Morgan Stanley analysts think that Intel underestimates the difficulties of contract business development
The speech of Intel CFO at the Bank of America conference has provoked the decrease in the company shares price, because David Zinsner had to admit that the market conditions at the moment were worse than the expectations of the company management.The pessimism was added by analysts at Morgan Stanley, who are not quite sure about the success of Intel's venture into the contract market.Image source: IntelThe processor giant intends to become a major contract manufacturer of semiconductor components in the next few years, offering customers no less advanced lithography than it will be used for its own needs.At any rate, the facilities under construction in Ohio will have mastered the production of chips using Intel 18A technology by 2025, and it will be offered to the company's customers at the same time.In fact, preliminary agreements have already been reached with some of the defense customers.Developing the contract business to a competitive level will require enormous material costs, and Morgan Stanley experts believe that Intel management underestimates the scale of difficulties that the company will have to face on the chosen path.All of this, of course, does not rule out the success of this business idea, but it may require a much larger investment than Intel currently expects.All these considerations have already led to the fact that the company's share price fell to a 52-week low ($39.18), and Morgan Stanley experts set the rate benchmark at $46 with a recommendation to sell Intel shares.
Automotive components supplier Denso is thinking about spinning off its chip business into a separate company
The importance of semiconductor components to the automotive industry could easily be seen in the magnitude of the crisis that has gripped production lines since late the year before last.The development of active driver assistance systems and the transition to electric propulsion will only increase automakers' demand for chips, so Denso is thinking about structurally separating its semiconductor business.Image source: DensoBy Denso CTO Yoshifumi Kato, this business brings the largest component supplier to the Toyota Motor assembly line up to $3.1 billion in revenue annually.The Japanese automaker's higher resistance to the semiconductor crisis has allowed Toyota Motor to break away from the rest of the market as its leader.Directly Denso is the second largest manufacturer of automotive components.In the segment of automotive semiconductor components, the company claims the fifth place in terms of revenues.Already now, Denso management is thinking about the feasibility of structural separation of the semiconductor business.If this happens, the chips of this brand will also be supplied for the needs of third-party manufacturers of automotive components.At the moment the company is concentrating its efforts on covering its own needs in semiconductor products.The development and production of such products will soon take up to 10% of Denso's research budget.The company expects revenues from chip sales to increase 19%.In February, Denso agreed with TSMC to buy a 10% stake in a joint venture with Sony, which will operate in Japan and manufacture semiconductor products.It was Denso's participation in the joint venture that prompted TSMC to consider using more advanced lithography, which initially was to be limited to Sony's required 22 and 28nm process technologies.Denso expects the semiconductor shortage to ease slightly at the end of the current fiscal year, which ends in March 2023.
Intel talks about the role of the contract business in its future development - it will make the company more flexible
David Zinsner, who represented Intel at the Bank of America conference, even though he is the CFO, answered questions related to the corporation's technology development strategy rather freely.At least he was able to explain the role Intel's own contracting business will play, as well as the principles on which the relationships with chip production contractors will be based.Image source: IntelIt is no secret that during the period of active investment in its own lithography technology Intel will rely on the production capabilities of TSMC.We know from unofficial sources, for example, that in the first half of this decade Intel will receive from the Taiwanese contractor ready-made 3-nm crystals, which will be integrated into its own products.Already by 2025, Intel intends to increase its competence in the field of lithography so much that it promises to regain the leadership in technological terms by the beginning of the second half of the decade.By that time, it expects to begin serial production of 18A technology products, which it will offer to third-party customers in sync with the release of its products.David Zinsner predictably began his reasoning with an analysis of the company's financial performance.The margin in the vertically integrated manufacturing capabilities (IDM) business has traditionally been close to 60%, whereas a good level for the contract business is considered to be a little over 50%.In this respect, as Intel's CFO admitted, combining both businesses threatens to \"dilute\" the company's profit margin relative to the original 60%.Besides, IDM and contract manufacturing, if they get along at Intel facilities, sometimes under one roof, will create a synergy in financial performance, as Zinsner assures.At least at the level of operating margins, progress will be made relative to the \"pure IDM\" option, where the company's production facilities are used only for its own needs.The Intel CFO ironically remarked that his boss Patrick Gelsinger clearly would not praise him for admitting that the PC market has no growth prospects.Representatives of the company believe that demand in this segment has reached a new base level, and now at least 300 million computers and laptops will be sold annually, and the market will never go down to 250 million.As noted today, within a year, downward fluctuations in demand cannot be ruled out.Commenting on the host's question about Intel's relationship with an \"Asian contract manufacturer\", which obviously meant Taiwan's TSMC, David Zinsner said: \"We want to keep some flexibility for one reason or another.Intel is expected to combine the development of its own production and the use of TSMC facilities.For example, Intel will not find it expedient to develop some processes on its own, or it will not be sure in the demand volume for the designed products.In this case, it is more profitable to outsource the production of chips to a third-party company, and then in case of a drop in demand Intel will have less losses than in the case of independent production.If the demand grows so much that Intel will not be able to meet it on its own, then again an outside contractor will come to the rescue.Present at the event, Intel executive vice president Sandra Rivera (Sandra Rivera) added that the company can continue to use the aging technology process for a very long time in the production of the same FPGA programmable matrix.They may be one or two generations behind basic lithography products, and it would not be feasible to keep their production on their own line in this case.In such a situation, production of programmable arrays can be entrusted to a contractor.In general, the approach to the packaging of computing solutions from heterogeneous chips opens up great opportunities for flexible management of production technology.Sandra Rivera also spoke about an example of unification of used technological processes.As you know, Intel was initially going to release Granite Rapids processors with Intel 4 technology, but Intel 3 technology process, from the point of view of migration, was not so complicated, promising a significant increase in transistor speed and density.Since Intel planned to use the Intel
Scientists have found a way to improve the cooling of processors and other chips by 150%, but it's not easy
Many people know the problem of cooling processors and other chips well.As the technological norms for manufacturing decrease, the problem becomes more acute as the energy density increases dramatically.Scientists have found that silicon can only increase the efficiency of heat dissipation by increasing its isotopic purity.Experiments have shown that the \"natural\" heat sink of chips can be increased up to 2.5 times.Image source: Junqiao Wu/Berkeley LabAs you know, all silicon in our nature consists of three stable forms of isotopes: about 92% is silicon-28 (28Si), 5% - silicon-29 (29Si) and 3% - silicon-30 (30Si).No one purifies silicon to make chips, although early studies showed that the silicon-28 isotope has the best thermal conductivity of all silicon isotopes, and impurities 29Si and 30Si impurities interrupt heat fluxes.It was believed that the gain barely exceeds 10% with significantly increasing costs of purification.Scientists from the Lawrence Berkeley National Laboratory.Lawrence Berkeley set up an experiment in which they tried to assess the degree of influence of silicon purification on the heat sink.They connected two thermocouples with nanowires 90 nm wide and measured the heat transfer.It turned out that pure silicon-28 improves heat transfer from the heating element to the other by 150%, not 10%, as previously thought.On closer examination it turned out that the surface of the nanowire was covered with silicon dioxide, which prevented parasitic heat dissipation across the surface of the wire, and the heat flow was concentrated in its center.Scientists do not give any recipes for the organization of the heat sink from the chips in a new way.But they showed the possibility of multiplying the efficiency of heat removal from the crystals, which will be worth the candle - the cost of producing isotope-pure silicon for future chips.
Western Digital revealed work on 162-layer BiCS6 flash memory and plans to create memory with more than 200 layers
Western Digital this week revealed its plans for 3D NAND memory in the coming years.The company said it is working with its Japanese partner Kioxia to develop smaller-area, 162-layer Flash memory chips and flash memory with more than 200 layers of capacity.Image source: Western DigitalThe next-generation BiCS6 3D NAND chip will arrive later this year.At first glance it will not be the most advanced chips by the number of vertical layers - only 162 layers, which against the background of just presented Micron 232-layer 3D NAND chips looks pretty modest.Nevertheless, the capacity of WD and Micron's solutions is the same - 128 GB (1 Tb).The area of BiCS6 is the most compact in the industry of 68 mm2.The company was able to achieve this by significantly reducing the physical size of memory cells, which helped to use a new material in their structure.BiCS6 memory will store four bits per cell (QLC).Reducing the physical volume of the cell with a combination of writing four bits to each cell should lead to a reduction in the number of overwrite cycles, but WD has not yet disclosed this value.At the same time, the speed of BiCS6 memory promises to be 60 percent faster than current solutions, allowing it to be used both for mass-market drives and high-capacity server SSDs.Increased density should also reduce the cost of production, which is important for everyone.Image source: Western Digital Another new development looks equally interesting - BiCS+ memory with more than 200 layers.It is claimed that it is being developed from scratch mainly for server SSDs.BiCS+ memory will appear by 2024 and will provide up to 55% increase in bits per wafer compared to BiCS6 memory, as well as speed increase of up to 60%.WD expects to see a 15 percent increase in write speeds, which is as important as anything else for NAND flash memory.Image source: Western DigitalWD is like other 3D NAND manufacturers, with the goal of having 500 or more layers of 3D NAND in the future.Produce such memory is only possible through a combination of multiple technologies, including vertical \"gluing\" of memory crystals.The company does not lose hope to release memory with writing five bits in each cell (PLC), but it is no less difficult than releasing 3D NAND with 500 layers, which puts off the time of its appearance.
AMD will unveil Radeon RX 7000 graphics cards on the latest RDNA 3 architecture graphics processors tonight. The start of the presentation,...