Taiwan Semiconductor Manufacturing (TSM -4.30%)
Q2 2023 Earnings Name
Jul 20, 2023, 2:00 a.m. ET

Contents:

• Ready Remarks
• Questions and Solutions
• Name Members

Ready Remarks:

Jeff Su

[Foreign language] Good afternoon, everybody, and welcome to TSMC’s second quarter 2023 earnings convention name. That is Jeff Su, TSMC’s and your host for immediately. TSMC is internet hosting our earnings convention name by way of reside audio webcast via the corporate’s web site at www.tsmc.com, the place it’s also possible to obtain the earnings launch supplies. If you’re becoming a member of us via the convention name, your dial-in traces are in listen-only mode.

The format for immediately’s occasion shall be as follows: First, TSMC’s vice chairman and CFO, Mr. Wendell Huang, will summarize our operations within the second quarter 2023, adopted by our steering for the third quarter 2023. Afterwards, Mr. Huang; TSMC’s CEO, Dr.

C.C. Wei; and TSMC’s Chairman, Dr. Mark Liu, will collectively present the corporate’s key messages. Then we are going to open the road for a question-and-answer session.

As regular, I want to remind everyone that immediately’s discussions could comprise forward-looking statements which are topic to important dangers and uncertainties, which may trigger precise outcomes to vary materially from these contained within the forward-looking statements. Please discuss with the secure harbor discover that seems in our press launch. And now, I want to flip the decision over to TSMC’s CFO, Mr. Wendell Huang, for the abstract of operations and the present quarter steering.

Wendell Huang — Vice President and Chief Monetary Officer

Thanks, Jeff. Good afternoon, everybody, and thanks for becoming a member of us immediately. My presentation will begin with monetary highlights for the second quarter of 2023. After that, I’ll present the steering for the third quarter.

Second quarter income decreased 5.5% sequentially in NT or 6.2% in U.S. {dollars} as our second quarter enterprise was impacted by the general international financial situations, which dampened the top market demand and led to clients’ ongoing stock adjustment. Gross margin decreased 2.2 proportion factors sequentially to 54.1%, primarily reflecting decrease capability utilization and better electrical energy price, partially offset by extra stringent price management and a extra favorable overseas change price. Regardless of the trade’s cyclical downturn, we proceed to spend money on R&D to help our N3 and N2 improvement.

Thus, working margin was 42%, down 3.5 proportion factors sequentially. General, our first quarter EPS was TWD 7.01 and ROE was 23.2%. Now let’s transfer on to income by expertise. 5-nanometer course of expertise contributed 30% of our wafer income within the second quarter, whereas seven-nanometer accounted for 23%.

Superior applied sciences, outlined as seven-nanometer and under, accounted for 53% of wafer income. Shifting on to income contribution by platform. HPC decreased 5% quarter over quarter to account for 44% of our second quarter income. Smartphone decreased 9% to account for 33%.

IoT decreased 11% to account for 8%. Automotive elevated 3% to account for 8%, and DCE elevated 25% to account for 3%. Shifting on to the stability sheet. We ended the second quarter with money and marketable securities of TWD 1.5 trillion or USD 48 billion.

On the legal responsibility facet, present liabilities decreased by TWD 62 billion, primarily because of the web lower of TWD 87 billion in earnings tax payable as we pay TWD 120 billion for 2022 earnings tax, offset by TWD 33 billion accrued tax payables for the second quarter. Lengthy-term interest-bearing debt elevated by TWD 53 billion primarily as we raised TWD 41 billion in company bonds. On monetary ratios, accounts receivable turnover days decreased two days to 32 days, whereas days of stock elevated three days to 99 days, primarily because of entry ramp through the quarter. Concerning money circulation and capex, through the second quarter, we generated about TWD 167 billion in money from operations, spent TWD 251 billion in capex, distributed TWD 71 billion for third quarter 2022 money dividend and raised TWD 41 billion from company bond issuances.

General, our money stability decreased by TWD 109 billion to TWD 1.3 trillion on the finish of the quarter. Free money circulation was unfavourable TWD 83 billion through the quarter as working money circulation was greater than offset by capital expenditures, partly because of the earnings tax fee of TWD 120 billion. In U.S. greenback phrases, our second quarter capital expenditures totaled $8.17 billion.

I’ve completed my monetary abstract. Now let’s flip to our present quarter steering. Based mostly on the present enterprise outlook, we count on our third quarter income to be between USD 16.7 billion and USD 17.5 billion, which represents a 9.1% sequential improve on the midpoint. Based mostly on the change price assumption of USD 1 to TWD 30.8, gross margin is predicted to be between 51.5% and 53.5%, working margin to be between 38% and 40%.

This concludes my monetary presentation. Now let me flip to our key messages. I’ll begin by making some feedback on our second quarter ’23 and third quarter ’23 profitability. In comparison with first quarter, our second quarter gross margin decreased by 220 foundation factors sequentially to 54.1%, primarily because of a decrease capability utilization.

In comparison with our second quarter steering, our precise gross margin barely exceeded the excessive finish of the vary offered three months in the past, primarily because of extra stringent price management efforts and a barely extra favorable overseas change price. We’ve got simply guided our third quarter gross margin to say no by 1.6 proportion factors to 52.5% on the midpoint, primarily as the upper stage of capability utilization price is offset by 2 to three proportion factors margin dilution from the preliminary ramp-up of our three-nanometer expertise. Waiting for the fourth quarter, we count on the continued steep ramp-up of our three-nanometer to dilute our fourth quarter gross margin by about 3 to 4 proportion factors. In 2023, our gross margin faces challenges from decrease capability utilization because of semiconductor cyclicality, the ramp-up of N3, abroad fab growth, and inflationary prices, together with larger utility prices in Taiwan.

To handle our profitability in 2023, we are going to work diligently on inside price enchancment efforts whereas persevering with to promote our price. Whereas we face near-term challenges, we proceed to forecast a long-term gross margin of 53% and better is achievable. Subsequent, let me speak about our 2023 capital finances and depreciation. Yearly, our capex is spent in anticipation of the expansion that can comply with in future years.

Given the near-term uncertainties, we proceed to handle our enterprise prudently and tighten up our capital spending the place applicable. We now count on our 2023 capital finances to be towards the decrease finish of our vary of between USD 32 billion and USD 36 billion. Our depreciation expense is now anticipated to extend by mid-20s p.c yr over yr in 2023, primarily as we ramp our three-nanometer applied sciences. Regardless of near-term stock cycle, our dedication to help clients’ structural progress stays unchanged, and our disciplined capex and capability planning stays primarily based on the long-term market demand profile.

We’ll proceed to work intently with our clients to plan our long-term capability and spend money on modern specialty and superior packaging applied sciences to help their progress whereas delivering worthwhile progress to our shareholders. Now let me make a number of feedback on our money dividend distribution coverage. The goals of TSMC’s capital administration are to fund the capital — the corporate’s progress organically, generate good profitability, protect monetary flexibility, and distribute a sustainable and steadily growing money dividend to shareholders. On account of our rigorous capital administration, in Could, TSMC Board of Administrators accredited the distribution of TWD 3 per share money dividend for the primary quarter of 2023, up from TWD 2.75 beforehand.

This may develop into the brand new minimal quarterly dividend stage going ahead. First quarter ’23 money dividend shall be distributed in October 2023. For 2023, TSMC’s shareholders will obtain a complete of TWD 11.25 per share dividend and no less than TWD 12 per share money dividend for 2024. Going ahead, as our capital depth begins to say no within the subsequent a number of years, the main target of our money dividend coverage is predicted to shift from a sustainable to a steadily growing money dividend per share within the subsequent few years.

Now let me flip the microphone over to C.C.

C.C. Wei — Chief Govt Officer

Thanks, Wendell. Good afternoon, everybody. First, let me begin with our near-term demand and stock. We concluded our second quarter with income of USD 15.7 billion, in keeping with our steering in U.S.

greenback phrases. Our enterprise within the second quarter was impacted by the general international financial situations, which dampened the top market demand and clients’ ongoing stock adjustment. Shifting into third quarter 2023. Whereas we have now not too long ago noticed a rise in AI-related demand, it’s not sufficient to offset the general cyclicality of our enterprise.

We count on our enterprise within the third quarter to be supported by the robust ramp of our three-nanometer applied sciences, partially offset by clients’ continued stock adjustment. Within the final quarterly convention, we stated we count on fabless semiconductor stock to rebalance to a more healthy stage exiting the third quarter. This assertion continues to carry true. Nevertheless, because of persistent weaker general macroeconomic situations, slower than anticipated demand restoration in China, and general softer finish market demand situations, clients are extra cautious and intend to additional management their stock into 4Q ’23.

Thus, whereas we keep our forecast for the 2023 semiconductor market, excluding reminiscence, to say no mid-single digit yr over yr, we now count on the foundry trade to say no mid-teens and our full yr 2023 income to say no round 10% in U.S. greenback time period. With such stock management, we additionally forecast the fabless semiconductor stock to exit 4Q ’23 at a more healthy and decrease stage as in comparison with our expectation three months in the past. Subsequent, let me speak concerning the HPC and TSMC’s long-term progress outlook.

As we have now stated earlier than, the huge structural improve in demand for computation underpinned by the trade megatrend of 5G and HPC continues to drive better want for efficiency and energy-efficient computing, which require use of modern applied sciences. These megatrends are anticipated to fill TSMC’s long-term progress. Even with a tougher 2023, our income stays properly on monitor to develop between 15% and 20% CAGR over the following a number of years in U.S. greenback phrases, which is a goal we communicated again in January 2022 investor convention.

The current improve in AI-related demand is directionally constructive for TSMC. Generative AI requires larger computing energy and interconnected bandwidth, which drives growing semiconductor content material. With utilizing CPUs, GPUs, or AI accelerator and associated ASIC for AI and machine studying, the commonality is that it requires use of modern expertise and a robust foundry design ecosystem. These are all TSMC’s strengths.

Immediately, server AI processor demand, which we outline as CPUs, GPUs, and AI accelerators which are performing, coaching, and affect capabilities accounts for about 6% of TSMC’s complete income. We forecasted this to develop at near 50% CAGR within the subsequent 5 years and elevated low teenagers p.c of our income. The insatiable want for energy-efficient computation is ranging from information facilities, and we count on it to proliferate to edge and N units of time, which is able to drive additional long-term alternatives. We’ve got already embedded a sure assumption for AI demand into our long-term capex and progress forecast.

Our HPC platform is predicted to be the principle engine and the most important incremental contributor to TSMC’s long-term progress within the subsequent a number of years. Whereas the quantification of the whole addressable alternative remains to be ongoing, generative AI and enormous language mannequin solely reinforce the already robust conviction we have now within the structural megatrend to drive TSMC’s long-term progress, and we are going to intently monitor the event for additional potential upside. Now let me speak about our N3 and N3E standing. Our three-nanometer expertise is probably the most superior semiconductor expertise in each PPA and transistor expertise.

N3 is already concerned in manufacturing with good yield. We’re seeing sturdy demand for N3 and we count on a robust ramp of N within the second half of this yr, supported by each HPC and smartphone purposes. N3 is predicted to proceed to contribute mid-single-digit proportion of our complete wafer income in 2023. N3E additional extends our N3 household with the improved efficiency, energy, and yield, and offers full platform help for each HPC and smartphone purposes.

N3E has handed the qualification, and achieved efficiency and yield goal, and can begin quantity manufacturing within the fourth quarter of this yr. With our steady enhancement of three-nanometer course of applied sciences, we count on robust multi-yield demand from our clients and are assured that our three-nanometer household shall be one other giant and long-lasting node for TSMC. Lastly, I will speak about our N2 standing. Our N2 expertise improvement is progressing properly and on monitor for quantity manufacturing in 2025.

Our N2 will undertake slim sheet transistor construction to offer our buyer with the most effective efficiency, price, and expertise maturity. Our slim sheet expertise has demonstrated glorious energy effectivity and our N2 will ship full efficiency and energy advantages to deal with the growing want for energy-efficient computing. As a part of N2 expertise platform, we additionally developed N2 with bottom energy rail resolution, which is finest fitted to HPC purposes. Bottom energy rail will present 10% to 12% extra pace acquire and 10% to fifteen% giant density increase on high of the baseline expertise.

We’re focusing on bottom energy rail to be out there within the second half of 2025 to clients with manufacturing in 2026. We’re observing a excessive stage of buyer curiosity and engagement in N2 from each HPC and smartphone purposes. Our two-nanometer expertise would be the most superior semiconductor expertise within the trade in each density and vitality effectivity when it’s launched and to additional lengthen our expertise management properly into the long run. This concludes my ready remarks, and now let me flip the microphone over to Mark.

Mark Liu — Chairman

Thanks, C.C., and good afternoon, everybody. Immediately, I need to speak about TSMC’s international manufacturing footprint standing replace. TSMC’s mission is to be the trusted expertise and capability supplier of the worldwide logic IC trade for years to come back. Our technique is to broaden our international manufacturing footprint to extend buyer belief and to broaden our future progress potential and to achieve for extra international skills.

Our abroad choices are primarily based on our clients’ wants and the mandatory stage of presidency help. That’s to maximise the worth of our shareholders and to meet our fiduciary obligation. In Arizona, we’re constructing a primary fab to offer U.S. most superior semiconductor expertise in mass manufacturing to help the wants for U.S.

semiconductor infrastructure. Our fab in Arizona began development in April 2021 with an aggressive schedule. We are actually coming into a important section of dealing with and putting in probably the most superior and devoted tools. Nevertheless, we’re encountering sure challenges as there may be an inadequate quantity of expert employees with these specialised experience required for tools set up in a semiconductor-grade facility.

Whereas we’re engaged on to enhance the state of affairs, together with sending skilled technicians from Taiwan to coach the native expert employees for a brief time period, we count on the manufacturing schedule of N4 course of expertise to be pushed out to 2025. In Japan, we’re constructing a specialty expertise manufacturing unit, which is able to make the most of 12, 16 and 22, 28 course of applied sciences. Quantity manufacturing is on monitor for late 2024. In Europe, we’re partaking with clients and companions to guage constructing a specialty fab in Germany, specializing in automotive-specific applied sciences primarily based on the demand from our clients and the extent of presidency help.

In China, we’re increasing 28-nanometer in Nanjing as we deliberate to help our clients in China, and we proceed to comply with all guidelines and laws totally. On the similar time, we proceed to spend money on Taiwan and to broaden our capability to help our clients’ progress. From a value perspective, the preliminary price of abroad fab are larger than TSMC’s fabs in Taiwan because of: One, the smaller fab scale. Two, larger prices all through the availability chain.

And three, the early stage of semiconductor ecosystem on these abroad websites as in comparison with a matured ecosystem in Taiwan. In our current conferences with senior authorities officers within the U.S., Japan, and Europe, we mentioned our plans to broaden our international manufacturing footprint to them. We additionally emphasised certainly one of our main accountability is to handle and reduce the fee hole to maximise returns for our shareholders. These discussions went very properly.

All sides perceive the important and integral function TSMC performs within the semiconductor trade, and we respect all the federal government’s ongoing help in working with TSMC to assist slim down the fee hole. We’ll proceed to work intently with all of the governments to safe additional help. Our pricing may even stay strategic to replicate our price, which incorporates the worth of geographic flexibility. On the similar time, we are going to leverage our basic aggressive benefit of producing expertise management, giant quantity, and economies of scale to constantly drive our price down.

By taking such actions, TSMC may have the flexibility to soak up the upper price of abroad fab whereas remaining probably the most environment friendly and cost-effective producer, regardless of the place we function. Thus, whilst we broaden our capability abroad, TSMC’s long-term gross margin of 53% and better and sustainable ROE of better than 25% is achievable, and we are going to proceed to maximise the worth for our shareholders. This concludes our key messages. Thanks in your consideration.

Jeff Su

Thanks, Chairman. This concludes our ready statements. Earlier than we begin the Q&A session, I want to remind everyone to please restrict your questions to 2 at a time to permit all of the members a chance to ask their questions. Must you want to increase your query in Chinese language, I’ll translate it to English earlier than our administration solutions your query.

[Operator instructions] Now, let’s start the Q&A session. Operator, can we please proceed with the primary caller on the road?

Questions & Solutions:

Operator

The primary one to ask questions Gokul Hariharan from J.P. Morgan. Go forward, please.

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

Yeah, thanks. Good afternoon, and thanks for lots of readability on the AI-related publicity. My first query is on the AI entrance. A whole lot of TSMC’s clients have been speaking about capability scarcity and having to form of queue up for capability for AI accelerators, together with GPUs and ASICs.

Might TSMC speak a little bit bit about what TSMC is doing on the capability facet, particularly on the superior packaging, but additionally in different areas? And when do you count on to get again to a point of demand and provide stability for these AI accelerators? Is it going to be solely someday subsequent yr? Otherwise you suppose it may occur faster primarily based on what you see on demand out of your clients and the capability plan?

Jeff Su

OK, Gokul. Thanks. Let me attempt to — please enable me to summarize your first query. So, first query from Gokul is that he notes that we’re — clients are seeing robust demand from AI-related, however they’re going through capability tightness or scarcity.

So, his query is, what are we doing when it comes to the capability facet, perhaps each when it comes to the superior packaging, in addition to the logic? After which when can we see the demand-supply imbalance returning to a greater, more healthy, balanced stage? Is it someday subsequent yr? Is that appropriate, roughly, Gokul?

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

Yeah. Thanks, Jeff.

C.C. Wei — Chief Govt Officer

OK. Gokul, that is C.C. Wei. Let me reply your query.

For AI, proper now, we see a really robust demand, sure. For the front-end half, we have no drawback to help, however for the again finish, the superior packaging facet, particularly for the CoWoS, we do have some very tight capability to — very laborious to meet 100% of what clients wanted. So, we’re working with clients for the brief time period to assist them to meet the demand, however we’re growing our capability as shortly as attainable. And we count on these tightening shall be launched in subsequent yr, in all probability towards the top of subsequent yr.

However in between, we’re nonetheless working intently with our clients to help their progress.

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

OK. C.C., perhaps one follow-up. Might you tell us what sort of capability growth is like — how a lot capability you are increasing on the CoWoS facet? Like several form of what sort of capability you’re including?

Jeff Su

OK. So, Gokul, simply along with the primary query, how a lot capability are we going to extend when it comes to CoWoS?

C.C. Wei — Chief Govt Officer

Let me offer you — I can’t provide the precise quantity, however let me offer you a roughly in all probability 2x of the capability shall be added. OK, Gokul.

Jeff Su

Gokul, are you there? If not, operator, perhaps we transfer on to the following participant. Gokul, are you there? OK. I feel there was disconnected. All proper.

Let’s transfer on to the following caller, please.

Operator

Subsequent one to ask questions, Bruce Lu from Goldman Sachs. Go forward, please.

Bruce Lu — Goldman Sachs — Analyst

OK. Thanks for taking my query. I nonetheless need to learn about just like the TSMC maintained a 15% to twenty% income CAGR once we lower this yr’s income to minus 10%. That is all we use, that 15% income CAGR to 2026, that has been tried about like 25-plus p.c for the approaching two years, which implies that the general semi-growth goes to extend like lots for the following two years.

And also you simply talked about that the AI solely accounts for six% with low teenagers probably. That isn’t large enough to get again to the development. So, what’s the underlying progress you’ve got with the worldwide semi within the coming years? And what are the important thing assumptions for the expansion for every section?

Jeff Su

OK. Bruce, let me attempt to summarize your first query. So, Bruce’s first query is on our long-term progress CAGR, which we have now stated is to be between 15% to twenty% from ’21 to ’26 CAGR interval. So Bruce’s query, this yr, C.C.

simply stated we are going to decline round 10%. Underneath his calculation, I feel he is saying, properly, this suggests you must develop 25% the following a number of years, which, in fact, is a CAGR, however nonetheless and so Bruce’s query is that, due to this fact, if that is the kind of progress, then should not that suggest a a lot larger progress stage for the general semiconductor excluding reminiscence trade? I feel that’s your query, Bruce. Am I appropriate?

Bruce Lu — Goldman Sachs — Analyst

Sure. What are the important thing assumptions for the quarter?

Mark Liu — Chairman

OK. Let me deal with this query. Your rationale is appropriate. Nevertheless, a number of the components might not be completely included.

For one factor, in your mannequin that the purchasers’ gross margin, 60% or plus, I do not suppose that can symbolize the typical buyer’s gross margin, perhaps some particular ones. Nevertheless, the opposite one is the market share. The market share issue, you assume the fixed. That isn’t one factor that might be completely different than in your components.

So, the semiconductor progress proper now, we’re forecasting 4% to five%. It might improve, however positively, as you stated, it will not improve to 10%. However these longer-term semiconductor excluding reminiscence progress remains to be but to be evaluated. Did I reply your query?

Bruce Lu — Goldman Sachs — Analyst

Sure. The rationale I do that’s I am assuming that you’ve like dominant market share in the most effective node and in addition the expansion is generally coming from the superior node, which your clients is meant to be larger. So, I feel in that the hole is extensive sufficient. That is what I am questioning proper whether or not I miss something, which is perhaps large enough to maneuver the needle that some would possibly investments that the administration may give us some colour. 

Mark Liu — Chairman

This can be a issue. So far as the market share worth, you may not be completely included in all of the components. That is my perspective. However I can’t dig into the numerical comparability at this level.

What I imply, the market share is not only the superior modern applied sciences, but additionally the share of the outsourcing.

Jeff Su

So, perhaps, Bruce, if I summarize once more, TSMC’s progress is pushed by each the underlying structural megatrends but additionally by our expertise management and differentiation. So, our CAGR is a mix of these two components.

Bruce Lu — Goldman Sachs — Analyst

I see. OK, thanks. My subsequent query is relating to to the steering adjustments. The earlier steering for the complete yr was low to mid-single digit, now it’s about like 10% decline.

So, the hole is like 5% of the whole income, which is like fairly sizable when it comes to income, TWD 3 billion, TWD 4 billion extremely concentrated within the second half of fourth quarter. Are you able to give us like what are the adjustments when it comes to this shortfall? What are the weak spot of the contract?

Jeff Su

OK. So, Bruce’s second query is taking a look at our 2023 full yr steering. He notes that final time, we had stated low to mid-single-digit decline. This time, we have now guided to round 10%.

So, his query is the delta of this appears to be all — numerous it additionally within the fourth quarter. Is there any specific section or market that’s driving this? And what are the components behind this? Is that appropriate, Bruce?

Bruce Lu — Goldman Sachs — Analyst

Sure, thanks.

C.C. Wei — Chief Govt Officer

OK. Bruce, let me reply the query. Sure, we did see one thing completely different. The primary is weaker than what we thought.

Three months in the past, we had been in all probability extra optimistic, however now it isn’t. Additionally, for instance, China economic system’s restoration is definitely additionally weaker than we thought. And so, the top market demand truly didn’t develop as we anticipated. So put all collectively, even when we have now an excellent AI processor demand, it is nonetheless not sufficient to offset all these sorts of macro impression.

So now we count on that the entire yr will develop into minus 10%. That is what we thought.

Jeff Su

And when it comes to by specific section or is there a specific market?

C.C. Wei — Chief Govt Officer

Nicely, thanks. You might be asking me the query? It is nearly an impression.

Bruce Lu — Goldman Sachs — Analyst

Yeah, I perceive.

C.C. Wei — Chief Govt Officer

Sure, its general market section is being impacted as a result of it is a mixture of macroeconomics.

Bruce Lu — Goldman Sachs — Analyst

So, can we conclude that aside from AI, nearly each software sees some weak spot within the second half?

C.C. Wei — Chief Govt Officer

You bought it.

Bruce Lu — Goldman Sachs — Analyst

Thanks.

Jeff Su

Thanks, Bruce. Operator, can we transfer on to the following participant, please?

Operator

The following one to ask the final query is Gokul Hariharan from J.P. Morgan. 

Jeff Su

Gokul, you are again. OK. 

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

Yeah. Sorry about that. So subsequent query, I simply wished to ask about TSMC administration’s view on the present stock cycle. It appears like this cycle is way — taking it for much longer to get via the down cycle in comparison with ’19 and 2015.

When do you suppose we form of backside out and do you are feeling that the restoration in subsequent yr goes to be a robust restoration? Or do you suppose it may be a extra gradual restoration? What are the form of plans that you’re putting in as we take into consideration subsequent yr’s restoration as soon as stock state of affairs normalizes? Thanks.

Jeff Su

OK. So Gokul’s second query is concerning the stock correction cycle. He notes this cycle appears to be taking for much longer to get via as in comparison with 2019 or — and 2015. So his second query is when do we expect this cycle can backside out? What is going to 2024 subsequent yr appear like? Can we count on a robust restoration? And what components are we taking a look at? Is that appropriate, Gokul?

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

Sure, thanks. Thanks lots.

C.C. Wei — Chief Govt Officer

OK. Gokul, it is a good query. Let me reply it in a brief one sentence is all concerning the macro. I imply I simply say the macroeconomy isn’t so — has develop into weaker than we thought.

Actually, larger inflation and rate of interest impression finish demand in all market segments in each area on this planet. As we stated, beneath such state of affairs, our clients are extra cautious of their stock management within the second half of this yr. So, whereas we count on the fabless semiconductor trade, their stock to be cleaner and more healthy current this yr, however had been a lot nearer to the seasonal stage — however our expectation for them was to proceed to handle their stock. And 2024, it’s nonetheless depending on the macro state of affairs.

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

OK. So, it sounds such as you’re nonetheless anticipating no less than early a part of subsequent yr to nonetheless be a little bit bit difficult, much like what it’s trying like proper now. Is that truthful to say?

C.C. Wei — Chief Govt Officer

We provides you with our feedback — Gokul, we provides you with our feedback subsequent time in 2024.

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

Thanks.

Jeff Su

Thanks, Gokul. Operator, can we transfer on to the following participant, please?

Operator

Subsequent one, we have now Charlie Chan from Morgan Stanley. 

Charlie Chan — Morgan Stanley — Analyst

Good afternoon. Thanks for taking my query So, my first query is concerning the abroad fab price appears to get larger. So, would TSMC contemplate to additional modify your pricing to soak up these elevated prices? And in addition, administration talked about that you just’re doubling or greater than doubling your elevated packaging given the AI rush order. Would that offer you an opportunity to reprice the bake-in foundry service? As a result of I bear in mind there was a form of below-company gross margin common.

Would that be an opportunity to convey that again to the company common? Thanks.

Jeff Su

OK, Charlie. Charlie’s first query is, I assume, relating to pricing, two components or two angles. First on the abroad fab provided that the prices are larger, would TSMC contemplate to additional modify our wafer worth? And in addition, alongside comparable traces associated to superior packaging, given we’re — C.C. stated we’re doubling roughly the capability, would we contemplate to additionally cost extra or larger provided that the returns of the backend are decrease. 

Mark Liu — Chairman

Let me reply the primary query first. Sure, the abroad fab will price larger, no less than for the close to future, the place their provide ecosystem isn’t mature but. And the labor price is from our expertise, truly, is a little bit bit larger than we anticipated. However to reply your query, sure, as we attempt to get the utmost authorities subsidy, and we additionally actually have a look at the — how the value worth for the abroad geographical flexibilities is all thought-about.

The purpose is to, one, to extend our buyer belief, make them proceed to work with us going ahead beneath the geopolitical considerations. Secondly is to maximise shareholders’ worth. To reply your query on worth is strategically sure. C.C., are you able to reply the second query?

C.C. Wei — Chief Govt Officer

I feel the second query is concerning the pricing or the — and the cohorts. As I reply the query, we’re growing the capability in as quickly as attainable method. After all, that is together with precise price. So, the truth is, we’re working with our clients.

And crucial factor for them proper now could be provide assurance. It is a provide to fulfill their demand. So, we’re working with them. We do every thing attainable to extend the capability.

And naturally, on the similar time, we share our price. After which–

Charlie Chan — Morgan Stanley — Analyst

Nice. Thanks. Could add a second query? It is a completely different matter. Is that OK?

Jeff Su

Certain. You get two questions, so positive.

Charlie Chan — Morgan Stanley — Analyst

Thanks. Thanks, Jeff. So, one other query is concerning the AI semi demand, proper, because you’re offering your income contribution, progress assumption that’s tremendous useful. However I am questioning how TSMC can choose the AI demand as a result of proper now, it is arms race proper now.

Our clients are very aggressive reserving capability. So, I am questioning how t firm can choose whether or not these AI semi demand is for actual? And in addition, when it comes to breakdown, I am questioning whether or not firm sees the ASIC, the content material is outgrowing GPU. I feel the extra essential one must be the primary a part of the query, particularly traders are involved whether or not AI is cannibalizing the CPU server demand. So these are form of the questions in our thoughts.

Thanks.

Jeff Su

OK. Let me summarize your second query, Charlie. Charlie is on AI demand. He needs to understand how can we choose the demand correctly as a result of clients are very aggressive, however, in his phrases, how do we all know that this demand is actual? After which additionally how can we see the demand particularly for ASICs because it pertains to AI? 

C.C. Wei — Chief Govt Officer

OK. You need to reply?

Mark Liu — Chairman

Let me attempt first then you definitely in all probability comply with up. This can be a very deep query. After all, we have now a mannequin, principally. The short-term fancy concerning the AI demand positively can’t extrapolate for the long run.

And neither can we predict the close to future, that means subsequent yr, how the sudden demand will proceed or will flatten out. Nevertheless, our mannequin is predicated on the info heart construction. We assume a sure proportion of the info heart processor are AI processors and primarily based on that, we calculate the AI processor demand. And this mannequin is but to be fitted to the sensible information afterward.

However typically, I feel the development of a giant portion of knowledge heart processor shall be AI processor is a positive factor. And can it cannibalize the info heart processors? Within the brief time period, when the capex of the cloud service suppliers are mounted, sure, it is going to. It’s. However as for the long run, when their information service — when the cloud providers having the generative AI service income, I feel they are going to improve the capex.

That must be according to the long-term AI processor demand. And I imply the capex will improve due to the generative AI providers. Something extra for you? 

Jeff Su

Yeah, Charlie, I feel a part of Charlie’s query can also be how can we see ASIC associated in AI improvement? 

C.C. Wei — Chief Govt Officer

Nicely, truly, the client even have excessive demand on the ASIC half for the AI software. And as Mark identified, short-term sudden improve, you’ll be able to extrapolate to be long run. However once more, let me emphasize that. These form of purposes in AI be it CPUs, GPUs, or AI accelerator or ASIC, all of them want modern applied sciences.

And so they all have one symptom. They’re utilizing the very giant die measurement, which is TSMC’s energy.

Charlie Chan — Morgan Stanley — Analyst

OK. Thanks.

Jeff Su

OK. Thanks, Charlie. Thanks. Operator, we’ll transfer on to the following participant, please.

Operator

Subsequent one to ask questions, Randy Abrams, Credit score Suisse.

Randy Abrams — Credit score Suisse — Analyst

Sure. I wished to shift to the profitability, perhaps extra for Wendell. Trying on the fourth quarter, you talked about the three to 4-point dilution from N3. I feel that’s 2 to three factors within the third quarter.

Is that what you are suggesting the directional change might be a little bit bit down margin profile. Or do you’ve got constructive offsets that might hold it extra secure? After which a follow-up on the margin, the place you mentioned it is a robust yr for margins on these components just like the vitality, ramp up three. However may you talk about 2024, do you suppose we’re going right into a interval of a bit tougher profitability the place you see components that we may comfortably get again to the 53% and above subsequent yr?

Jeff Su

OK. Thanks, Randy. So, Randy’s first query is on gross margin. Fourth quarter with the N3 dilution of three% to 4%, does that imply directionally, fourth quarter margin is sequentially down? Are there any constructive offsets? After which for seeking to 2024 for the complete yr, if Wendell may give some feedback about 2024 gross margin, will it even be difficult? Or can we nonetheless really feel assured in a 53% and better gross margin?

Wendell Huang — Vice President and Chief Monetary Officer

OK, Randy. Ranging from the second half of this yr, as we stated, we confronted sure price challenges together with the ramp of N3, which is able to dilute about 2 to three proportion factors in third quarter and three% to 4% within the fourth quarter plus the upper electrical energy price. However we’re not giving our steering on the fourth quarter at this second. We’re simply spelling out a number of the challenges that we’re seeing.

And naturally, we’re going to proceed to drive down our prices and promote our price to make sure that we may have a superb return on the node. That is for this yr. For subsequent yr, we’re seeing — we’re not speaking about the entire gross margin, however we nonetheless see that N3 will dilute about 3 to 4 proportion factors of subsequent yr’s gross margin. And though the yield shall be higher subsequent yr.

On the similar time, the proportion of income contributed by N3 shall be larger. So, net-net, the place we additionally see some dilution from the N3 subsequent yr. However the margin — the steering shall be given out subsequent yr.

Randy Abrams — Credit score Suisse — Analyst

OK. I bought a fast follow-up to the primary query. I feel the previous few notes was 2 to three factors dilution within the first yr or two of ramp. The issue for it bigger, is it the upper capital depth or one thing completely different with three versus 5 and 7, it appears like a little bit bit extra dilution.

Wendell Huang — Vice President and Chief Monetary Officer

Yeah, the growing course of complexity does add on to the challenges of a more recent node. Nevertheless, the opposite essential issue is that our company averages has develop into larger than earlier than. We used to have 50% gross margin. We’re now speaking about 53% and better gross margin. 

Randy Abrams — Credit score Suisse — Analyst

OK. And the second query, I wished to ask the way you’re fascinated about capex, simply netting a number of issues, the geographic growth, the three, after which the beginning of two-nanometer, the primary ramp up or instrument move-in versus the combined outlook you are taking a look at for macro for a ballpark capex into subsequent yr? And if I may, perhaps inside it ask if the Arizona fab delays, does that push out the place you talked about the low finish of steering, push out a few of this yr that might give some raise for subsequent yr?

Jeff Su

OK. So, Randy’s second query is on capex. He needs to know principally specializing in 2024 capex, do a number of the delays within the Arizona fab push out capex from this yr to subsequent yr, as we broaden abroad as we spend money on N2, however on the similar time, because the macro stays unsure, how does this impression 2024 capex?

Wendell Huang — Vice President and Chief Monetary Officer

Yeah. Randy, the push out of fabs does push out some a part of the capex, however that does not have an effect on a giant half. For 2024, it is too early to speak concerning the general capex. Nevertheless, our capex, when you — as we stated earlier than, yearly, we spend the capex to seize the long run progress alternatives.

And up to now few years, our capex has risen very quick to seize the mega development. And going ahead, the following few years, once we begin to harvest these investments, we consider that capex will start to stage off when it comes to greenback quantity. And that can result in — begin to decrease the capital depth within the subsequent a number of years.

Jeff Su

Sorry, Randy, sorry, do —

Randy Abrams — Credit score Suisse — Analyst

My fast follow-up. I feel you talked about up to now, you could possibly use your five-nanometer to help the ramp of three. Given the AI and a few of that pickup, do you continue to see that potential that might assist optimize capex? Or do you want to hold it for current node? And that is my last one. Thanks.

Jeff Su

Yeah. So, Randy, simply can also be asking then how does instrument commonality play a job in our future capex?

Wendell Huang — Vice President and Chief Monetary Officer

Yeah. We at all times construct two [Inaudible] between nodes to offer a better flexibility. We talked about final time the robust multiyear demand from N3, we’re in a position to help that utilizing a number of the instruments from N5. We’re not going to touch upon the capex past this yr.

Nevertheless, as I simply talked about, yearly, the capex spend to seize future progress alternatives.

Randy Abrams — Credit score Suisse — Analyst

Thanks.

Jeff Su

OK. Thanks, Randy. Operator, can we transfer on to the following participant.

Operator

Subsequent one to ask questions is Laura Chen from Citi.

Laura Chen — Citi — Analyst

Thanks very a lot for taking my query. Good afternoon, gents. Very respect, C.C. and Mark sharing on TSMC’s view on the longer-term outlook in AI.

So, I am simply questioning how does TSMC consider your backend capability growth to sync up with the front-end wafer facet? Since there isn’t a drawback within the front-end foundry house, had been you form of involved about potential overcapacity within the backend facet past subsequent yr? Or truly, we might even see extra upside on the foundry wafer facet, so our, say, like superior node utilization price could go larger into subsequent yr? Thanks. That is my first query. 

Jeff Su

OK. So, Laura’s first query is taking a look at our growth of superior packaging or backend versus the frontend wafer, as we’re increasing the backend, however not the entrance finish, does that suggest that, first, that our frontend wafer notably main node, we count on the utilization to extend subsequent yr? After which conversely, or is there a threat that we’re over-expanding or overcapacity threat for the packaging facet?

C.C. Wei — Chief Govt Officer

OK, Laura. Let me reply the query. AI immediately is a very popular matter. A whole lot of my clients proper now elevated their demand and that can improve their front-end demand, in fact.

TSMC nearly have the key share or the most important share, let me say that, within the front-end wafer. In accordance with the entrance finish [Inaudible] we actually work intently with our buyer and to determine what’s the again finish that they want. And so, on that perspective, we’re planning our cohort as a capability, though in all probability nonetheless not sufficient, however we’re working very laborious to extend it. Overcapacity not immediately is a priority.

Immediately concern’s not sufficient capability to help all of the very robust demand. 

Jeff Su

OK. Thanks, C.C.

Laura Chen — Citi — Analyst

Yeah. Thanks. That is very clear. Thanks very a lot, C.C.

And my second query can also be concerning the gross margin outlook. If I am incorrect, please appropriate me. I recall that on the earlier cycle like seven or five-nanometer, the capability normally shall be thrice within the third yr of the brand new expertise ramping. So, I am questioning it’s nonetheless the case for N3.

Specifically, we have now seen that important capability depth improve could result in some margin strain, notably within the first few years. So I am simply questioning how does TSMC stability your expertise management and in addition the margin saturation? Thanks.

Jeff Su

Laura, you stated thrice — sorry, are you referring to the income contribution? Sorry, you stated N7 — 

Laura Chen — Citi — Analyst

No, the capability, the capability. Yeah. 

Jeff Su

Sure. OK. So, all proper, let me attempt to summarize your query. I feel Laura is asking N7 and N5, we considerably expanded the capability.

So, what’s the case for N3? After which additionally when it comes to the profitability of N3 or gross margin to be extra particular because it compares to N5 and N7 beforehand? Is that roughly appropriate, Laura? 

Laura Chen — Citi — Analyst

Sure. Thanks, Jeff.

Jeff Su

OK.

Wendell Huang — Vice President and Chief Monetary Officer

OK, Laura. Let me reply this query. As I simply talked about, the N3 because of the growing course of complexity is changing into tougher than the earlier nodes. However on the similar time, we are going to proceed to promote our price and drive down the fee on the similar time.

We nonetheless consider that N3 shall be a long-lasting — or a big node for TSMC. With all of the efforts and we nonetheless consider that the entire firm’s gross margin shall be 53% and better. OK?

Jeff Su

Laura, does that reply your query?

Laura Chen — Citi — Analyst

These solutions are very environment friendly.

Jeff Su

Sure. OK. Thanks, Laura.

Laura Chen — Citi — Analyst

Sure, thanks.

Jeff Su

Operator, let’s transfer on to the following participant, please.

Operator

Sure, proper now, we have now Rolf Bulk from New Road Analysis. Go forward, please.

Rolf Bulk — New Road Analysis — Analyst

Thanks for taking my query. This quarter, you are utilizing the legacy node 16 and 28-nanometer specifically had been down round 15% to twenty% Q-on-Q. And my query is, had been there any specific finish markets that prompted this decline? And the way do you concentrate on the restoration of these legacy nodes? Ought to we nonetheless count on restoration within the fourth quarter of this yr? Or is that extra 2024? Thanks.

Jeff Su

OK. So, Rolf’s first query is trying on the mature nodes, reminiscent of 16 and 28, he is aware of that — he notes, sorry, that these all noticed sequential declines within the second quarter. So, his query is, what’s driving this — what finish markets are driving this decline? And what’s the expectation for this within the second half?

C.C. Wei — Chief Govt Officer

Nicely, let me reply that query. The mature nodes wafer truly or the product truly attempt to be companionship for the smartphone or for the PC market or for the HPC. So, whereas the whole unit of smartphone develop into weaker and PC develop into weaker, so it is a excessive — the modern expertise node being — additionally demand dropping and so the mature node, that is collectively. Did I reply your query? 

Rolf Bulk — New Road Analysis — Analyst

Sure, thanks. That is very clear. For my second query, are you able to deal with cobalt and superior packaging typically and in addition the weak spot that you just see within the the rest of what you are promoting? Might you perhaps touch upon the proportion of your capex spending that can go towards main nodes, specialty nodes and packaging this yr in comparison with final yr?

Jeff Su

OK. So, Rolf’s second query is for 2023 capex, which our CFO has stated towards the decrease finish of the 32% to 36% vary, can we give a breakdown between modern specialty applied sciences after which the packaging, testing, mass-making, and others?

Wendell Huang — Vice President and Chief Monetary Officer

Modern expertise accounts for between 70% to 80% of our complete capex within the yr. Mature — specialty expertise, between 10% to twenty%, and the remaining are break up between superior packaging and EBO and a few others.

Jeff Su

OK, thanks, Rolf.

Rolf Bulk — New Road Analysis — Analyst

Excellent. Thanks.

Jeff Su

All proper. Operator, let’s transfer on to the following participant, please.

Operator

Subsequent one, we have now Sunny Lin from UBS. 

Sunny Lin — UBS — Analyst

Thanks very a lot. Good afternoon. Thanks for taking my questions. So, my primary query is on three-nanometer ramp-up.

As we’re going via set quarters of mass manufacturing, I feel it’s essential to have a reasonably good visibility for buyer engagement for the approaching few years. So, I ponder, now how ought to we take into consideration the general ramp of three-nanometer if we evaluate with 5 and seven-nanometer? If we have a look at 5, you reached towards 18% of income within the second yr of mass manufacturing after which about 24% of income within the third yr, whereas for three-nanometer, I feel the considerations by smartphone clients have been on price. Then the query shall be if HPC is important sufficient to nonetheless drive a significant pickup of three-nanometer, and so could be vastly appreciated when you may present us any form of ideas. Thanks very a lot.

Jeff Su

OK. So, Sunny’s first query is on the ramp of three-nanometer. Her query actually, I consider, is coming from a proportion of income contribution. She needs to understand how is the ramp-up three-nanometer? After which, can it contribute to the income like 5 and 7 up to now?

Wendell Huang — Vice President and Chief Monetary Officer

Yeah. As I simply stated, we consider N3 shall be a long-lasting and enormous node for TSMC. Now when it comes to proportion, I feel it is typically much less essential as a result of our general company income is way, a lot larger today than earlier than. So, I feel you must also take that into consideration, however greenback amount-wise, it is a a lot larger node, yeah.

Jeff Su

And C.C. additionally stated it is a multiyear robust structural demand. Yeah, sorry. OK. 

Sunny Lin — UBS — Analyst

Obtained it. Nicely, so I’ve a fast follow-up on three-nanometer profitability. And so, Wendell has offered a reasonably good perception concerning the dilution for 2024. However traditionally, a brand new node would take about seven to eight quarters to get to company common after mass manufacturing.

I perceive now company common gross margin can also be larger, however any expectations that N3 will develop into in keeping with company common gross margin? 

Jeff Su

Yeah. So, Sunny’s query is trying on the three-nanometer. Her query is basically that with three-nanometer and course of complexity. Sunny you are asking actually can it attain the company common over time?

Sunny Lin — UBS — Analyst

Or is there a timeline that you’re anticipating?

Jeff Su

And a timeline to achieve? 

Wendell Huang — Vice President and Chief Monetary Officer

Yeah. Sunny, as I simply talked about, it is changing into tougher for the main nodes due to the method complexity will increase lots. It applies to N3. So, will probably be difficult for N3.

We truly talked about that in the beginning of final yr already, will probably be difficult that for N3 to achieve the company common in seven to eight quarters time-frame like earlier than, yeah. However nevertheless, a part of it’s actually due to the upper company margin that we presently have. 

Sunny Lin — UBS — Analyst

Obtained it. Thanks. My second query is on two-nanometer. And so, if we have a look at your goal for two-nanometers enchancment over three-nanometer when it comes to pace and energy, the improve appears to be truly lower than three-nanometer over five-nanometer.

So, I ponder what’s truly the implication of GAH transition to price versus efficiency? Is the goal considerably conservative? Or there’s different technological challenges that we have to contemplate?

Jeff Su

OK. Sunny’s, second — properly, it is actually her third, however the query is on N2. She notes that the efficiency and the development appear to be lower than three-nanometer versus five-nanometer. So, may we speak extra about that?

C.C. Wei — Chief Govt Officer

Yeah, let me reply the query. Sunny, you’ve got an excellent commentary. Yeah, you’re proper. As I evaluate node to node from 5 to 3, the development, it turns into much less from three to 2.

However let me level it out, normally, we’re speaking concerning the efficiency, the pace and in addition the density, in order that the geometry shrinkage. Now, we deal with the facility consumption discount, which remains to be a full node efficiency as a result of as time goes by, increasingly more buyer, actually, they’re growing towards better energy effectivity. This is essential for the info heart, crucial for the server. And that is what we’re engaged on.

So, did I reply your query, Sunny? 

Sunny Lin — UBS — Analyst

Obtained it. Thanks for the colour and good to know that you’re on monitor to ship a second era of two-nanometer in 2026.

Jeff Su

Thanks, Sunny. Operator, let’s transfer on to the following participant, please.

Operator

Subsequent one to ask questions is Brett Simpson from Arete.

Brett Simpson — Arete Analysis — Analyst

Yeah, thanks very a lot. The primary query is for C.C. I used to be fascinated about getting a learn on the client reception you are getting for the brand new variants for N3, I feel you talked about N3P, N3X. Are clients nonetheless as targeted on N3? Or are you seeing a desire for them emigrate to the brand new variants reminiscent of N3P, N3X quite than N3E? And it is a follow-on.

For AI, when can we truly begin to see N3 adoption for N3? Thanks.

Jeff Su

OK. So, Brett’s first query is taking a look at our three-nanometer households and the continual enhancements that we at all times have, he’s asking, what’s the buyer reception of N3P and N3X? How does this evaluate or cannibalize N3? And when can we count on AI associated to undertake 3-nanometer household options? 

C.C. Wei — Chief Govt Officer

Let me reply the final one first. AI software already adopting that our N3 expertise node. We proceed to enhance our expertise as we at all times do. So, we have now an N3E, N3P and N3X.

X is the precise efficiency that is for very excessive pace, very excessive, let me say, efficiency computing for a number of the CPUs software. However N3Es are broadly accepted by all of my buyer and the design ranging from N3E and we assist them, OK, for a few of them go to the N3P. So, all collectively, each model, each variation, there’s numerous buyer engagement proper now. 

Brett Simpson — Arete Analysis — Analyst

OK. Thanks, C.C. And perhaps simply my second query for Mark. Mark, you had been speaking about increase the ecosystem in a number of the abroad markets just like the U.S., and also you’re speaking about ability scarcity.

However are you able to speak about what you suppose the like-for-like wafer price variations to function within the U.S. versus Taiwan? I feel your TSMC founder talked beforehand a couple of 50% premium to function within the U.S. Are you able to simply make clear if that is prone to be that prime? After which when would you count on the money help from the U.S. CHIPS Act to be made out there to TSMC? Thanks.

Jeff Su

OK. So, Brett’s second query is for Chairman. He needs to know, in, principally the fee hole, how large is the fee hole of fab within the U.S. versus in Taiwan.

Founder has stated 50% or extra. Is it that prime? After which concurrently, with the CHIPS Act, how and when can we count on to obtain the incentives to help? 

Mark Liu — Chairman

Sure, Brett Simpson, I feel the founder is true. I imply at this level, if we’re utilizing the present provide chain and labor price, certainly, as that variations. Nevertheless, we attempt to work with the U.S. administration.

Initially, on the subsidy — money subsidy and tax funding tax credit score, that’s to cowl the hole within the first 5 years roughly. When the instrument is depreciated, then the ecosystem turns into outstanding. That’s what’s the materials prices, chemical prices and the labor price. And we’re working with our provider to arrange a number of the extra environment friendly provide websites to be decrease, and the U.S.

administration has determined additionally to subsidize our suppliers. So that’s nonetheless within the works. How a lot it will probably additional lower? I do not know. However I feel both manner, we are going to strengthen our pricing values and have the ability to hold the company profitability as we forecasted now.

Jeff Su

Thanks, Chairman. OK. Thanks, Brett. Within the curiosity of time, operator, we’ll take questions from the final two members within the queue, please.

Operator

Sure. Subsequent one to ask query, Mehdi Hosseini from SIG. 

Mehdi Hosseini — Susquehanna Worldwide Group — Analyst

Sure, thanks for taking my query I am going to return to the gross margin. And I feel you highlighted the truth that for ’23, you are still monitoring to 53% gross margin on a USD foundation. That will suggest that This autumn might be flat to up. And I simply need to higher perceive how that is monitoring.

I am not asking for a information on This autumn, but when 2023 gross margin goes to be 53% plus, that might suggest This autumn flat to up. Is that appropriate?

Jeff Su

All proper, Mehdi. I feel we’ll let Wendell reply this query. However Mehdi is asking principally, are we saying that 2023 shall be 53% and better?

Wendell Huang — Vice President and Chief Monetary Officer

Mehdi, we’re not giving out steering past the third quarter. So, we’re not saying what Jeff simply stated. What we’re saying is barely a number of the unfavourable components will have an effect on the second half of the yr. As to 53% and better, that is a long-term progress margin targets for TSMC.

Jeff Su

Yeah, we didn’t present a steering for 2023 particularly, Mehdi. As Wendell simply stated, 53% or larger is our long-term goal, which we consider is achievable. Do you’ve got a second query?

Mehdi Hosseini — Susquehanna Worldwide Group — Analyst

OK. Thanks. Sure. Your up to date information suggests the revenues within the second half could be up 10% to 12% versus the primary half.

Clearly, the step-up is decrease than prior expectations. What I need to higher perceive is how ought to we take into consideration continued stock correction amongst your buyer versus new product ramp by a number of the different clients. Is there any manner you’ll be able to differentiate these two developments?

Jeff Su

OK. Mehdi is asking with our full yr steering. It implies a extra delicate second half seasonality. So, he needs to understand how a lot energy of the brand new buyer product launches is offset by continued stock correction, type of if we are able to present extra colour on that.

C.C. Wei — Chief Govt Officer

That is a tricky query to reply. Mehdi, your commentary is true. Second half, the seasonality is extra delicate than the earlier years. However, in fact, we have now N3 ramping up and for the brand new product launch.

However what’s the impression, easy methods to separate them, no, I can’t share an excessive amount of of the element of that.

Jeff Su

OK, Mehdi. Thanks. Operator, then let’s transfer on to the final participant, please.

Operator

Sure, the final one on queue is Charles Shi from Needham.

Charles Shi — Needham and Firm — Analyst

Thanks. Hey, thanks for squeezing me in. I’ve two questions. The primary query is I need to ask about AI, particularly round TSMC’s monetization of the AI development.

We did hear some commentary that for sure AI purposes, TSMC is promoting chips for a number of hundred bucks, however TSMC clients can truly promote for tens of 1000’s of {dollars} to their finish clients. So, I imply some traders I spoke with actually really feel that it pains them to see TSMC to create a complicated expertise, that deserve, in all probability, to offer a better worth than this. So, the query actually is, how does TSMC take into consideration perhaps higher monetization going ahead for the aptitude to provide all these AI chips? And actually, I need to tie again to 1 factor administration talked about within the ready remarks. The AI progress, 50% CAGR, how a lot of that’s quantity? And the way a lot of that might be the pricing tier when it comes to TSMC’s anticipated progress over the following few years in AI? Thanks.

Jeff Su

OK. Charles’ query. First is on AI. Once more, principally, he is asking about monetization or capturing worth, as an example.

He notes that TSMC, we could also be promoting chips for a number of hundred {dollars}, however our clients are in a position to promote it for tens of 1000’s or much more. So, is TSMC freely giving an excessive amount of of the worth? Can we higher promote our price or monetize to seize better worth with the AI development?

C.C. Wei — Chief Govt Officer

Nicely, Charles, I used to make a joke on my clients saying that I am promoting him a number of hundred {dollars} per chip, after which he offered it again to me with USD 200,000. However let me say that we’re completely satisfied to see clients doing very properly. And if clients do properly, TSMC does properly. And naturally, we work with them and we share our price to them.

And basically, we need to say that we’re in a position to deal with and seize a serious portion of the market when it comes to a semiconductor part in AI. Did I reply your query? 

Charles Shi — Needham and Firm — Analyst

Sure. What concerning the half about 50% CAGR, how a lot of that’s quantity? And will we count on some pricing aspect in that long-term progress?

Jeff Su

Sorry. Yeah. So, the second half is about near 50% CAGR for AI, sorry, server with AI processor. How a lot of that’s quantity? How a lot of that’s worth?

C.C. Wei — Chief Govt Officer

We can’t separate out, however let me share with you once more, we speak to the purchasers as a result of we have now a serious share of all of the modern expertise node. So, we all know that we are able to make our judgment. And so, we forecast of fifty% CAGR. How a lot of that’s on the frontend, backend, or others? I’m not in a position to share with you about it.

However let me guarantee you that TSMC goes to seize a serious portion of the market when it comes to semiconductor part.

Jeff Su

OK, Charles?

Charles Shi — Needham and Firm — Analyst

Thanks. Can I ask you the second query?

Jeff Su

Final query.

Charles Shi — Needham and Firm — Analyst

Sure, no drawback. The final one. After I requested concerning the capex, I feel I heard a remark perhaps from Wendell concerning the greenback quantity capex might be going to stage off the frontier. And I feel the administration used to level us, analysts, to have a look at the 2010 to 2013 interval with a excessive capital depth.

Capex truly went up $3 billion to $4 billion per yr to $10 billion. And truly, after that, TSMC capex did stage off round $10 billion stage for roughly 5 years till 2019. By telling us the capex might be going to stage off, are you telling us or are you alluding to perhaps there shall be some steady-state capex numbers going ahead, perhaps ranging from ’24 or ’25, round $30 billion-ish stage? That is only a clarification on that remark to stage it off. Thanks. 

Jeff Su

OK. So, Charles’ second query is on our capex. He needs to know capital spending beginning to stage off. I feel Wendell stated within the subsequent a number of years.

So not any particular, however what does that imply? Is it going to remain across the $30-some billion stage? Or what does that imply by spending leveling off?

Wendell Huang — Vice President and Chief Monetary Officer

Yeah. Charles, as I’ve talked about up to now few years, our capex elevated dramatically from $10 billion to $36 billion final yr. As we begin to harvest these investments, the rise in capex shall be slower than earlier than. That is what I imply by leveling off. 

Jeff Su

OK. All proper, Charles?

Charles Shi — Needham and Firm — Analyst

Obtained it. Obtained it. [Inaudible]

Jeff Su

OK. Thanks.

Charles Shi — Needham and Firm — Analyst

Sure, thanks.

Jeff Su

OK. Operator, this concludes our Q&A session. Earlier than we conclude immediately’s convention, please be suggested that the replay of the decision shall be accessible inside half-hour from now. The transcript will develop into out there 24 hours from now, each of which you’ll be — discover and out there via TSMC’s web site at www.tsmc.com.

So, thanks for becoming a member of us immediately. We hope everybody continues to remain properly. Have a superb remainder of the summer season, and we hope you’ll be part of us once more subsequent quarter. Goodbye, and have a superb day.

Period: 0 minutes

Name members:

Jeff Su

Wendell Huang — Vice President and Chief Monetary Officer

C.C. Wei — Chief Govt Officer

Mark Liu — Chairman

Gokul Hariharan — JPMorgan Chase and Firm — Analyst

Bruce Lu — Goldman Sachs — Analyst

Charlie Chan — Morgan Stanley — Analyst

Randy Abrams — Credit score Suisse — Analyst

Laura Chen — Citi — Analyst

Rolf Bulk — New Road Analysis — Analyst

Sunny Lin — UBS — Analyst

Brett Simpson — Arete Analysis — Analyst

Mehdi Hosseini — Susquehanna Worldwide Group — Analyst

Charles Shi — Needham and Firm — Analyst

Extra TSM evaluation

All earnings name transcripts