And then I have a follow-up.Jeff Su Okay. So Krish’s first question is on N3 capacity. Note we said they will be fully utilized this year. He wants to know the capacity that we build or plan for N3 this year and in the second half. How does that capacity amount compared to what we expected a year ago?C. C. Wei Well, I can answer the question. Actually, the demand is higher than we thought a year ago, and that’s why that we need to work very hard to meet customer’s demand. Did that answer your question, Krish?Krish Sankar Got it, got it. And then just a quick follow-up. You spoke about AI being a positive and all the innovation happening in generative AI today. Just from a TSMC standpoint, is this fair to assume that what you’re going through today is more on the training stage and therefore is more semiconductor and wafer intensive, but when you go into more inference, that intensity has to decrease?
Is that a fair assumption? Or do you think that this level of intensity will continue growing from a TSMC standpoint for AI?Jeff Su Okay. So Krish’s, second question is regards to generative AI and these type of applications. His observation is that the majority is training today, which seems to be beneficial. But as training moves to inference, would it be less semi-intensive or semi content intensive? And then, therefore, would that be, i.e., lower benefit to TSMC?C. C. Wei Krish, I mean, that today, your observation is right because, right now, most of the AI concentrate or focus on training. And in the future, it will be inference. But let me say that, no matter what kind of application, they need to use a very high-performance semiconductor component, and that actually is a TSMC’s advantage.
So we expect that semiconductor content starting from a data center for [indiscernible] to device and edge device or those kind of things, put all together, they need a very high-speed computing with a very power-efficient one. And so we expect it will add to TSMC’s business a lot. Qualitatively, as I said, we didn’t know yet. We hope that in the next few quarters, we can give a more clear picture.Jeff Su Okay. Thank you. Operator, in the interest of time, we will take the last 2 questions from the last 2 participants, please. So can we proceed to the next participant?Operator Next one is Brad Lin from Bank of America.Brad Lin I have two questions, one on the internal organization change and the other on the recently announced strategic alliance.
So while the global expansion is a key focus of TSMC, I noticed that TSMC set up a new unit called overseas operations office, OOO. What are the targets and impact that the TSMC management will look for from this new unit? That’s my first question.Jeff Su Okay. So Brad’s first question is about our internal organization change. We set up an overseas office we call OOO. So he wants to know what is the purpose of this organization or office. What is its targets and purpose?C. C. Wei Well, the purpose is very simple, because we need to have all the organization. Now the fab is overseas fab’s number and the amount would be more and more. So we need to have a coherence or — and the culture, everything aligned to the headquarter to TSMC’s core value.
So we established this overseas operation office to make sure that headquarter’s support to each overseas fab, what is sufficient and enough. And so that the performance will be aligned or matched with the TSMC’s fab in Taiwan. But more importantly, because of we have this organization, so we can help them to be succeed and, in the future, can be more profitable.Jeff Su Thank you, C.C. Brad, do you want to — yes, your second question.Brad Lin Sure. So my second question will be on the recently announced strategic alliance. As there no migration is the foundry’s competition focus, and we know that TSMC leadership in the leading edge. We recently see TSMC hang up with NVIDIA, Synopsys and ASML on 2-nanometer production and beyond. And what is the target?
And how is the progress so far? And TSMC is currently the only foundry within that group. Should we expect an even larger gap to peers with this? Or should we allow more companies or competitors to join this group?Jeff Su Brad, sorry. Let me clarify, make sure we understand your second question. So your second question is talking about, you call it, an alliance. But I think you referred to ASML and NVIDIA. So are you referring to this announcement recently of the — what is called the cuLitho?Brad Lin Exactly, yes, computational litho.Jeff Su Yes. Okay. So — okay. Got it. So Brad really wants to know what is management’s view towards this recently announced cuLitho. What is the implication? And what does this mean for TSMC’s competitiveness, sorry, going forward?C.
C. Wei Okay. That’s a good question. It is initiative invented by TSMC’s customer NVIDIA. And actually, we are working with them. And this, particularly the software and the hardware, together will help speed up computational lithography by moving expensive operation to GPU, which will help us deploy lithography solution like inverse lithography technology, deeper learning more broadly or et cetera. And because of this one, we got involved with our customer and our supplier. And we expect that this will give us some advantage over our cost improvement and also competition.Jeff Su Sure. Thank you, Brad. Operator, then can we move on to the last participant, please?Operator The last questions are from Charles Shi from Needham & Company.Charles Shi I have a question about your CapEx. Maybe I want to start with one item you disclosed in your filings.
On your balance sheet, you have an item called equipment under installation and construction in progress. That number has hit a record high, I think, over USD 40 billion as of the end of fourth quarter ’22, which kind of means there are $40 billion investment. You put money in, but you are not harvesting that investment yet. But there’s no revenue dollars being generated. But now I think you said you’re tightening the CapEx for 2023, but you’re reiterating that $32 billion to $36 billion CapEx. That seems to be adding on top of that $40 billion. A lot more investment is still expected in this year. So my question really is about this. Are you expecting a significantly higher incremental revenue opportunity in 2024 and beyond to justify that $40 billion plus, maybe another $32 billion investment?
