And we’ve been talking about this 2 times installed MIPS shift — MIPS capacity over the last program. That was actually the z15 program over z14 because we’re only 5 quarters into z16. But if you fast forward right now into z16, five quarters in, we’re 120% to 130% up on an installed MIPS capacity already. And you know the tail of that usually happens in the outer path of the next three or four quarters before we come out with the new one. So we feel pretty good, and that’s why we had the confidence entering this year, Arvind, and I, to take up the transaction processing model to actually low single-digit growth from being down mid-singles. So we see that continuing. Thanks for your question.
Patricia Murphy: Okay, we’re at the top of the hour. But let’s squeeze in one more question, please.
Operator: Thank you. Our final question comes from Ben Reitzes with Melius Research. You may go ahead.
Ben Reitzes: Hey, thanks a lot. Thanks for sneaking me in there. I wanted to ask you, Arvind, in layman’s terms with regard to AI, I find a lot of investors want to understand IBM’s place better. And in terms of AI, where you have your consulting and then hand it off to software, are you finding that IBM is well positioned to help people pick the right models to train, to lower their cost? What is it about IBM with the consulting and software that has your unique place set for growth and some of the optimism that you sound like you have heading into next year?
Arvind Krishna: Ben, thank you for the question, and I deeply appreciate your asking it in that form. Look, when we look at — let me start with our extreme advantage when clients have data that they like to train models on and they absolutely want to preserve the intellectual property in that data, they do not want the learnings, I mean, I can go through three forms. Almost everyone will say, look, we’re not going to take our data and give it to anybody in direct form. Okay. Next, how about learnings in the model itself from the data? Do you use that implicitly or explicitly or not at all? Three, the very fact that you’re asking a certain kind of question can in turn reveal information. I’ll use the word from my college training of site channel information.
I’m sure that all of you also understand what that word could imply. So for people who are worried about all that, which we estimate to be at least 30% of the overall AI opportunity. Think about a bank dealing with regulatory compliance. Think about a pharma company dealing with reports to the FDA. Think about a health care company worried about some, kind of, side effect on something. Think about a chemicals company worrying about their proprietary formulations, not just watson, the literature. So as you go through all of these, there are lots and lots of examples where people want to train models. Maybe it’s a refinement of a larger known model. So to your point on other models, and I’ll come to that. So we found that there’s a lot of opportunity, and those are the questions that people are asking us.
