Microchip Technology Incorporated (MCHP) Earnings Call Transcript & Summary

Good evening, and welcome to the Microchip's Q2 Fiscal Year '27 Financial Results Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. I'll now turn the conference over to Eric Bjornholt, Senior Vice President and CFO. Thank you, Eric. You may begin.

All right. Thank you, and good afternoon, everyone. During the course of this conference call, we will be making projections and other forward-looking statements regarding future events or the future financial performance of the company. We wish to caution you that such statements are predictions and that actual events or results may differ materially. We refer you to our press releases of today as well as our recent filings with the SEC that identify important risk factors that may impact Microchip's business and results of operations. In attendance with me today are Steve Sanghi, Microchip's President and CEO; and Sajid Daudi, Microchip's Head of Investor Relations. I will comment on our second quarter fiscal year 2027 financial performance. and then Steve will provide commentary on our results and discuss the current business environment as well as our guidance. We will then be available to respond to specific investor and analyst questions. We are including information in our press release and on this conference call on various GAAP and non-GAAP measures. We have posted a full GAAP to non-GAAP reconciliation on the Investor Relations page of our website at www.microchip.com, and included reconciliation information in our earnings press release, which we believe you will find useful when comparing our GAAP and non-GAAP results. We have also posted a summary of our outstanding debt and leverage metrics on our website. I will now go through some of the operating results, including net sales, gross margin and operating expenses. Other than net sales, I will be referring to these results on a non-GAAP basis, which is based on expenses prior to the effect of our acquisition activities share-based compensation and certain other adjustments as described in our earnings press release and in the reconciliations on our website. Our non-GAAP financial results met or exceeded the guidance we provided on August 6, 2026. Net sales in the September quarter were $1.622 billion, which was up 9.2% sequentially and up 42.2% from the September 2025 quarter. We had several infrequent items that positively impacted net sales by about $53 million in the September quarter that we do not expect to repeat in the December quarter. These items were all at 100% gross margin and included approximately $21 million impact of our pricing change on inventory in the distribution channel. $22 million in licensing revenue from patent licensing transactions in excess of what we expect in the December quarter and a $10 million benefit from a long-term contract with the customer. We have posted a summary of our net sales by product line and geography on our website for your reference. On a non-GAAP basis, gross margins were 66.5% and including capacity underutilization charges of $26.6 million. Non-GAAP gross margins were positively impacted by higher licensing revenue in the quarter as well as the impact of our pricing change that was implemented in the September quarter. Non-GAAP operating expenses were at 27% of sales and operating income was 39.5% of sales. Non-GAAP net income was $559.2 million and non-GAAP earnings per diluted share were $0.97, which was $0.04 above the midpoint of our guidance. On a GAAP basis in the September quarter, gross margins were 65.9%, total operating expenses were $603.5 million and included acquisition intangible amortization of $90.2 million, special charges of $7.9 million, which was primarily driven by our activities associated with our closure of Fab 2 and acquisition-related expenses. Share-based compensation of $62.2 million and $5.3 million of other expenses. The GAAP net income attributable to common shareholders was $373.6 million or $0.68 per share and included a tax benefit of $0.13 per share related to a valuation allowance release on tax attributes. Our non-GAAP cash tax rate was 5.7% in the September quarter, and we expect to have a tax rate of about 6.5% for fiscal year 2027. Our tax rate is down from our previous forecast for a variety of reasons, including the impacts of a reduction in inventory reserves and a higher percentage of our taxable income being in jurisdictions with lower tax rates. Our inventory balance at September 30, 2026, was $1.117 billion. We had 184 days of inventory at the end of the September quarter, which was up 9 days from the end of the June quarter. Inventory increased in the quarter as we received some high dollar value wafers from our foundry partners in anticipation of future growth and build some finished goods inventory. Included in our September quarter ending inventory was 12 days of long life cycle, high-margin products whose manufacturing capacity has been end-of-life by our supply chain partners. Inventory at our distributors in the September quarter was at 25 days, which was flat compared to the June quarter and at the lower end of what we have experienced historically. Our cash flow from operating activities was $419.2 million in the September quarter. Our adjusted free cash flow was $377.1 million in the September quarter. As of September 30, our consolidated cash and total investment position was $457 million. Our net debt decreased by $31.7 million during the September quarter, and this is after funding the acquisition of Hailo. Our adjusted EBITDA in the September quarter was $705.4 million and 43.5% of net sales. Our trailing 12-month adjusted EBITDA was $2.162 billion, our net debt to adjusted EBITDA was 2.36% at September 30, 2026. We expect our net debt to adjusted EBITDA to continue to reduce as we progress through fiscal year 2027. Capital expenditures were $23.1 million in the September quarter, and we expect capital expenditures for fiscal year 2027 to be about $100 million. Depreciation expense in the September quarter was $36 million. I will now turn it over to Steve, who will provide some additional commentary on our September quarter results and our guidance for the December quarter. Steve?

Thank you, Eric, and good afternoon, everyone. I will reflect a bit on our performance in the September quarter. The September quarter was an excellent quarter for Microchip with broad-based growth as we beat the high end of our guidance in net sales and our non-GAAP financial metrics like gross margin percentage, operating expense percentage, operating profit percentage and earnings per share all met or beat our guidance. Our industrial and data center end markets were the 2 biggest drivers of our growth across our end markets, new design wins are increasingly supporting that growth. Our end market split for the September quarter as a percentage of net sales, excluding licensing revenue, was industrial at 34.6% and data center at 16.7%, aerospace and defense at 16.1%, Automotive at 14.2% and communication at 8.2%; consumer appliances at 7.3% and compute at 2.9%. From the September quarter 2025 to the September quarter 2026, net sales from industrial grew 50.3% and Data center grew 55.7%. Aerospace and Defense grew 41%, automotive grew 19.3% and Communication grew 43.5%, Consumer appliances grew 26.7% and compute declined 1.6%. Net sales from our data center business unit was over $250 million in the September quarter and is well on its way to be $1 billion in calendar year 2026. For calendar 2026, about half of our data center business comes from our standard products and about half comes from our Data Center Solutions business unit. On the standard product side, design wins are building across many products. In power, our digital power controllers are winning across AI, data center, power applications, from server power supplies to battery backup units. In system management, we offer the parts that sense, control and connect a server or rack and our gigabit Ethernet switches and USB hubs are gaining design win momentum in management networks. With respect to our Gen6 PCIe switch and retimer products, we had 19 design wins at the end of September quarter, up from 12 design wins at the end of June quarter. As of today, we now have 24 total design wins. These incremental wins will begin to ramp net sales next year and help position us across a growing range of next-generation data center platforms. As these deployments scale so does our opportunity across the broader data center portfolio, including timing, security and networking. I would like to give one more organizational update. As you know, after my return to Microchip as CEO 2 years ago, I reorganized the company into 5 pillars. These pillars were microcontrollers, analog and power, networking and connectivity, high-performance compute and AI on the edge. There is a Senior Vice President heading up each of these pillars, our Data Center Solutions business unit reports into the networking and connectivity pillar. Because of our focus on our Data Center Solutions business unit called the DCS, I have decided to pull up the DCS business unit to report directly to me at the executive staff level. The head of that business unit is Brian McCarson, who has spoken at these earnings calls before and will now report directly to me. Our design wins are not limited to data centers in aerospace and defense, we're winning designs across regions, products and applications from satellite communications and laser terminals to unmanned systems by adapting our commercial technologies for Space and Defense. In one recent case, a defense technology company selected our radiation-tolerant secured FPGA over its incumbent supplier for 2 space avionics programs. As defense production rates ramp so does our opportunity. Security requirements are also growing across our markets from data centers to defense to the edge. At the edge, a leading enterprise network provider, recently selected our security solution to protect its wireless networks with quantum resistant encryption became a requirement. Together, data center, aerospace and defense, industrial, robotics, automotive and the intelligent edge are the secular trends we see driving growth over the long-term. I will now talk about our business environment and the guidance going forward. Our bookings were very strong in the September quarter, and our book-to-bill ratio remained well above 1. We started the December quarter with a higher backlog compared to when we entered the September quarter. The September quarter saw a pivot in the business environment with customers completing their inventory digestion to shortages appearing in certain areas because of lengthening lead times at foundries, set suppliers and in our own fabs. Lead times for the majority of our standard products historically have been 4 to 8 weeks. But now with our die inventory substantially reduced and foundry and OSAT capacity constraints, we are seeing lead times broadly stretch out on many of our products. Our customer requests for expedited shipments have increased significantly from a few quarters ago. We are receiving many orders for requests inside the lead time and they're going unsupported in the quarter. We are advising our customers to give us longer-term backlog to help better match our new lead times which ready a lot across our product lines. Our new lead times are anywhere from 6 weeks to 32 weeks. Eric just explained the 3 infrequent items in the last quarter. First, it was an unusually strong quarter in our licensing group with a couple of large deals closing in the September quarter. Our licensing business is usually lumpy. The second item was the onetime effect of our recent price increases, which raised the value of our distribution inventory. The third was a revenue recognition from a long-term customer contract -- those infrequent items are not present in the December quarter, our December quarter is also our weakest quarter seasonally because there are a lot of holidays in the quarter. Therefore, the guidance for December quarter should be viewed in that context. Our December quarter also includes the impact of 2 small acquisitions that we completed. Hailo in the AI on the Edge space in Vorago in the A&D space, we expect that Hailo will add about $4 million to our sales in the December quarter and Vodago will add about $6 million to our net sales in the December quarter. Taking all of these factors into account, we expect our net sales in the December quarter to be between down 3% and up 1% with the midpoint of minus 1%. At the midpoint, this would be up 35.4% from the year ago quarter. December quarter is usually seasonally down 3% to 5%. This seasonal effect, combined with the infrequent items we experienced in the September quarter puts our December quarter guidance about 5% to 7% above normal seasonality. We expect our non-GAAP gross margin to be between 64% to 65% of net sales. This is because of several reasons. First, the 3 infrequent items in the September quarter were all at 100% gross margin. Secondly, our net sales guidance is lower than what we would have expected a quarter ago. Thirdly, we have substantial unsupported backlog on high-margin products that our supply chain cannot fulfill. And finally, we are running lower level of utilization internally than what we were planning a quarter ago. All these factors are impacting our gross margin guidance. We expect our non-GAAP operating expenses to be between 26.5% and 27% of sales. We expect our non-GAAP operating profit to be between 37% and 38.5% of sales. We expect our non-GAAP diluted earnings per share to be between $0.87 and $0.95 per share. At the midpoint, non-GAAP earnings per share would be up 107% over a year ago quarter. Regarding our long-term model, we maintain our long-term model on a non-GAAP basis to be 65% gross margin, 25% operating expense and 40% operating profit. With that, operator, will you please poll for questions?

[Operator Instructions] Our first question comes from the line of Timothy Arcuri with UBS.

Steve, I was wondering if you could help unpack a lot of these puts and takes. It seems like you said that December revenue was a little lower than what you would have thought 3 months ago and Fab utilization is also a little lower, yet bookings are stronger. Is this due to constraints that are downstream from you? And maybe -- does that give you more visibility into March. So you could maybe give us a sense, maybe normal seasonal, I think in March is up like 3%. So it sounds like maybe you feel a little bit better about March as a result of all this. If you can just unpack that for us, that would be great.

Yes. So I think if you look at the last quarter, it was kind of noisy. The 3 items that we talked about were in our guidance, the licensing deals were in our guidance, the recognition of $10 million of customer contract was in our guidance and so was the price increase. So the upside last quarter came from product sales, so we did very well. Now when you look at the current quarter, lead times have stretched out and a lot of the high-margin bookings we received on a lot of advanced products in pretty much every business segment cannot be fulfilled in this quarter. So they -- some of them are scheduled for next quarter. So as a result, I think we're guiding the December quarter to be -- it had lower than we would have thought about a quarter ago. The backlog for March quarter today is much better than December quarter was at the same point in time. So currently, we're optimistic. We're not giving any guidance for the March quarter. But I think some of the constraints we are talking about are now scheduled into March quarter. So we are hopeful that March quarter will be a stronger quarter.

And then maybe just as a quick follow-up. So on the PCIe Gen6 wins, I think, Steve, you said 14 design wins coming out of June and it's now up to sounds like you're building some high-value inventory for that stuff, too. So when does that start to become really big revenue? And if you can kind of size what the incremental could be from those wins, if possible, that would be great.

So we -- we're unable to dollarize the wins. The part went into production last quarter and we have a small amount of shipments in the current quarter, and they will increase every quarter from here on. Actually, there is a sizable shipment this quarter and they start increasing then from the next quarter on, the design wins are very strong and they're turning on at various times next year. So by the start of the 2028 you can expect almost all 24 designs to be in production, and we're winning essentially design a recall.

Yes. Just one clarification to the way Tim asked that question as he said PCIe Gen6 and the 24 design wins is both PCIe and the retimers. So I just wanted to clarify that.

That is correct. I think we are basically keeping them together. They're very much in the same application, many times, retimer goes with the PCIe, we don't always win the both, in some cases, customer doesn't need the retimer, they only need the PCI switch, but they're pretty much products are in sync. In the past, we have broken out how many on PCI Gen6 and how many in retimers. And I don't have that today, but we have no problem in breaking that out.

Okay, Steve.

Our next question comes from the line of Matthew Prisco with Cantor Fitzgerald.

I guess to start, if you could maybe discuss order linearity in the quarter and maybe what you've seen through October and November thus far? And any color you can offer on those booking trends across end markets would be appreciated.

So I think bookings were quite strong all the way through September. So nothing really to point out there. And then October was another good month of bookings where backlog -- overall backlog continued to grow. November has just been a few days, but nothing odd to report.

Yes. So bookings have continued well. The backlog has continued to build. The best way to look at our December quarter is to take the $53 million out of the September quarter, which were the 3 infrequent items. And if you do that, then our December quarter is up about 2.3% from our September quarter versus normal seasonality for December quarter is down 3% to 5%. So therefore, I think we got no issues in how the quarter is how the product chip backlog is, how the bookings are, all that continues well. But the 2 licensing deals, a unique customer program that added $10 million of revenue recognition last quarter. And this strange effect of when you ship the inventory to distributor you assume there's a certain line of [ disti ] price protection on that. And then the rest of it, you recognize as revenue, when you increase the price then that [ disti ] price protection becomes less, and you have to capture that as a onetime revenue in the quarter. And that was about $21 million. So those items made it very noisy. But if you feel through all that, and we're giving you complete visibility, if you take those out, then the quarter is up $2.3 million -- 2.3%. Now in that 2.3%, there's $10 million from 2 acquisitions, $4 million from Hailo and $6 million from Vorago. We broke that out too. So take that out to July and they will -- the 2.3% will go down to maybe 1.5% or something versus down 4% at the midpoint. So there's no problem in how the business is doing and we remain optimistic for the next quarter. Backlog is good, bookings are coming in. That's how you should view the quarter in that context.

Very helpful. And then maybe you can offer some more color on what you're seeing in terms of those constraints today? How those foundry OSAT constraints change over the last 90 days. And I believe you added internal constraints this quarter. So how are you responding to these constraints overall? And how do you expect that situation to evolve as we look into 2027?

So incrementally, we have additional capacity allocation every quarter, which will continue to provide growth, but we could use more than that. So on various different nodes, we have more capacity the following quarter. Therefore, some of the orders from this quarter are scheduled into next quarter. And I believe some of the orders from the next quarter will move into the June quarter, where we have more capacity. We're not unique. I think the entire industry is seeing it. Demand is stronger than the capacity available with whatever is happening in the market right now with data center, AI infrastructure concerns whether they will all be built or not built I think there's a silver lining in it. The whole thing it was just too much hot air. And if some air comes out and capacity becomes available at OSATs and foundries for broad-based business. I think there is silver lining in it.

Our next question comes from the line of Joe Quatrochi with Wells Fargo.

You saw a nice acceleration in industrial demand. this quarter. I was wondering if you could just talk about what you're seeing there? And maybe just on that last line, you said how do we think about your exposure to just everything around the data center infrastructure build like maybe outside of the data all.

So I think when it comes to industrial, industrial is the broadest market segment there is. I mean what is industrial. It's not a narrow set of things. Industrial is factories, cities, hotels, lighting, robotics, medical, industrial is very, very broad, and we're seeing opportunities in industrial. Some of it got driven by industrial was sort of the last one to recover, I think, 2 or 3 quarters ago. So it's seeing substantial recovery from that point and a lot of new design wins in every single area, robotics, medical, process control, factories, farming, equipment appliances, everything is growing. So it's very, very difficult to kind of explain industrial beyond that because it's so broad. The second point of your question is on the data center. So we are early in the data center penetration because we didn't have a lot of Gen5 switch revenue, and you've known from past that we were very late to market on it, and therefore, we lost the Gen5 business. So our -- with Gen6, we are regaining a lot of the market share and a lot of it is not in our revenue. So for us, it's all upside, where if it slows down, it could be a downside for the people who have a lot of that business. In terms of other half of our data center exposure, which is through microcontroller, analog, power, switches, security, USB, Ethernet, these are very broad-based products where data center is really a small portion of the overall shipments to all the end markets. And if certainly, data centers won't get built, a lot of that upside is not in our forecast.

As a follow-up, as we're thinking about just kind of the gross margin puts and takes into the December quarter, how should we think about this underutilization charges is a nice improvement this quarter -- do you keep thinking that's somewhat linear as we move forward?

So it will not be linear. This quarter, we do not -- we are not anticipating any sort of big improvement. We actually built finished goods last quarter and don't really want to do that again this quarter. So we're slowing down some of the assembly and test operations. We're still doing the ramp in our wafer fabs where Steve had mentioned in his prepared remarks, the die inventory is getting low in certain cases. So it's a balance, but you will not see the large reduction in utilization or underutilization charges this quarter that you saw last quarter. But that's still upside in front of us in future quarters.

December quarter historically is kind of weak. So we took the ramp cautiously, but it will accelerate again next quarter because then we go into a good March quarter, followed by 2 strong September and -- June and September quarters before we get the -- another week December quarter next year. So utilization should start to improve, but not linearly.

Your next question comes from the line of Chris Caso with Wolfe Research.

Yes. The question is on pricing. And the impact on the December quarter from some of the pricing actions that you took previously, what was the magnitude of that in December quarter? And I seem to recall from last quarter, that was one of the factors you had expected to be an offset because, again, as you said, you knew some of those 3 items that would not recur in the December quarter. I seem to recall that pricing was going to be one of the offsets for the December quarter.

So you are correct. We expected that to be offset. And the remaining pricing is being effective this quarter because pricing change was done in the middle of the last quarter. However, because of what we described before, some of the fairly large dollar item products with very high margin supply chain cannot fulfill this quarter, which is kind of offsetting the improvement we would have seen.

Right. Okay. So -- and I guess then going forward, would it be right to assume that as you -- as supply improves, as you get the ability to supply some of these higher-margin products perhaps some of the products where the price has gone up, but you can't ship it right now that you would see some of that benefit as you go into calendar '27? Is that the right way to think about it?

Are you talking about benefit in revenue or benefit in something else?

Well, I guess, benefit in revenue from obviously being able to ship the product and the higher price. And then, I guess, more importantly, from the margin side because I think I understand what you're saying is that you didn't see as much of the pricing benefit that you had expected in the December quarter because you couldn't ship those products. So therefore, as you finally are able to ship those products, do you get that margin benefit?

We suddenly get the revenue benefit and we get some of the margin benefit, but the pricing in the industry is from our suppliers is continuing to go up. There's another round of price increase across the industry starting basically in January from most major foundries and OSATs. So some of that will -- because some of that we counted in our price increase when we did the price increase, so that's why you're seeing a very good gross margin for the September quarter and pretty good gross margin in December quarter. If you're expecting it to go back to [ 66.5 ] we're currently not projecting that our long-term target is [ 65 ]. And we think we're going to have to fight that price increase with the suppliers and figure out how we deal with it, whether whether there is a check on our gross margin or we pass it on to the customers further, we haven't figured out that out yet.

Our next question comes from the line of Vivek Arya with Bank of America Securities.

Steve, I wanted to start with more of a macro question based on your industry experience. When you look at all these -- the volatility in semiconductor stocks, do you think that signals to you there are earnings cuts coming? Are we at peak demand? Just kind of what is your crystal ball telling you in terms of where we are in the cycle? And do you think all these worries are justified based on what you are seeing on the ground?

Well, Vivek, I think there is data center and non-data centers. There is nothing to worry about in the nondata center. Industrial is doing fine. Automotive is continuing to recover, its unit count is not growing as much low single digit, but the content is growing significantly and we're winning across new car architecture across the board, and we are doing very well. Aerospace and defense is doing well. Industrial is doing well. Automotive is doing well. communication is doing well. The issue is in data center. It was all that just -- I've been kind of saying it, it could not be true. It's just -- the demand was explosive. You can't just have that large of a percentage of economy, the rate, everybody expected it to grow. So I think some of that is coming in check, whether -- it comes in check that the real estate of data centers can't be built or the power can't be supplied or the new Congress wouldn't let it happen or circular financing finally -- you can't do it anymore. It's not cost effective or people can't monetize. Whatever the reason is or a combination of reason is, the second derivative on data center has to come down substantially. And I think you see market working on it. To me, it's not new. I think it was predictable.

Right. So what do you think that means for Microchip? Because I think in your commentary, you used the phrase that you thought September was a pivot, right? And you're seeing lead times get stretched out. Do you think a lot of that demand is coming because there's a crowding out from the data center? Or do you think there's kind of real kind of demand upturn. So more specifically, if you look at the next several quarters of Microchip, do you think there is still the possibility of above senior growth?

Yes. I think, yes. Like I said, there is data center and non-data center, the nondata center demand is normal. It's doing well. Data center demand has been explosive. If that comes down, it's affecting us would be very small because we're very much in the early stage. We're not doing billions of dollars of revenue in the data center and all of our rest of the markets are growing and data center is 16.7% of our business, not 100% like many of the other companies. Where the crowding out is not in demand, the crowding out is in capacity. Data center has been crowding out capacity for trailing edge nodes because of power requirements and analog requirements and all these other things, microcontrollers and others. So data center has been crowding out capacity at the OSAT and in many cases, we have lost capacity, where some of it we have to bring it inside, some of it we have to move it elsewhere. And it takes time to requalify -- and therefore, the backlog could not be served and will be served some in the next quarter or some in the quarter after. So I think a deceleration on the data center will probably have a positive effect on our business because some of that capacity frees up.

Our next question comes from the line of Vijay Rak with Mizuho Securities.

Just a quick question on the aerospace defense side. It's kind of grown in line with Microsoft's overall growth. When we actually see aerospace defense orders start to come back? Like, are these long need dollars where you have good visibility, and a quick follow-up.

Well, so aerospace and defense grew 41% September over September. So we think it's growing very well. The demand for aerospace and defense is very strong at the primes, but primes have to build up capacity, some starting from real estate. Some of them are being asked to increase the missile production by 5x or 4x. Well, they don't have the floor space to ramp the production to 4x. So they're adding capacity, building and machinery and all that, and we're starting to get the orders, but they're scheduled out when they need it. So I think it's -- aerospace and defense is not a data centric kind of business $10 million today and $200 million tomorrow, it will grow consistently, but it's not a hockey stick because the industrial complex cannot grow at that rate.

Got it. And then on the price increases you mentioned, foundry or set price increases. How much flexibility do you have to kind of pass through or how much world are you anticipating in terms of the impact of those as you look at 27% in terms of margin or -- marginal any headwinds there?

Well, we consider our gross margins very healthy. I mean our last quarter because of some infrequent items was higher than target, otherwise, it's pretty close to target within 100 basis points of that. And we don't want to take a second price increase to our customers so quickly when we just did a price increase in the middle of August. So just like we did that last time, we absorbed it for a while. And if the costs kept going up, then we cast the price increase, I mean we're kind of doing similar. Right now, we're planning to not do a price increase and absorb it. And at these kind of gross margins, we don't feel that our gross margins are low.

Our next question comes from the line of Harlan Sur with JPMorgan.

Book-to-bill is strongly above 1, our backlog even now continues to grow. And I assume that this strong demand profile is across not just advanced foundry based products, but across our internally manufactured products as well, right? Your overall lead times are extending. So your front-end utilizations, I would have expected to move higher this quarter, especially with your commentary on low-dibank inventories as well. So why wouldn't front-end utilization charges continue to decrease this quarter? I understand back-end utilization might be stepping down a bit, but why aren't front-end utilizations continue to increase during December and therefore, lower underutilization charges.

So Harlan, the front end is reducing in the current quarter, but the back end is not going to be as fully utilized as it was the quarter before. So it's the combination of those things, and we don't expect a material change in the overall underutilization charge in the quarter.

I think one other point is that, there's a perception that on in-house build products, there should be no constraint and there should be no lead time issues. The problem is the inside-produced wafers also go through the OSAT. Some of it goes through an internal factory. But some of the packages we don't do, it goes to use. So it sees the same constraints at our foundry wafer. See, and in a lot of cases, some of the capacity at our major OSAT suppliers has been displaced. AI companies which ones came in and took over the whole factory and they pushed everybody out. We brought some in, we brought it some to other OSAT manufacturers, but that is the crowding lot of AI. It's affecting processing of inside wafers as well as outside wafers in assembly and test. And we are rapidly improving that situation ordinarily, we do inside versus outside analysis when there's a payback to do inside. Not everything we can do cheaper inside. Sometimes, we cannot keep the tool fully busy because we could not fully utilize the tools. So therefore, we keep it outside. But when we get displaced outside, we can't produce it a very good margin product, then cost doesn't matter, and we want to bring it in and produce it. So we're making those moves, but those moves take 6 months. And some of this displacement only happened just a couple of months ago. So those are the things we are struggling with. And that's why I said that some deceleration of the AI demand actually would be quite positive early on.

Got it. Okay. I appreciate that. And then as my follow-up, Steve, we are seeing a lot more adoption overall of integration of machine learning and AI capabilities into all of these sort of embedded applications across industrial, automotive, even consumer, right? And for many of these applications, you don't need a stand-alone MTU like your acquired product family from Hailo, but you could leverage embedded NPU accelerators, right? Can you just remind us your 32-bit MCU product lines, do you have product families with embedded MPU and are you seeing accelerated adoption of NPU-based MCUs for these types of embedded like machine learning AI applications?

So our in-house microcontroller and microprocessor product lines did not have AI acceleration in them no hardware acceleration. We had some software acceleration. We did have some hardware acceleration available on our FPGAs and our Data Center Solutions business unit. But not on our microcontroller business units and the microprocessor business unit, we did not. When we bought Halo in our microprocessor business unit, we were developing products. The first of those products was going to be available in about 2.5, 3 years. And the second of them are going to be available in about 4, 4.5 years because they were serial in nature. The product line that Hailo brought to us accelerates the road map by 5 years. So therefore, we have canceled the 2 products that we were developing in the NPU business unit because they have been replaced by Hailo. In fact, I have merged the Hailo and MPU business unit together and we are advancing the Hailo road map, and we have canceled our NPU road map in the microprocessor business unit. That's how important halo has been for us.

I see. Okay. That makes a lot of sense.