Vicor (VICR) 2026财年第三季度业绩电话会议：专利费增长与Fab-2扩建lg...

核心要点

• 2026财年第三季度产品及专利许可费收入达1.971亿美元，环比增长37.5%，同比增长78.5%。
• 专利许可费收入环比增长119.1%至6660万美元。管理层将许可收入流描述为经常性的，但季度贡献可能仍会出现波动。
• 毛利率扩大至68.3%，环比提升1,030个基点，主要得益于专利许可费收入的增加以及产品出货量、效率和产品组合的改善。
• 净利润为6330万美元；按4760.9万股稀释后股数计算，GAAP稀释后每股收益达1.33美元。
• 一年期未完成订单环比增长7%至4.06亿美元，第三季度订单出货比大于1。公司在第四季度的前两周录得1亿美元的新增订单。
• 管理层预计第四季度产品收入将环比增长约10%，假设本季度没有达成新的许可协议，专利许可费收入将至少达到4300万美元。

关键财务数据

业务与运营表现

先进产品在本季度的收入结构占比有所提升，从第二季度的65.7%升至总收入的70.4%。得益于出货量增加、效率提升以及更有利的产品组合，产品毛利率亦有所改善。

技术许可成为了主要增长动力。Vicor按第二阶段单台专利许可费率，向某领先的AI原始设备制造商（OEM）授予了其首个VPD许可。管理层表示，专利许可费收入与被许可方对技术的使用量成正比，既可以按单台收取，也可以通过覆盖预计使用期限的全包协议形式来收取。

公司表示，其核心业务及许可业务的需求依然强劲。第四季度前两周的新增订单已达1亿美元，其中包括来自Vicor主要客户的第二代VPD模块大额订单。管理层还表示，其ATE（自动测试设备）业务目前的年收入已超过1亿美元。

Vicor强调，第二代VPD和分比式电源架构（FPA）是解决AI系统中日益增长的负载点电流需求的核心。据管理层介绍，其全FPA方案可将AI处理器的功率损耗降低10%至15%，同时提供超过IVR（集成稳压器）40倍的电流增益，以及VR（电压调节器）3倍的电流密度。

管理层业绩指引

管理层预计第四季度产品收入将环比增长约10%。先进产品收入预计将再次实现两位数的环比增长。

Vicor预计，若本季度未达成新的许可协议，第四季度专利许可费收入将至少达到4300万美元。管理层称这是一个保守的基线预测，因为即便没有新交易，现有的协议也能产生额外的许可费。

预计短期内法律费用平均约占专利许可费收入的20%，这反映了与许可业务相关的或有成本。

在产能扩张方面，Vicor预计将于2027财年第二季度在Fab-2（2号晶圆厂）进行初始设备部署。公司计划优先为Fab-1（1号晶圆厂）面临瓶颈的工艺环节增加设备。管理层表示，Fab-2和Fab-3可支持超过五倍的产能扩张，并预计在无需进行外部融资的情况下为这两座厂房提供资金，不过未提供详细的资本支出估算。

风险与关注事项

• Fab-1正接近满负荷运转，特定工艺环节已在短期内构成了产能限制。
• 据管理层称，专利许可费收入本质上是经常性的，但由于许可签署时间、追溯补缴款项以及现有协议下的使用情况，各季度的业绩可能仍不均衡。
• 法律成本随着专利许可费收入的增加而上升，预计短期内将接近许可费收入的20%。
• 尽管管理层报告称与超大规模企业及OEM厂商沟通积极，但除主要客户外，第二代VPD的采用仍处于早期阶段。
• Fab-3的建设时间尚未确定。管理层计划部分根据客户需求和产能预留讨论来做出决定。
• 随着Vicor的ITC（美国国际贸易委员会）执法程序推进以及后续被许可方费率提高，潜在的许可结果仍存在不确定性。

分析师问答环节亮点

分析师关注的焦点是专利许可费收入的可持续性。管理层强调，许可业务现在应被视为经常性收入而非一次性收入，同时也承认其仍存在波动性。第四季度4300万美元的许可费展望未计入新协议，且可能通过现有许可协议下的活动超预期达成。

在制造产能方面，管理层表示Fab-2将于2027财年第二季度开始补充受限的工艺环节，而非立即复制Fab-1的所有运营环节。这种分阶段的方法旨在更快、更具成本效益地提升产能。

管理层将购买Fab-3土地解释为提供更长期的灵活性，而非预示着将立即开工建设。位于现有一栋33万平方英尺建筑内的Fab-2仍是近期重点。

关于客户采用情况，管理层表示，除主要客户外，围绕第二代VPD的讨论仍处于早期阶段。不过，Vicor报告称在Open Compute峰会上进行演示后市场兴趣浓厚，并表示已掌握正在开发中的芯片动态以及潜在的近期机会。

分析师还询问了基于IVR的系统。管理层表示，Vicor可以通过以1.8伏供电来支持使用IVR的客户，同时坚持认为其自身的分比式电源方案可为大电流轨提供更为高效的解决方案。

业绩电话会议完整文字记录

---

完整财报电话会议逐字稿

管理层陈述

Operator

Ladies and gentlemen, thank you for standing by. Welcome to the Third Quarter 2026 Vicor Earnings Conference Call. [Operator Instructions] Please be advised that today's conference is being recorded. I would now like to turn the conference over to Jim Schmidt, Chief Financial Officer. Please go ahead.

James Schmidt

Thank you. Good morning, and welcome to Vicor Corporation's earnings call for the third quarter ended September 30, 2026. I'm Jim Schmidt, Chief Financial Officer; and I'm in Andover with Patrizio Vinciarelli, Chief Executive Officer; and Phil Davies, Corporate Vice President, Global Sales and Marketing.

Earlier this morning, we issued a press release summarizing our financial results for the 3 and 9 months ended September 30, 2026. This press release has been posted on the Investor Relations page of our website, www.vicorpower.com. We also filed a Form 8-K today related to the issuance of this press release.

I remind listeners, this conference call is being recorded and is the copyrighted property of Vicor Corporation. I also remind you various remarks we make during this call may constitute forward-looking statements for purposes of the safe harbor provisions under the Private Securities Litigation Reform Act of 1995.

Except for historical information contained in this call, the matters discussed on this call, including any statements regarding current and planned products, current and potential customers, potential market opportunities, expected events and announcements and our capacity expansion as well as management's expectations for sales growth, spending and profitability are forward-looking statements involving risks and uncertainties.

In light of these risks and uncertainties, we can offer no assurance that any forward-looking statement will, in fact, prove to be correct. Actual results may differ materially from those explicitly set forth in or implied by any of our remarks today. The risks and uncertainties we face are discussed in Item 1A of our 2025 Form 10-K, which we filed with the SEC on March 2, 2026. This document is available via the EDGAR system on the SEC's website.

Please note the information provided during this conference call is accurate only as of today, Tuesday, October 20, 2026. Vicor undertakes no obligation to update any statements, including forward-looking statements made during this call, and you should not rely upon such statements after the conclusion of this call. A webcast replay of today's call will be available shortly on the Investor Relations page of our website.

I'll now turn to a review of our Q3 financial performance, after which Phil will review recent market developments, and Patrizio, Phil and I will take your questions. In my remarks, I will focus mostly on the sequential quarterly changes for P&L and balance sheet items and refer you to our press release or our upcoming Form 10-Q for additional information.

As stated in today's press release, Vicor recorded product and royalty revenue for the third quarter of $197.1 million, up 37.5% sequentially from the second quarter of 2026 total of $143.4 million and up 78.5% from the third quarter of 2025 total of $110.4 million. Advanced Products revenue, which includes royalty revenue, increased 47.5% sequentially to $138.9 million, and Brick Products revenue increased 18.4% sequentially to $58.3 million.

Royalty revenue increased 119.1% sequentially to $66.6 million. A fraction of this increase due to initial royalty payments from the new licensee was nonrecurring. As we add more licensees, growth in royalty revenues will be less lumpy. Shipments to stocking distributors increased 24.8% sequentially and increased 16.2% year-over-year. Exports for the second quarter decreased sequentially as a percentage of total revenue to approximately 36.1% from the prior quarter's 46%.

For Q3, Advanced Products share of total revenue increased to 70.4% compared to 65.7% for the second quarter of 2026, with Brick Products' share correspondingly decreasing to 29.6% of total revenue.

Turning to Q3 gross margin. We recorded a consolidated gross profit margin of 68.3%, a 1,030 basis point increase from the prior quarter. Q3 gross margin increased 1,080 basis points from the same quarter last year. Gross margin improved both as a result of increased royalty revenue as well as better product gross margins as we achieved gains in volume, efficiency and mix.

I'll now turn to Q3 operating expenses. Total operating expense increased 29.6% sequentially from the second quarter of 2026 to $62.5 million. Within operating expense, legal expense, a substantial portion of which is contingent on royalty revenue, increased by approximately $11.1 million from the second quarter to $16.1 million. On a run rate basis, we expect legal expense to average approximately 20% of royalty revenue in the near term.

The amounts of total equity-based compensation expense for Q3 included in cost of goods, SG&A and R&D was $1.203 million, $2.655 million and $1.483 million, respectively, totaling approximately $5.3 million.

Turning to income taxes. We recorded a tax provision for Q3 of approximately $15.4 million, representing an effective tax rate for the quarter of 19.6%. Net income for Q3 totaled $63.3 million. GAAP diluted income per share was $1.33 based on a fully diluted share count of 47,609,000 shares.

Turning to our cash flow and balance sheet. Cash and cash equivalents totaled $513.1 million in Q3, an increase of $59.6 million sequentially. Accounts receivable net of reserves totaled $102.5 million at quarter end, with DSOs for trade receivables at 34 days. Inventories net of reserves increased 6.2% sequentially to $110.9 million. Annualized inventory turns were 2.3. Cash flow provided by operating activities totaled $69.7 million for the quarter.

Capital expenditures for Q3 totaled $16.1 million. We ended the quarter with a construction in progress balance primarily for manufacturing equipment of approximately $26.4 million and with approximately $48.6 million remaining to be spent.

I'll now address bookings and backlog. Q3 book-to-bill came in above 1 and 1-year backlog increased 7% from the prior quarter, closing at $406 million. 2026 is the year in which Vicor's innovative products and technology licensing practice came into focus within the industry. With incremental capacity, we expect a 10% increase in Q4 product revenue and at least $43 million in royalty revenue, assuming no new license agreements are closed this quarter.

To achieve the product revenue growth objective, we are expecting another double-digit sequential increase in product revenue in our Advanced Products portfolio. In Q3, we achieved operating ratios for both profitability and expense approaching our long-term financial model. Further revenue growth will result in further margin expansion. Phil?

Philip Davies

Thank you, Jim. Demand across our main business segments and in our IP licensing practice is strong with $100 million in bookings in the first 2 weeks of Q4, which included a large order for second-generation VPD modules from our lead customer.

The purchase of 2 sites for Chip Fab-2 and Chip Fab-3 reflects the need to expand capacity in support of increased demand. Substantial capacity expansion will be necessary based on AI industry trends and customer enthusiasm for our second-generation VPD. Meetings held with major hyperscalers and OEMs at last week's Open Compute confirmed that first-generation VPD systems using VRs, which infringe Vicor patents, are struggling to keep up with the requirements of next-generation AI systems. Our targeted audience was impressed by Vicor's second-generation VPD with its distinct manufacturability and performance advantages.

Open Compute this year focused on 800-volt DC rack distribution through inefficient intermediate bus voltages from 12 volts to 6 volts or possibly 1.8 volts to a point of load approaching 0.5 volts, underscoring the inherent limitations of VRs and IVRs. Lack of adequate current density and current gain at the point of load sets the stage for broader adoption of Vicor's second-generation VPD, providing high current density and much higher current gain.

With its transient performance eliminating voltage undershoots, Vicor's proprietary all-FPA enables a 10% to 15% reduction in AI processor power loss with over 40x the current gain of IVRs and 3x the current density of VRs. Demo systems for second-generation VPD and all FPA have been set up in our Silicon Valley applications laboratory and are available for selected OEM and hyperscaler engagements.

On the licensing front, we granted our first VPD license to a leading AI OEM based on Stage 2 unit royalty rates and responded to inquiries from 2 hyperscalers seeking all-inclusive licenses quoted at Stage 2 and Stage 3 prior to the recent ITC decision to investigate VPD infringement. Even though leading OEMs have already obtained a license to our VPD patents, others may not have yet realized that they are importing or planning to import infringing products.

Catching their attention, Vicor has requested the ITC to issue a limited exclusion order barring importation of unlicensed computing systems that infringe Vicor's VPD, like the ITC did before in an earlier investigation involving our NBM patents. As hyperscalers and other OEMs come to realize the impact of an importation ban, they will take licenses based on escalating royalty rates established in our licensing practice, which favors early adopters.

The decision to focus on a set of top 100 customers and top channel accounts or TCAs with our global distribution partners across our 4 markets and 4 regions globally is clearly paying off with both bookings and opportunity pipeline growth. Our customers are on the cutting edge of high-growth applications with the toughest power delivery challenges and specifications. Our leading ATE customers for ASIC and memory test systems, which now represents over $100 million in annual revenues shows no signs of slowing down, and we are actively discussing capacity reservations to ensure supply in coming years.

Major new product introductions are also underway in our industrial and aerospace and defense businesses with market expansion now occurring with new nontraditional companies that have entered the A&D market with new autonomous technologies. Successful execution is achieved with focus, and we are very focused.

Thank you, and we will now take your questions.

Operator

[Operator Instructions]

And the first question is going to come from Quinn Bolton with Needham & Company.

分析师问答

Quinn Bolton

Congratulations on the strong product and licensing outlook. I guess I wanted to start, Jim, just you had mentioned that the royalty revenue in the third quarter of $66.6 million did include some catch-up royalties. Wondering if you could help size that in the third quarter? And then I've got a couple of follow-ups.

James Schmidt

It might be best to just think about the $43 million that I mentioned, Quinn, as the baseline amount that we can count on now with no new deals as a starting point for Q4.

Patrizio Vinciarelli

Yes so thinking -- Quinn, thinking about this, what's the meaning of nonrecurring? I think that which was level nonrecurring has been recurring pretty periodically, right? So as we pointed out, in Q3, there was a contribution that is nonrecurring. But then early in Q4, we got some other contribution.

So I think we should all expect that our licensing income is going to be growing, while we're still in a licensing arrangement with a relatively small subset of all of the hyperscalers and OEMs. It is still going to be lumpy. There's going to be more "nonrecurring contributions." It's not going to be until there's a crossing of the chiasm with respect to the need to license Vicor technology for freedom to operate in AI applications given the strong component of a Vicor IP that is inherently involved in these systems.

But with some lumpiness, it is going to be a significant growth business that we're now estimating to reach levels next year and the year after that beyond the levels that I had forecasted a year ago.

James Schmidt

And Quinn, I'll just add to what Patrizio said, it's better for, I think, the audience and the analysts to think about the royalty stream as recurring now and not a nonrecurring activity, just to put it in perspective. The nature of the agreements are recurring.

Quinn Bolton

Got it. Yes. No, I think I understand that if you sign new licenses, there may be catch-ups with those additional license agreements, but the $43 million sounds like a good go-forward base. And then obviously, we can add on to that as the company signs new licenses.

I wanted to turn to the second and third chip fab. I think we're all expecting you to announce a second site. So it was great to see you announce both a second and a third site. But can you give us a sense for the second site, when would you think that, that facility would come online and begin contributing to product revenue? Is that a sort of 12-month time line? Is it 18 months? What's a good sort of time line for investors to be thinking about with that second site?

Patrizio Vinciarelli

We're likely going to have some initial equipment deployment in Q2 -- Q2 of '27.

Quinn Bolton

And would that be able to produce sort of full set? Or is that just first equipment deliveries, Patrizio?

Patrizio Vinciarelli

Yes. So it's going to be incremental steps with respect to those processes where we first ran into capacity limitations in Fab-1, right? So in Fab-2, we're going to bring up lines that address the shorter-term limiting factors and move on from those to other ones.

Quinn Bolton

Got it. So it may not be sort of full production in the mimic of Fab-1, but you'll bring in those production steps where you're most constrained. So the impact would be to increase the company's total manufacturing beginning in the second quarter?

Patrizio Vinciarelli

That's right, yes. For those process steps where we still have ample capacity at Federal Street, there is no need, there is no rush to deploy additional equipment there. So we're going to be able to do it incrementally and more cost effectively and more predictably.

Quinn Bolton

Got it. And then the last one for me. Phil, you mentioned good meetings at Open Compute last week. And just kind of wondering if you could address what kind of demand are you seeing for the factorized power, the 48 core solutions as the VRs and IVRs do look increasingly challenged. When do you think you might see sort of adoption of a second-gen VPD outside of your lead customer based on some of those conversations?

Philip Davies

So conversations are still early, Quinn, in that regard. We do have line of sight to new chips that are under development that are sort of further out, if you like. But I do think that there could be opportunities in the shorter term based upon the inefficiencies of VRs and IVRs. So we'll be taking a look at that as we move forward.

Patrizio Vinciarelli

So again, we're looking at all these opportunities in [ December ] reference of capacity limitations over the next 12 months. It's not that we can engage with any hyperscaler OEM that has got a need. Virtually, all of them do. Because as you know well, there's been a lot going on with respect to an 800-volt bus and converting that to 6 volts or 12 volts or back to 6 volts on the way to 1.8 volt.

The real issue with respect to enabling AI doesn't have to do with what happens in effect some distance away on 800-volt bus or even intermediate bus, the real constraints have to do with the point of load. And they have to do, as we've been articulating because of the fact that IBA, the intermediate bus architecture, which is how the competitive industry has been doing it, is gathering in tension between efficient power distribution and efficient duty cycle and the requisite current density and current gain at the point of load.

As I've been saying for quite some time, I think all the focus about 800 volt is really missing the boat with respect to the real challenge. It's not that there isn't a benefit to 800 volt. There is a benefit. It's a few percentage points of system efficiency. But in that point of load, power system will keep AI from progressing on its performance trajectory, and that's what we're focused on.

Operator

And our next question is going to come from Justin Clare with ROTH Capital Partners.

Justin Clare

First, I wanted to start off. So backlog increased to $406 million this quarter. And I was just wondering if you could discuss the visibility that you have now into kind of fully utilizing Fab-1, when you expect to reach that milestone. So just looking beyond the backlog, kind of what visibility do you have into additional customer orders or commitments from customers that would support that full utilization?

Patrizio Vinciarelli

So as of yesterday, the backlog is actually closer to $400 million. But beyond that, we see robust demand. Phil and I were on a call yesterday with a major customer seeing demand doubling every year for the next several years. And they have a lot of visibility with respect to the request for their systems.

So the challenge in the short term is to bring up additional capacity to do so ahead of demand, taking care of customers. We have many customers, most customers that are single source with us, where we have obviously a particular responsibility with respect to ensuring that we're not going to be getting their growth. A lot of it is directly AI related, but some of it is indirectly related in that the AI growth path is fueling, as you know, a growth path for all of the portions of the industry that play one support role or another.

Justin Clare

Got it. Okay. And then just shifting to the capacity announcements here that you made for Fab-2 and Fab-3. I was wondering if you could just clarify, you indicated that those additional sites could increase capacity nearly 5x. Is that 5x figure related to kind of physical manufacturing space, production output or revenue-generating capacity? And I'm just wondering if you could potentially share what the revenue potential might be as you move from Fab-1 to Fab-2 and Fab-3, assuming full utilization of those facilities?

Patrizio Vinciarelli

So around Fab-1, we are space constrained. In other words, we have a relatively small parcel of land. So we made sure with both Fab-2 and Fab-3 to avoid getting limited by, in effect, elbow room within that particular site. Both sites have significant upside that could support even more than a 5x expansion.

And again, the methodology for expanding capacity, as you might recall, we looked first at just industrial parcels, then looked at buildings in order to reduce the time to accessing a new facility. We ended up doing one of each within a triangle is about 30 miles off to the west and to the north of Fab-1, relatively short distance, but with a significant level of dependency with respect to power fees and all that would be a consideration regarding redundancy with respect to this incremental capacity. The 5x is a conservative number. It could be more than that because, again, unlike Fab-1, we're not going to be limited by elbow room within either one of these sites.

Justin Clare

Got it. Okay. And then just one more. Wondering if you could share what you expect in terms of the capital investment for Fab-2 and Fab-3 and what the anticipated timing of that investment would be?

Patrizio Vinciarelli

So we are not prepared to share any details today. We're in the planning phase with respect to Fab-2. I think possibly as of our next conference call, we will share some quantitative data with respect to capital investment. But it's safe to say that we're going to be able to fund Fab-2 and in due course, Fab-3 without need for raising any capital. We're adding cash at an increasing rate. And Fab-2, with this investment, is well within our capability. We want to make a dent with respect to our cash flow.

Operator

And the next question is going to come from Richard Shannon with Craig-Hallum Capital Group.

Richard Shannon

Jim, I'd love to put the numbers you mentioned in your prepared remarks here to get to a fourth quarter revenue number. You mentioned a royalty number. I think you mentioned Advanced Products' growth or total product growth, I'm not sure which one. That was coming up with revenues kind of in the $180 million to $185 million range for the fourth quarter. Am I doing the math right? Or if not, can you correct me?

James Schmidt

I'm going to give you a specific question (sic) [answer], but I do want to make clear, Richard, that the number I quoted for 10% product revenue growth is total product revenue growth, okay? And what I said about Advanced Products' revenue growth within Advanced Products will be yet again another double-digit increase in that category of products.

Patrizio Vinciarelli

Let me give you more color with respect to this. So as we discussed earlier, the licensing side of the business has got lumpiness, potential for additional deals. We are taking the long-term horizon. We're not in a hurry to get any deal done. As time goes on, the cost to hyperscalers and OEMs for VPD license goes up at a pretty fast clip as we progress through the investigation of the ITC.

But there could be deals done this quarter. And for that reason, we really can't give you a guidance that we would be comfortable with given the level of uncertainty. And for that reason, what we provided with in terms of helping you set expectations is with respect to products side, that which we can predictably expect to deliver, which is a low double-digit increase in product revenue, right, hardware revenue, separate apart from licensing income for which we provided a minimum of $43 million, and that assumes no new deals are done within Q4.

James Schmidt

Richard, one other point to mention. The $43 million, as we've said and Patrizio just said, doesn't rely on any new deals. But inherent in the licensing agreement we have, there's potential for incremental royalty revenue just as a function of the existing arrangement, and we can't predict that exactly.

Patrizio Vinciarelli

That's right. So it could be that the number that we are forecasting at the minimum of $43 million ends up larger than that even without new deals. So it's a very conservative number.

Richard Shannon

It's a baseline, and that's what investors like to see. So I just want to verify that.

My second question is related to licensing kind of the notion of run rate here. You mentioned kind of a $43 million run rate, again, assuming nothing else gets done, and let's start with that notion in mind here. You've mentioned that most of the licenses you've done here in the past have been 2-year terms here. And I think the first one is coming up for renewal and will be expired in the not-too-distant future. And so I want to get a sense of your expectation of renewals of those so that we can assume a real recurring run rate that's going to go for many quarters in a row. So if you could comment on that, that would be great, please.

Patrizio Vinciarelli

Yes. So I think it's becoming more and more apparent that Vicor pioneer role with respect to a density power system is multifaceted, right? As of -- as recently as a year ago, even within a shorter time frame, all the focus was on NBMs as part of IBA sort as a stepping stone to the pal road for competitive alternatives. Again, we don't use NBMs because the factorized power system doesn't require intermediate bus converter, but virtually all of the competitive alternatives do.

The focus clearly has recently shifted to VPD, which is on a different scale of significance, right? Its significance in terms of enabling AI power systems is much greater than NBMs because it's not the intermediate bus converter, but it's the point of load system, which is where the spotlight, where the focus should be. And so there is greater value associated with that, and we're capturing that value. We capture that value with the first VPD license. We'll be capturing that value with additional VPD licenses or inclusive licenses that involve the combination of NBMs, VPD and other vital technology that over time is also to become an issue for unscrupulous competitors to contend with.

Operator

And the next question will come from [ John Dillon ] with DNB Capital.

Unknown Analyst

Congratulations to the whole Vicor team, incredible quarter. Phil, my question is really for you. It's, with the exception of Vicor, I've never heard of any company claiming that power supplies could enable higher performance processors. Yet Cerebras just announced a new CS4 AI processor and Angela Yeung, the Senior VP of Products, stated that the wafer scale engine stayed the same as the CS3.

So the CS3 had the same memory and logic, but had 2x the increase in speed. And with 3 of these together, they had 10x the throughput of 3 wafers. And she contributed a big part of the increase was the ability to deliver higher density power to the same wafer. So Phil, I find this extraordinary. Can you walk us through that and let us know what the reactions are from your new target customers now that you're demonstrating the benefits of the second-gen VPD?

Patrizio Vinciarelli

So John, we're not going to make any comments that are specific to any customer, right? I'm going to answer your question in general terms with specificity.

Every generation of AI processor is stepping up its current requirements by a dramatic percentage increase. What used to be very high current requirements not that long ago, a few hundred amps, right, following the era of Intel processors that were barely consuming 100 amps. It's now at the 1,000 amp level and on its way to thousands of amps. And then there are leading-edge applications such as wafer-scale engines where you're dealing with tens of thousands of amperes going to 100,000 amperes.

That's where the action is. That's where inference and other AI enabling capabilities get realized, and they're fundamentally fueled by current. They're fueled by current at ever decreasing voltages, which makes factorized power with current multiplication, the technology of choice. And conversely, fundamentally handicaps the competitive approach where the upper voltage is, in effect, averaged down by mixing a low-voltage source, i.e., one that is inefficient in terms of its power distribution with ground, which is fundamentally [ Joules ].

That kind of technology is inherently handicapped with respect to keeping up with AI demands. And conversely, our technology was envisioned a long time ago, and we kept on improving it, right? We got lots of new developments related to factorized power, it's IP, it's components. And these are reflected in the level of performance that we're bringing out with every new generation, right?

The initial demonstration wiggles for second-gen VPD reached 3 amps per square millimeter. We are not that far away from a 5, 6 amp per square millimeter capability. We're going from what used to be in a time frame where we're powering a GPU many, many years ago with current multipliers to the level of 3, 3.5 millimeters, we're now 1.5 millimeter. In 6 months, we're going to be at 1 millimeter. So it's this kind of technology, the packaging technology, power conversion engines, control systems, unique components that enables AI.

Philip Davies

So John, just to build on that, I think just to build on what Patrizio said, I think that, that was on display last week at Open Compute with our meetings with several big customers in terms of their level of excitement around our second-generation VPD. They've been waiting a while to see it. We've been focused on our lead customer. But now we're here. We've got the demo system set up. They were excited by what we showed them. So I think it's now up to us to roll this forward and to bring these deals home.

Patrizio Vinciarelli

Within our capacity constraints -- near-term capacity constraints.

Unknown Analyst

Okay. Now before you've talked about pricing per watt. So with twice the power with the fifth gen solutions, are average selling prices going up?

Patrizio Vinciarelli

Let's not get into that. I think what we can say is that our total competitive standing, which obviously leads with performance, but needs to deal with cost is becoming stronger and stronger because we have the right approach to solving these point-of-load challenges. And with that right approach, inherently, there are scalability and cost advantages that will materialize as we scale up from a nearly full Fab-1 to Fab-2 and Fab-3. We're going to have the economies of scale. We believe we're on the right track and the rest of the industry is foolishly going down at that end road.

Unknown Analyst

Excellent. And the IVR customers you mentioned last time, how is that progressing? And that will be my last question.

Philip Davies

So again, as we talked about, John, IVRs, it's very early days for IVRs in terms of really delivering solutions to customers. They're intrigued by using IVRs on some of the rails. I think for the high current rails, we have the best solution. So it might be all Vicor. It might be some Vicor and some IVRs. We don't know yet. We'll see how that plays out. But IVRs have a way to go yet in terms of proving their performance sort of claims.

Patrizio Vinciarelli

So we had several meetings last week on the IVR front with 2 IVR companies and an hyperscaler interested in solutions, IVR solutions, using one of these 2 companies. And so we were told by the hyperscaler that when it comes to feeding the IVRs at 1.8 volt, we will be the only solution.

I think it remains to be seen. The jury is still out with respect to how soon these IVRs will become a practical reality. The good news is that the hyperscalers saw that we have a solution that doesn't involve IVRs. So fundamentally, they can come to us for either alternative, one that involves IVRs with that which IVRs provide, which is to be honest and fair, they provide flexibility, at least now in theory. But in reality, they will, over time, provide flexibility. But that flexibility comes at a significant cost.

Whenever you sit down and you deal with the real numbers, not the marketing numbers, the insertion cost to, in effect, process power through an IVR stage is close to -- it's upwards of 10%, and it could be pretty close to 15%. And for that 10%, 15% of extra power dissipation, all that you're getting, all that the customer is getting is a current gain of 2.5. Now, it does provide flexibility. But from a power system efficiency perspective, it's a very challenged proposition.

We're open-minded though. We're perfectly willing, eager to play a support role with respect to IVR application. And fundamentally, the same point-of-load technology enables us to deliver dozens of apps at 0.5 volts with the requisite current density, bandwidth and all the right attributes. That same technology can deliver 1.5 volt to an IVR. And if customers, for whatever combination of reasons, want to go down that path, we'll be there to support them.

Operator

And our next question will come from Richard Shannon with Craig-Hallum Capital Group.

Richard Shannon

Let me ask a follow-on here. In the press release and in the call here today, you've used a couple of terms that I'm unfamiliar with, Stage 2 and Stage 3 licenses. If you can explain those, that would be a big help here.

Patrizio Vinciarelli

Yes, I'd be happy to. Actually, if you go to our website, we now have a licensing page. And on the licensing page, you can get an answer to the question. And it's a quantitative answer to the question as it relates to royalty rates, which change as a function of the stage.

So fundamentally, for any Vicor patent in our portfolio that has not been asserted, a prospective licensee can take a license at Stage 0. That's the most cost-effective license. It is most cost effective because Vicor has an edge to get into deals with law firms to assert the IP and go through, in effect, the enforcement effort. But if the patents are being infringed and we have to assert them to the fact that there's a cost to doing that as time progresses and the enforcement effort advances, we increase the royalty rate in relationship to what we call Stage X.

So after Stage 0 comes Stage 1, 2, 3. So 3 begins when the ITC institutes an investigation. And the first roughly 10 months of that time frame, which is during all the phases of the ITC process before a hearing, we have increments in the royalty rate, as you can see from the website page at the level of about 5% per month. As we get into the hearing and then the initial determination by the administrative law judge at the ITC, the final determination, an exclusion order and enforcement of the exclusion order by Customs and Border Patrol, the cost to, in effect -- late licensee that in effect, has taken a wait-and-see attitude or has approached this from the perspective of taking its chances, it keeps escalating because our cost has gone up. And with that, we deserve to fetch more with respect to the patent and its value.

So all of this is intended to encourage early adoption, encourage licensee not to take their chances, not to roll the dice. We got a track record of knowing what we're doing, and we're going to keep doing that. And we're very determined that we're very well equipped with partners that have a vested interest in the right outcome, which is not the case for our opponents. So that's how the various stages of this process relate to time and the cost of the license, its royalty rate. And if you look at the web page, there are curves that define this progression.

And part of the reason for this is to have a level playing field, right? It's very important that we treat each and every licensee fairly. All of these licensees have got most favored the licensee provisions, which means that you can grant to a following licensee a license on better terms than a prior licensee under similar conditions. Well, similar conditions, among other things, means at what stage the prior licensee took a license as compared to a new licensee. So we have documented our practice, among other things, to ensure that we have a complete transparency and fairness and consistency in our methodology.

Richard Shannon

Patrizio, thanks for that detail here, and I see the web page here, very comprehensive and complete. My follow-on question is regarding the AI OEM that you signed a license with in the third quarter. A 2-part question. The first one is, why was that seemingly mostly a nonrecurring or onetime license versus something that might be ongoing? And second of all, can you state which stage this third quarter AI OEM was licensed at?

Patrizio Vinciarelli

So we didn't say that that's nonrecurring to the contrary. Maybe, the source of potential confusion was my comment to the fact that, that which has been characterized now several times as nonrecurring keeps recurring. So by definition, it's not nonrecurring.

But beyond that, what we recorded in Q3 from this first pure VPD license is far from nonrecurring, whichever way you want to look at that word. In fact, there was at the beginning of Q4, another significant license PO that came from the same new licensee. So these licenses are structured, whether they're all inclusive or per unit to be, in effect, proportionate to the use of the technology by the licensee.

And obviously, as time goes on and licensees make greater use, they pay royalties in direct proportion to that. We have different ways of making this happen, fundamentally 2 approaches, either on a per unit basis or by estimating what the use by the licensee is going to be over a couple of years and having an all-inclusive license that deals with that time frame before we take on the next couple of years.

James Schmidt

Richard, just to amplify what Patrizio said, it's better to think about the royalty revenue as recurring, okay? It's a licensing practice that is now up and running. It's generating royalty revenue. It's proportional to the use of what otherwise would have been infringing technology. So it's very important that everybody understand that. We've used a couple of quarters of catch-up, but really think about this as recurring revenue now. It's in the P&L.

Patrizio Vinciarelli

They keep on coming at increasing levels. So I think it was perfectly appropriate, let's say, a couple of years ago to look at a lump sum contribution in one quarter as nonrecurring. Back then, that was obviously the conservative way to think about it. But now to Jim's point, obviously, there's been a lot of recurrences of nonrecurring things, and they're no longer nonrecurring.

Operator

And our next question will come from Justin Clare with ROTH Capital Partners.

Justin Clare

Just wanted to follow up on the capacity expansion here, particularly for Fab-3. Wondering how we should interpret the land acquisition that you made for Fab-3 at this point. Is that a commitment to add additional capacity? Or is that -- or should we view that more as an option to potentially accommodate longer-term growth? And then just trying to think through how long it might take for you to get to fully utilizing 5x the capacity that you have with Fab-1.

Patrizio Vinciarelli

So we bought that land. We closed a week ago or thereabouts. So -- but we're not going to act on it in terms of putting up a building there for now. The logic of having that flexibility was in effect, to make the most of the process that we went through earlier this year in investigating these various options that carry different trade-offs, right?

So on a new campus, we have total flexibility with respect to the complement of buildings, how they relate to each other in terms of next fabs. That's very valuable in terms of medium, longer-term planning as distinct from the need to act in the very short term with respect to expanding capacity within an existing building. So we now have the wherewithal to take very short-term steps with the time scale of 5, 6 months on the one hand, within existing infrastructure, including a 330,000 square foot building as well as doing it from the ground up in a different site.

We're going to act first on the Fab-2, which is in Merrimack to the west of Andover. And it's yet to be determined when we want to pull a trigger on Fab-3. Again, Phil and I yesterday were having a call with the CEO of a major customer wanting to, in effect, reserve capacity over the next few years. These kinds of discussions are going to facilitate decisions with respect to when to pull the trigger on the next incremental capacity.

James Schmidt

And it might be worth mentioning the call they had was not with an AI customer.

Justin Clare

Okay. Interesting. And then just one other follow-up. You mentioned that you plan to, I think, start adding equipment in Q2 for Fab-2. And so just wondering, do you anticipate running at near full capacity for Fab-1 next year and that kind of justifies the ramp-up of Fab-2 into next year?

Patrizio Vinciarelli

Yes. Yes. So unfortunately, for the near term, we have certain capacity limitations, right? And they are, to be clear, dependent on specific platforms related to particular process steps. They are such that we can grow revenues. As Jim pointed out earlier, we expect low double-digit increase in product revenues this quarter. But as we've been explaining over the last several calls, we're getting closer to getting to full capacity utilization of Fab-1, hence, the need to step with some of the process steps into Fab-2 starting Q2 of next year.

Operator

Thank you. And this is all the time that we do have for questions today. So this will conclude today's conference call, and thank you for participating, and you may now disconnect.

原文链接