Maxlinear and Silicon Motion, Vicor and Monolithic Power and New Product Launches
Maxlinear has a legal battle on it's hands, and what's the new product Vicor is talking about?
Last week the best show in semiconductors was the Maxlinear and SIMO merger news. I will catalog the entire saga and outline my thoughts on where things go.
MaxLinear’s Acquisitive Nature
Before we talk about SIMO, let’s talk about Maxlinear’s acquisitions. MaxLinear acquired an asset in April 2020 that was almost the perfect fit. As Intel shed assets, they sold their home gateway platform, which makes the semiconductors that go in-home modems and routers. Intel once dominated this business with their CPUs, but given the struggles at the core business, Home Gateway was an afterthought. The results were astounding.
Revenue increased 86%, and EPS increased 205% in 2021 over 2020. Part of that is the revenue comparability of an acquisition, but even the following year, Maxlinear grew 25%. In the high-flying semiconductor industry, Maxlinear few the highest. I put Nvidia on here to put things into context.
The logic was simple. Intel’s home gateway business had a huge share in the logic of a home gateway, and Maxlinear had a share of RF other devices that sold into the home gateway. They could sell into the large install base and raise ASPs. Home run.
In 2021 this is what they had to say about the size of their WiFi business.
if you go back to 2020, we did kind of $25 million to $30 million of revenue from the WiFi business. Now that was early days of WiFi 6. We really didn't -- I mean we had a product in WiFi 5, but it wasn't very significant. We relate to market. So, winning in WiFi 6 was super critical, getting there early. So, we won a lot of shares. That business doubled the following year to north of $50 million. This year, we're expected to do well north of $100 million and then have talked about being north of $200 million next year. It's great that we're getting the attach, meaning we're selling alongside of our SSE and now we've got penetration within the fiber market.
They ended up with this WiFi business hitting around 200 million in run-rate revenue, an astounding almost 8x in revenue from 2020 to 2023. Higher ASPs drove part of that, but part of that was by winning a larger share of attach rate to the Intel home gateway business.
Maxlinear pulled off the perfect acquisition. So as empire builders do, they decided it was time to build an empire. They did this by turning their eyes to Silicon Motion, or SIMO.
Silicon Motion Technology Rationale and Saga
Maxlinear is rolling off one of the best deals in a public company’s life, and now they are talking about how they want to do more deals and then they buy SIMO. I’ve written about it before, but the logic was that Maxlinear could become a Marvell-like company by acquiring Marvell’s worst business, memory controllers.
The logic almost makes sense. But it’s coming to the Marvell conclusion backward, with the business that Marvell tried to diversify away from. But Maxlinear had the possible PAM4 business, WiFi, and other baseband RF; it almost made sense to try to compete with Marvell head-to-head as a broad networking peer. This was likely to try to achieve the Marvell valuation.
The problem is that the deal was consummated in the good times, and memory is the most volatile and impacted. The deal was announced in May 2022, and the transformative deal was supposed to be the next homerun. They bought SIMO on peak earnings. EPS has moved from ~8.00 to 3.15, and the company that should have been very accretive is dilutive.
But what’s worse for the SIMO shareholders, SAMR held up the deal process, and the deal spread was consistently 60-90%. Part of that was because China hadn’t, and we expected it would not approve a semiconductor deal, but part of that was that earnings collapsed, so it seemed less likely that the deal would even happen.
Well, SAMR, out of nowhere, approved the deal. I think they approved it because it would screw over Maxlinear, given the new price. I wrote about a rumor in March, and the rumor was the deal was approved. The rumor turned out to be correct! So where do Maxlinear and SIMO go from here?
Deal Implications
First, I am mostly using this tweet thread as source material. Please read it. I am not an arbitrage expert, but I have some opinions regardless. First, it would be horrible if Maxlinear bought SIMO; it just financially does not make any sense anymore, especially how it’s structured. The cash has a worse return when they struck the deal, and the stock consideration was even more expensive; it’s a lose-lose deal.
SIMO shares moved quickly on the news, while MXL moved quickly down that they would have essentially the biggest dilution event ever. They are issuing 1 dollar bills to buy 80 cents, given the relative valuation of the companies.
Then Maxlinear pulled the legal card out and said, nevermind we are breaking the deal. Their rationale is mostly laid out in the technical details of the merger. There are three prongs here.
(ii) Silicon Motion has suffered a Material Adverse Effect (MAE)that is continuing
(iii) Silicon Motion is in material breach of representations, warranties, covenants, and agreements in the Merger Agreement that give rise to the right of the Company to terminate, and
(iv) in any event, the First Extended Outside Date has passed and was not automatically extended because certain conditions in Article 6 of the Merger Agreement were not satisfied or waived as of May 5, 2023.
I am no lawyer, but I will at least walk through each item they bring up in the merger document. The MAE is probably the hardest to prove, and given how the Twitter MAE argument didn’t work, it will be hard to prove MAE. According to the document, it will be governed by Delaware law.
that provisions related to the definition or occurrence of a Company Material Adverse Effect or a Parent Material Adverse Effect will be governed by, and construed in accordance with, the law of the state of Delaware.
So MAE is going to be a hard road to prove. But one Maxlinear will have to go on, given how horrible the deal is for Maxlinear.
What about the material breach? This is the language I find interesting. It’s self-referential.
material breach of any representation, warranty, covenant or agreement set forth in this Agreement has principally caused, or resulted in, the Effective Time not occurring prior to the Outside Date; provided, further, that if on the Original Outside Date, all of the conditions in Article 6 other than Sections 6.1(c) (solely with respect to any Regulatory Law or any Order issued pursuant to or in respect of any Regulatory Law) or 6.1(d) have been satisfied (other than conditions that by their nature can only be satisfied on the Closing Date), or have been waived by Parent and Merger Sub and the Company, as applicable, then the Outside Date shall automatically be extended to May 5, 2023
So that brings me to the First Extended Outside date. Pretty much there are automatic extensions in the outside date. Below are the dates and extension timelines.
First Extension: If certain conditions in Article 6 are met or waived on the Original Outside Date, except for specifics regarding Regulatory Law or Orders, the Outside Date is extended to May 5, 2023 (First Extended Outside Date).
Second Extension: If the same conditions as in the first extension are met or waived on the First Extended Outside Date, the Outside Date is extended to August 7, 2023 (Second Extended Outside Date).
Ten Business Days Extension: If all conditions to Closing are met before the Outside Date, then the applicable Outside Date will be extended by ten (10) Business Days.
Two Business Days Extension: The Outside Date is also extended to two (2) Business Days after the end of any pending Matching Period or Marketing Period.
Let’s evaluate certain conditions; oh, it’s an MAE! So it becomes circular, but it looks like there is a second extension date, so Maxlinear is not out of the woods for the deal period.
I cannot find what warranties and covenants are in the merger document. But realistically, the MAE and the agreements are the prongs Maxlinear will be used to argue in court.
And that’s exactly what SIMO is going to take advantage of. They are going to enforce the merger. SIMO probably believes they have a case to force the merger through the MAE.
So where does that leave us today? The trial will take place in Singapore but be under Cayman law. And the definition of MAE will be using Delaware Law. Talk about confusing.
I don’t know enough about MAE or merger law to know who wins, but the asymmetric upside favors SIMO. They have the most to gain, while Maxlinear has the most to lose.
Given how hard MAE is to enforce, I expect this to be an intense saga. The lawyers likely will be the biggest winners. I expect this to be a protracted legal battle.
Psst if you liked this analysis, please consider sharing and subscribing. Support from readers keeps me in the game, and I appreciate that. I couldn’t do what I do without all of you. Okay, onwards to Vicor.
Vicor and Monolithic Power
Anyways I also wanted to spend this time to talk about Vicor briefly again. I decided to boost the Vicor short seller report; I regret it. Yes, they are not in the H100, and that’s clear. But they insist that they will be in a new product by Nvidia. Shares rallied 40% since this was a crowded short.
I think I have an idea of what is going on here now. Let’s discuss.
They beat the hell out of earnings, but the backlog also collapsed, which doesn’t seem to make sense if they have a new customer win. Not exactly a comforting outcome.
Vicor reports Q2 EPS $0.38 vs FactSet $0.23
Reports Q2:
Revenue $106.7M vs. FactSet $98.0M
Backlog for Q2 ended 30-Jun-23 totaled $217.3M, a 47.0% decrease from $410.0M for the corresponding period a year ago, and 19.9% sequential decrease from $271.3M at the end of Q1 of 2023.
They insist that they expect revenue and backlog to increase soon. And it’s a bit of a messy situation. Vicor’s plan? Sue their competitors in the ITC so that the large GPU customer has to use their products in the US! Genius!
We also expect a sequential increase in operating expenses, primarily as a result of funding the legal work associated with cases filed earlier this month at the International Trade Commission and in federal court in the Eastern District of Texas against foreign manufacturers of power modules and computing systems infringing Vicor patents covering non-isolated bus converters, NBMs.
Legal work associated with these cases and related legal expenses are expected to grow substantially over the next year. Legal expenses are, however, less than the royalties paid to Vicor by licensees of our patents. In our ITC case, we are seeking an exclusion order precluding importation into the United States of power modules, servers or AI cards that infringe our patents
At the same time, they are ramping up a large fab that is coming online in September. They expect that their fourth-generation lateral-vertical PDN will ramp and provide a better performance as their fab ramps, leading to a design win in the key GPU customer. When asked if it was a new customer, they say its for an existing customer.
Do you think that's on -- it's new generation for the existing customer. And it's a chip-set that can be deployed either in a lateral PDN, which is substantially handicapped from a power system perspective to the point that it limits power delivery, power capability, processor performance, in that it gives rise to large losses within the copper of the substrate to the GPU and then the towers. It gives rise to further losses within the [ silicon ] itself, owing to the limitations of a lot of our delivery applied at the 1,000-amp level.
With 4G chip-set, we can enable a lot of solution with the same handicaps or with a vertical element using the same chip-set, a lateral-vertical solution, which is unique, highly differentiated in that improved system efficiency by about 10% and removes a number of limit actions relating to processor performance.
They say this customer will be deployed with lateral and then swapped to lateral-vertical cards over time. Customers rarely make design changes in their production chips. It’s weirder because we have seen the boards of Nvidia’s products and haven’t seen the characteristic gold power delivery.
I find the odder thing is that their backlog hasn’t increased to account for what should be a material ramp-up in revenue. They say that next quarter they will ramp backlog in Q3 and Q4.
No, I think that's exactly right. We have existing backlog and our plan is to obviously begin to ramp with that particular backlog. And then in Q3, Q4, we'll getting increased bookings for the follow-on 2024.
They are saying they can be multi-source for the AI customer (Nvidia), which seems odd given what we know about chip ramps and product launches. They think they can fit their product in the same floor plan as the existing multiphase solution (MPWR).
So the lateral solution using the same ChiP-set is not the "drop-in" because with different components, the layout itself is different, but it shares a common lateral PDN, given fundamentally the floor planning that had been done, which floor planning was, again, predicated on a multi-source, multiphase solution. So we're first in effect within the same general floor plan, fitting in a solution which is better in many respects, mature and nice, better performance, in general, but handicapped by the lateral PDN. So that handicap is a common denominator limitation that can be overcome.
They are also a bit better at saying that Multi-phase power isn’t going away and are saying that scaling things are hard, and Vicor has the better product.
It's not the end of the line for multiphase, right? It's -- multiphase has been around for a long time, and it will not suddenly die. There is a lot of case, but you have me say this before, it can't keep up with the performance of a factorized power system, particularly leveraging our 5G components with a 3x step up in condensity and much better performance all around.
Nor can it keep up with the industry demands with respect to escalating [ cost ] requirements producing voltages for all the competing AI systems that are under development are going to be brought to market over the next few years. I see a fundamental disconnect between what the systems are going to need and what multisource, multiphase can support, particularly as we get into full vertical power delivery type of systems, i.e. fast lateral, even fast lateral-vertical.
Let’s turn to Monolithic Power and see what they are saying about this. They had a very weak earnings season because AI is ramping so aggressively. They are saying they expect some redesigns that allow competition in the market.
I think in most scenarios, and this particular customer is representative, is they want to take a leadership position through innovation, and they found us an equal partner for that. But it's in everybody's best interest that it be a competitive, not a single source. And so yes, we'd always anticipated that there would be redesigns that allow competition into the market.
Redesigns sound very good for Vicor. Meanwhile, Monolithic is saying we are going to be in the new products; MPS is even better than the last generations, and we believe we are better than the competitor’s solution.
Our visibility is, okay, probably is well documented. Okay. If you listen to all these major AIs, probably you'll see MPS. Okay? And you will see the MPS in the next few quarters of potential. And -- but I can talk to you about the technical issues. Okay? I think that the MPS is -- so far the solutions is far better than our competitor. So we -- for the next generation of AI processor, we're not working on it now. Okay. We're working a year ago, even 18 months ago.
So this brings me to ask, what exactly is this new product that Vicor thinks they can win, and the product that MPWR thinks they will at least have a second source competitor? I have some speculation on product announcement and timing that I am, of course, going to put behind a paywall.