This is a longish draft that I wrote before leaving on a trip out of the country. It needs work but I thought I woud post it as is to try do two things.
First is to signal that I am taking a time out here.
Second is to give my thoughts about why the Samsung deal ought to be unique, and why the market seems not to believe that that will be the case.
We are still missing key details about the January 19 Rambus/Samsung settlement. But on the plus side, there are obviously some good things here for Rambus. These include the cash, the headlines (“$900 million victory”), the hopes for a “longer-term relationship,” the end to an almost-four-year dry spell in Rambus licensing, and the Rambus licensing team’s proclaiming (oddly) that it will not now “rest on its laurels.”
On the other hand, RMBS is still (even after two buoyant conference calls, a strong performance on CNBC, and extremely good intervening litigation news from the ITC) trading relatively low. It is about even from a week before the deal, and at about half the $48 price RMBS achieved in early 2006. So the stock market apparently has its doubts. These include the key question of whether (despite vigorous denials in the last quarterly conference call) similar, extremely favorable terms will be extended to other long-stalled deals in the Rambus licensing pipeline (described in early 2006 as “capacity constrained” and “literally in the dozens”).
A. The Rambus Trade-off and Worldview Behind the Samsung Deal.
To me, the key impact of this deal depends, frankly, on whether it was a template for others (the CEO vigorously denies this) and whether Rambus was wise in trading “litigation upside” (against its only deep-pocketed antitrust defendant) for “leadership products” and “longer–term relationship” upside.
Let’s accept the CEO’s vigorous denials and focus on the trade-off.
Two press quotes from the current Rambus GC, Tom Lavelle, make clear that this litigation vs. longer term relationship trade-off was was exactly the trade-off that Rambus intended to make:
There's two kinds of settlements . . . . There's the settlement where the one side is getting rid of the litigation, and there's another kind where the parties create a longer-term relationship — and that's what this is.
He also said:
It was a really good day because this is the business we want to be in. We want to be in the business of licensing — crazy as this sounds, we don't want to be in the business of litigating.
In the Rambus quarterly earnings release last Thursday, the Rambus CEO went even further, stating:
The agreement signed last week with Samsung Electronics is transformational for Rambus and will accelerate the market adoption of our patented innovations and leadership products.
As an aside, before going further, I should share some of my own bias and a bit of history.
First, bias: I have had, as my posts here somewhat reflect, my own issues (on several levels) with current Rambus management. I also have a background that includes substantial litigation experience. With this background, I have long seen the interrelationship between the “licensing business” and the “litigation business.” They are linked. The former requires the later. It is tough to sell ideas. They are easy stolen or modified. And it is tough to compete (at any reasonable rate of return) with the in-house engineering teams of your customers. I used to do a lot of in-bound licensing (at another job) for a major manufacturer. Our perpetual goal? Remove those license fees from our COGS (“cost of goods sold”) as soon as possible.
From this perspective, I worry about how Rambus now overemphasizes what some at the company consider the “friendly” licensing of Rambus’ “leadership products.” http://www.investorvillage.com/smbd.asp?mb=3666&mn=409398&pt=msg&mid=7676613
. I worry that Rambus management risks more than naiveté, but also credibility problems, misallocated resources and other bad decisions. Without taking anything from the remarkable engineering skills at Rambus, my concern is that Rambus’s focus on “leadership products” and on (what it at least perceives to be) “longer-term relationships” outstrips any reasonable measure of their track record, relative current contribution or (given current circumstances) relative future promise.
Second, a small bit of history: In mid-2002 or so I attended a panel of the National Science Foundation. The topic was patent law and its support of useful capital formation. The main speaker that day was Rambus’ then–Chairman of the Board, who is also one of Rambus’ two founding venture capitalists (and for whom I continue to have great respect). I will never forget the response of one member of the NSF panel, a senior and very well–regarded federal judge, Charles Legge, of the Northern District of California. He leaned forward in his chair, interrupting our Chairman’s description of hoped-for technology leadership and friendly licensing and said simply: “So you are in the litigation business.”
B. What Rambus Got in Its Samsung Deal.
Beyond the headlines, what Rambus got from Samsung falls into four buckets: two upfront cash payments totaling $200 million, plus a $200 million from the sale to Samsung of newly issued Rambus stock (about 9 million shares at about $21 a share, with half of the shares half covered by a put back to Rambus at that same price), plus five years of quarterly payments to Rambus of ($25 million a quarter, adjustable after the first year to something in a $10 million to $40 million range), plus an MOU.
1. The Upfront Cash. The upfront cash disappointed many here, falling far short of what would have been required for the past under past Samsung license rates, EC rates or the court-ordered Hynix rates. http://www.investorvillage.com/smbd.asp?mb=3666&mn=477382&pt=msg&mid=8464300
Among those voicing strong doubts was the Rambus shareholder who has devoted what appears to be the most time of any attending court hearings and reading the record: http://www.investorvillage.com/smbd.asp?mb=3666&mn=480409&pt=msg&mid=8480680
As to the upfront payments and whether they adequately cover past infringement, it is useful to focus on the correct metric. The EC rates are a tempting possibility. But by their terms the EC rates are not applicable to past infringement.
So the better metric for what was owed from past DRAM and controller infringement is, in my view, what Samsung previously agreed to. (I recall that court documents and hearings made thesepublic, putting these rates on the order of 3.5% for worldwide DDR sales and about 5% for worldwide DDR controller sales). Another alternative is what a court would award. What a court would award is, of course, unclear. But we can project that to be 4.5% on U.S. DRAM sales using the Hynix patent trial numbers, which only addressed DRAM – not controllers – and which the judge reduced from a jury award of about 10%.
Using any of these metrics, it is very clear that Samsung got an extraordinary deal with respect to its past infringement. That deal is likely much better than I can quantify since I have no data easily at hand on Samsung systems (e.g. TVs, cell phones, laptops, camera) or flash memory shipments.
2. The Ongoing Quarterly Payments. As to the ongoing $25 million per quarter payment stream, the EC is a useful metric there. It appears that the $25 million from Samsung per quarter is (however it is adjusted in the future per the deal’s terms) significantly low. It appears to be significantly less than what Rambus would have received under the patent royalty rates that Rambus very recently negotiated with the EC for DRAMs and DRAM controllers (even putting aside the coverage that is now in the deal for other Samsung products like cell phones, laptops, TVs and flash memory).
The DRAM and controller math is fairly straightforward. If Samsung is 35% or more of a $30 Billion DRAM market, the EC-approved rates (which are 1.5% on DRAM and 2.65% on controllers, dropping to 1% and 2% when key Rambus patents expire in April 2010) would have generated royalties to Rambus running well over $100 million a year. Beyond this, significantly more would have been owed from Samsung to Rambus under the EC-approved rates on DRAM controllers. And more still would be due (though not covered by the EC rates) on flash memory and other Samsung products.
Bottom line, it looks like Rambus took a very significant haircut off of its already-deeply discounted EC rates (on which the ink is barely dry). My guess is that that haircut -- looking only at DRAMs (easy to quantify) and controllers (harder to quantify) -- amounted to at least 50%. But it could be much higher in reality since I do not have data on Samsung system or flash shipments.
Four other aspects of the future impact of the Samsung deal (versus a possible deal under the EC rates) also are likely to come into play:
* There would have been, under the EC license, no surrender (which we now see, essentially for nothing) of Rambus’ antitrust claims against Samsung.
* There also would not have been (under the EC plan) any “perpetual, fully paid up” license to Samsung at the end of the five years. So renewal after five years under the EC approach would, presumably, have been much easier for Rambus than it will be now under the settlement. (Renewal prospects are important in deciding whether and to what extent to apply a P/E ratio to the Samsung stream of payments.)
* Unlike the EC plan, the Samsung settlement does contractually commit Samsung to continue to pay after the first year. Under the EC plan, any licensee can cancel after one year of payments. Still, if Samsung does breach its multi-year commitment and stop paying, Samsung will, under the settlement, unlike under the EC scheme, apparently face sharply capped exposure (contract damages only, it appears). It is unclear. Will Samsung ever thus refuse to pay Rambus for what it owes during the contract term? Maybe. It did just that in 2001, when (just before a scheduled Rambus earnings conference call) Samsung withheld one key quarterly patent payment in order (apparently) to pressure Rambus for a better deal (which it got).
* The effective per-unit rates to be paid to Rambus over the next five years may go lower still relative to the EC rates. This may happen (depending in part on the terms of a largely unspecified adjustment mechanism) if prices go up or Samsung picks up market share. And after this deal one can expect Samsung to improve its market share – both in memory (where the unspecified, probably inadequate adjustment in quarterly payments applies) and in controllers and other products (where there is apparently no adjustment). Samsung can be expected to gain sales (a) by using more Rambus technology, and/or (b) because Samsung customers now need not worry about being sued for using unlicensed DRAM or controllers, and/or (c) by having Hynix and/or Micron go bankrupt (or just risk bankruptcy) in the Rambus antitrust case.
3. The MOU, Stock Sale and “Longer-Term Relationship.” One thing is clear. As reflected in the numbers and in its handling of its potent antitrust claims, Rambus management obviously ascribes tremendous value to the MOU in its settlement – as well as to the “longer-term relationship” and work on “leadership products” that that MOU signifies.
The Rambus CEO acknowledge this in the last conference call – citing the frustrations he faces given his underutilized engineering staff when he has no manufacturing partner. Rambus management (which mostly comes from one of the few great manufacturing companies in Silicon Valley, Intel) must see the importance of that MOU as, in fact, overwhelming – certainly (given what they gave up) to be valued in the hundreds of millions of dollars, perhaps in the billions.
There are at least two problems here that I see.
First, to my knowledge Rambus has never (in its twenty year history) publicly claimed to have made material profits from its “leadership products.” They have never claimed that aggregate royalties exceeded development costs on any of them.
Second, there is the question of just what it is that an MOU gets you. An MOU is (literally) a “Memorandum of Understanding.” It is not an agreement and is almost never enforceable. Similar past “understandings”, friendships, and hopes of cooperation had added little or nothing to Rambus revenues or the Rambus bottom line. Cadence (July, 2004), Infineon (March, 2005), AMD (December, 2005), and Intel (November, 2007) all come to mind. Perhaps Rambus now has reason to believe that it now has a more reliable partner. But history (with Samsung itself, and with other “longer-term relationships”) calls this into question.
It might be (we do not know for sure) that the MOU with Samsung has – contrary to convention -- some teeth in it, making it enforceable. That seems unlikely, however, particularly given that Rambuis has (now given many opportunities) said nothing along these lines. Also, per the 8-K, Rambus’ releases of its litigation claims against Samsung become effective immediately upon the completion of Samsung’s $200 million in upfront cash payments (which will be made this year.) Those Rambus litigation claims, once released, will thus provide no leverage for further compliance with the MOU – even if (as seems unlikely) it is enforceable.
It also seems highly unlikely that Samsung will be highly incentivized (as some hope) by its purchase of $200 million worth of Rambus stock. This deal term (while diluting Rambus shareholders significantly and effectively providing a voting block to managment) feels like it is (to a significant extent) window dressing – helping Rambus to reach the claimed “$900 million victory.” Samsung can sell the stock in eighteen months. Half the stock is covered by a put, with which Samsung can (in eighteen months) recover the same price it paid Rambus. And, in any event, the downside risk or upside opportunity for Samsung in Rambus stock feels, frankly, like a negligible incentive to a company that reports 2008 shareholder equity of $90 billion, with over $3 billion dollars in reported profits last quarter.
C. What Rambus Gave Up to Samsung.
One thing that Rambus gave up (beyond its stock) is (I fear) the notion that it does not pay for prospective licensees to litigate. Rambus has given this kind of a discount for past and ongoing infringement to a licensee/litigant before (e.g. to Samsung itself in 2001 and in the Infineon deal in 2005) but never before to this extent, and never before when it so obviously (from all public information) had the upper hand in its court cases.
A prompt antitrust trial date may have been another thing Rambus gave up or put at risk in its Samsung deal. The antitrust case was on the edge of jury selection from all appearances. Boxes of evidence had been moved into the courtroom. The January 19 Samsung settlement was among the reasons Judge Kramer relied upon on January 21 when he continued the trial and declined to set a new date.
Beyond these two (perhaps unforeseen) consequences, what Rambus gave up in its deal is the future use of its IP by Samsung (with some major open questions) and its legal claims against Samsung.
1. Open IP issues. As to the future use of Rambus IP by Samsung, the exact parameters of that are unclear in several respects.
What, exactly, is covered by the “fully-paid-up, perpetual license”? Is it a capture license (covering all patents that Rambus has or applies for in the five year term and before)? That will make renewal in five years even more difficult.
Also, the perpetual license is said to cover (per the original press release) “certain current DRAM products.” What does that mean? Per the 8K, we now know a little more, i.e. that it covers essentially all the DRAM types that Samsung currently ships. The perpetual license may or may not (based on the 8K) exclude a perpetual license for DRAM controllers. It is unclear: Does the term “DRAM products” include “DRAM controllers?” It might. Such an outcome would also add to Rambus’ renewal woes.
In any event, does the “fully-paid, up perpetual license” also effectively exclude DRAM standards or designs developed during the course of the five year period? It seems to. Rambus refers to the perpetual licensing of “current” DRAM products. But does this limit work? This was a major issue in the past. Rambus tried to assert such a limit in its 2000 patent license to Samsung for SDRAM and DDR. The carefully crafted limiting language (in two parts of the 2000 license) was clear. But Samsung essentially ignored it, claiming that its past DDR2 shipments (now forgiven as part of the settlement) were also licensed.
One final area of uncertainty: to what degree is there protection here against the kind of error that Rambus made with AMD, the last major deal that structured by this licensing team?
In that last deal, Rambus did nothing to protect itself by requiring more royalties (beyond the agreed-upon fixed amount) if (as occurred) AMD acquired ATI (a company with billions of dollars a year in significant additional infringing sales). See http://investorvillage.com/smbd.asp?mb=3666&mn=410600&pt=msg&mid=7695690
This time around there is, we are told, some protection for Rambus from an AMD-like outcome. At the conference call, we learned a little about this. We were told by the head of the licensing team, Sharon Holt, that there was some “variance” in the quarterly payments to reflect “Samsung business results.” Now, with the 8-K, we know a bit more. During years 2 through 5 the quarterly payments can vary, we now know per the 8-K, from $10 to 40 million a quarter. This variance will depend (how is unclear) on varying Samsung DRAM sales, with other adjustments (unspecified in the 8K or in public statements so far) for possible mergers or acquisitions by Samsung. There seems to be no adjustment contemplated, however, for improved Samsung sales (through organic growth at least) of infringing controllers or infringing systems. So at this point (absent new “leadership product” design wins) we cannot feel too good about all those new Samsung laptops, smart phones and plasma TVs at Best Buy over the next five years.
2. The Surrendered Rambus Legal Claims. And then, of course, there are the legal claims (“the litigation business”) that Rambus gave up. They were certainly significant.
The Rambus patent claims were always strong given past Samsung licenses (at very high rates relative to the settlement) and the Hynix trial results. And the patent claims have recently been looking even better. We have seen significant recent progress with the ITC, the promised prompt issuance by Judge Whyte of his Samsung spoliation opinion, and the likely outcome of the CAFC Hynix/Micron appeal.
The Rambus antitrust claims against Samsung seemed (from public information) even stronger than the patent claims. They were backed by, inter alia., admissible criminal convictions by Samsung and Samsung employees, convictions of its co-conspirators, and “dynamite” emails implicating Samsung. The Rambus damages expert was (and still is) going to be able (as the trial Judge has now ruled) to present the jury with an opinion (based on actual past worldwide sales and actual agreed-upon royalty rates for RDRAM and RDRAM controllers) putting the Rambus damages (before automatic trebling) at over $4 billion. It is surprising to see these antitrust claims (as asserted against the only deep pocket defendant) surrendered for what (beyond the “leadership products” MOU) appears to have been essentially nothing.
Of course, Rambus could, under the right circumstances, have afforded to settle with one or two of the three antitrust defendants for a relatively low amount. That is because, under the law of joint and several liability, the remaining defendants are responsible for the balance of all damages (with no allocation by market share – as Rambus recognized in its last conference call – and with any applicable settlement amount from Samsung deducted only after trebling).
There is a hitch to this argument, however. How valuable (ultimately) are the claims against the remaining defendants? They are apparently not strong financially. Both have hemorrhaged cash for years. Hynix stock is dropping. Its major owners cannot find a buyer for their block of stock and it claimed last year (to Judge Whyte after losing its first Rambus case) that it had to put a lien on one South Korean factory to secure its judgment owed to Rambus. What if Hynix and Micron cannot pay the trebled antitrust damages that Rambus has now claimed and go bankrupt? The Rambus GC was asked about this in the conference call announcing the Samsung deal. He said that that would be a nice problem to have, but he had not yet looked at it and did not know the answer. http://www.investorvillage.com/smbd.asp?mb=3666&mn=478696&pt=msg&mid=8470414
D. My Best Wishes.
I have been wrong (in varying ways) about Rambus and Rambus management in the past. It is likely I will be wrong again – maybe including about the above concerns. Perhaps Samsung is a one-off deal. Maybe this MOU will bear fruit. Time will tell. For now, I should say that I have been very grateful for the support of many here and I look forward to some time away from all things Rambus.
(Written solely on my own behalf)
e-mail to a friend printer-friendly