Nvidia’s Q1/2016 analyst call had three interesting themes that caught SemiAccurate’s ear, Icera modems, OEM business, and the Qualcomm/Samsung IP suit. All three are confirmation of news we brought to you months or years ago, and none of the news is what Nvidia claims it is..
Lets start out with Icera, the radio/modem company Nvidia bought a few years ago. Progress was glacial and the first LTE radio from Nvidia surfaced in late 2012. To the best of our knowledge it was single mode only, didn’t work very well, and was featured in only one product, the Asus VivoTab 600 RT LTE. There may have been a few other minor or in-house designs that featured an Icera LTE modem but none had anything close to the reach of the Asus 600. That device did not set sales records.
The other product of note with an Icera modem in it was the Tegra 4i which, despite its name, was a Tegra 3 with a few bits of a Tegra 4 pulled in and a modem added. When it debuted the 4i was a generation behind in cores, ARM A9s when the world was well into A15 based products, and an incompatible GPU. The modem was similarly not up to par and there were no third-party devices. At its debut during CES 2013, CEO Jen-Hsun Huang held up a reference design rather than a partner device. That is industry shorthand for being dead in the water, no partner devices means no partners, and we explained why this state of affairs was so dire in the link above.
During MWC 2013, there was one “third-party” Tegra 4i device at the Nvidia booth, a second tier French wireless provider without their own network if memory serves. Nvidia representatives insisted the device, a clunky plastic brick that was a dead ringer for the reference platform from CES, was designed by the telcom itself. No one SemiAccurate talked to believed that but some may have shipped to customers. That too was the only non-Nvidia Tegra 4i device that SemiAccurate was aware of.
This is the long way of saying that Nvidia had no customers for their modems. You can ascribe that to scheduling, price, quality of the supplied parts, bias, sunspots, or anything else you want. What you can’t say is that there were enough sold to cover the design and certification costs of any device they were in, every product bearing an Nvidia modem failed badly in the market.
And last week Nvidia officially pulled the plug on Icera with a public announcement that they would be winding down the group later this year if a buyer was not found for the division or its assets. Some people reacted to this with shock but not SemiAccurate readers. Why? Because we told you that Nvidia had unquestionably pulled the plug on their phone business in December of 2013. To be a bit more exact, at the time we said, “Even if insiders weren’t whispering to SemiAccurate that phones are effectively dead at Nvidia, there is only one conclusion you can draw. Nvidia is out of phones but don’t look for them to admit it for a long time if at all, but it is unquestionably over.” Now almost 18 months later, Nvidia admitted to what SemiAccurate told you in 2013. Told ya.
The next theme is a little more anecdotal and it surrounds Nvidia’s talk about their OEM business. If you listen to the Q1/2016 analyst call, you will hear an interesting theme repeated over and over about their OEM business. The company claims that their OEM income is only a tiny part of their overall revenue, something SemiAccurate agrees with completely. A bit over a year ago this wasn’t true, Nvidia was overwhelmingly dominant in OEM design wins for GPUs from Apple to no name players.
What changed over the past year or so? Do you think that Nvidia suing or threatening nearly 100% of their OEM customer base with a patent trolling lawsuit had anything to do with losing business there? Don’t forget SemiAccurate told you about the baseless threats and many of the targets over well over a year before the company sued Qualcomm and Samsung. Samsung does make PCs you know. So does Apple and all the others we told you about months before things went public.
Threatening your customers with patent litigation or actually suing them is a sure way to lose design wins for the next several decades. Given who has not been approached with a ‘generous offer’ surrounding patents, it is no wonder Nvidia’s OEM business is effectively dead. Can you think of one PC maker who does not sell ARM based phones or tablets at the moment? Which category is growing for OEMs? Don’t look for Nvidia’s OEM business to grow significantly again, ever, regardless of product quality. Similarly don’t look for Nvidia to actually explain why their OEM business is now a minor portion of their sales. You can just think of it as lowering silicon volume tremendously or driving fixed costs like masks up tremendously.
The last part is the most interesting and it involves Nvidia’s full-on blitz during the Q1/2016 call about the value of their lawsuits against Samsung and Qualcomm. The company said they were going to spend high tens of millions of dollars in FY2016 on legal expenses, something SemiAccurate believes is a fair number. They also claimed they were expecting a fat return on their investment, a claim we don’t have nearly as much faith in. Why? Track records and the patents.
On the track record side, Nvidia has been approaching OEMs since late 2012 according to our sources. If you listen to the official explanations of the ‘facts’ behind the case, they claim to have a valuable IP portfolio which others are infringing. The value of the patents has two sides to it, what the company really feels they are worth, something SemiAccurate will cover in a future article, and what others think they are worth. Both can have starkly differing values attached, and we know what Nvidia publicly thinks they are valued at. Both of these are also distinct from the public stance of course.
What potential licensees see their value at is diametrically opposed to Nvidia’s public stance. Why do we say that? After 2-3 years of ‘negotiations’ and bluster, Nvidia has not signed a single licensee. We told you about the count and costs, updated it with the new numbers in the second and third rounds of offers, gave full licensee ‘counts’, and explained why it still has no chance to succeed. (Note: If Nvidia signs just one more licensee, they will be at a grand total of one.)
In patent litigation like this there is a tried and true path that companies take to establish the value of their patents and get more companies to license instead of fighting. The short version is that you pick on a tiny 3rd tier player in the market and offer them a sweetheart deal, sometimes even one that is to their financial benefit. In exchange they ‘license’ your tech and publicly proclaim how wonderful it is during PR events. You then work your way up the ladder to bigger 3rd and then 2nd tier players gradually cranking up the licensing fees and more importantly establishing a precedent.
By the time you get to the big fish, in this case players like Samsung, Qualcomm, and Apple, you have several precedents establishing the ‘validity’ and ‘value’ of your claims. It may be total BS but going into court with 5+ successful licensees looks a lot better than zero even if you had to pay for those five. This is a time honored patent trolling methodology and sadly it does seem to work, or at least work in the US. In theory this will prevent the more combative targets from fighting, or at least make them more receptive to negotiating before hand.
So Nvidia has had 2+ years to buy a few minnows before knocking on the doors of the sharks. So far the minnows that lined up beside Nvidia, sweetheart deal or no, is zero. Licensing costs have plummeted to a fraction of what they were, everything has come down, and still no licensees, not even a minor tier nine nobody with nothing to lose. Does that tell you anything about the ‘value’ of the Nvidia patents? The last angle to this is again out of the scope of this article but you should be asking yourself what Nvidia actually values their patents at, public and PR bluster aside. Again not the public bravado, but their real internal valuation.
So in the end, Nvidia’s Q2/2016 analyst call was very interesting but not for the reasons they probably wanted it to be. Icera is dead but we directly told you that 18 months ago, the only outstanding question is why Nvidia waited this long to announce it. Their OEM business is a fraction of what it was 1-2 years ago mainly because they are threatening most if not all of said OEMs with patents suits like we again told you about years ago. Lastly the patent bluster is nothing more than a sham likely meant to prop up the stock price with visions of windfalls balanced by low odds of success. If there was any merit to the suits, Nvidia personnel would be awash in minnows every time they went on stage. They aren’t. Do you believe the hype?S|A
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018