It's an investment thesis so simple it could be written in a Google (GOOG) doc. Short or sell Nvidia (NVDA). The upcoming Tegra 4 from Nvidia is not going to be powering the next version of Google's Nexus 7 tablet. And, frankly, it's really no surprise because, despite the hyperventilating of Nvidia's PR staff, the Tegra 4 is the wrong SoC for pretty much anything below a 10" or bigger specialty tablet, and even then it is iffy. Qualcomm (QCOM) is, a Snapdragon 600, and likely would have had won both the original Nexus 7 and the Microsoft (MSFT) Surface RT last year had TMSC's 28 nm process not been having yield trouble, delaying the Snapdragon S4 rollout.
Now that is behind them and Qualcomm can and is shipping seemingly billions of its Snapdragons in all variants, there is little to no room for the Tegra 4 in the market place. Even if the Tegra 4 is faster at some level than the best Qualcomm has to offer, it is yet again another iteration of the Intel (INTC) vs. AMD (AMD) debate of building a faster path to a dead end. Who cares if you can eke out slightly better SunSpider scores if your SoC uses twice the power and costs four times as much to make?
This question is the bane of enthusiasts and most of the tech press - cool technology does not equate with good product. And from where I sit, the Tegra 4 is the blocked punt in the CPU equivalent of a three and out in football.
Look at the reasons cited by Google as to why it chose the Snapdragon over the Tegra 4 - cost, integrated LTE and, most importantly, availability. Simply put, in every way, Qualcomm's SoCs are simply better value and more reliably available than anything Nvidia has or will have anytime soon. Nvidia doesn't have TMSC to help prop them up for another four quarters.
Simply put, I have to agree with Charlie over at Semiaccurate who has been pounding the table for a long time saying that the 4+1 core structure started with the Tegra 3 was not only a poor design choice but it was also the third in a series of under-delivered promises made to phone and tablet OEMs. Quad core SoCs running Android in 2012 were a stupid marketing ploy. And then they made things worse by adding another core to manage, driving up complexity to "plaid."
The 4+1 core is vastly more complicated than a properly constructed dual core or even vanilla quad-core processor. Complexity adds to cost and cost equals no uptake by OEMs in phones because there is no room for being outside of design spec. Even the mighty Samsung (SSNLF.PK) had to turn to Qualcomm to power some of the latest Galaxy S4 line because big.LITTLE is hard to pull off and the Exynos Octa is not ready.
So, not only did Nvidia not learn from the Tegra 3 -- taking those 10 million pity sales thanks to a screwed up 28nm marketplace as a sign of a job well done - it repeated it because nothing says success like failure. Since then, it has tried to tell everyone it's something it isn't, namely capable of competing with anything else on the market. And when the customer that handed you 6 million of those 10 million in sales bolts for your rival, someone has to be willing to finally yell, "Iceberg dead ahead."
And it's not only Google that has shunned Nvidia here. So has Asus and other OEMs that bought into the hype of the Tegra 3 only to have it fail to live up to expectations and cost them money. In order to even get the Nexus 7 and Surface RT contracts, Nvidia had to build them the reference platform to prove it could hit a design spec which was realistic; down-clocking the Tegra 3 to the point of making it painful to use in an 11" form factor.
Honestly, while I really like a lot of what Microsoft has done in the past year, the Surface RT was a sincere mistake in every conceivable way. But, given the state of tablet SoC's last year, there wasn't much for it to work with. They didn't have the clout in the channel to pry some of those precious Snapdragons away so they settled for the least ugly solution available.
Honestly, Nvidia is the easiest short idea in technology. Its main business in discrete cards is dying. Games are moving to mobile. Even consoles are going to be a tough sell this winter at any price point above $399. Its primary partner in crime, Intel, is likely going to throw it under the bus with Broadwell only kind of sort of supporting a wide enough PCIe bus to allow discrete cards to function. It has no prayer of competing long-term in the ARM SoC world, regardless of how good its engineers are, and will not be dictating the path of graphics software development for games since it had no viable solution for any of the next generation game consoles.
It should say something loud and clear when AMD in 2012 was viewed as more reliable and more capable than Nvidia to Sony (SNE), Nintendo (NTDOF.PK) and Microsoft. How many Tegra 4's will Nvidia sell this year if it even ships working silicon? How many Jaguar-cores will AMD? There is a reason why AMD won all three of them and it isn't because Nvidia didn't want the business.
Then let's watch the carnage in the graphics space as AMD continues to work towards mobile dual graphics with first Richland and then Kaveri, which should bury any thought of a place for Nvidia at that table either. It's not that all of its products are bad, it's that not enough of them will be good enough real soon now.
I have to laugh at the latest report from Trefis trying to tell us that Nvidia's chips are close to competitive from a cost perspective. $25 for a Tegra 3? A 2012 SoC selling at a comparable price to a 2013 one from Qualcomm is going to seriously accrete to the bottom line? Does it even make any money on those at that price? The Tegra 4i is a quad A9 with an LTE modem that the market has moved on by before it has even shipped.
Where will that compete at the price it'll cost to make? Nowhere close to the cost of a quad-core A7 like MediaTek is producing for the low-end of the market, but that's where it will have to compete to get wins. Nvidia is not Intel. It can't buy its phone design wins with inferior chips and buy media coverage with it. Nvidia may still be making money but it will not be for long if it cannot sell its flagship product and at this point who will buy it? HP is rumored to be kicking the Tegra 4 tires but it is seriously cozying up to AMD for Kabini and Temash allocations. Somehow the price and margin conscious HP (HPQ) and Tegra 4 just don't seem like a good fit, no matter how desperate HP is to make an Android tablet.
The Tegra line was supposed to be Nvidia's way out of this mess all those years ago, but after you've ticked off the OEMs three times for not delivering on your promises that path is now honestly closed. Nvidia should have let Intel buy it when the rumors were floating around a few years ago. By now the two could have merged into a formidable opponent to Qualcomm, Apple (AAPL) and Samsung. As it stands now I fully expect it to be sold for parts and IP by the time the dust settles on this era of computing.
Iceberg dead ahead.