I have gone on record to claim that NVIDIA (NVDA -3.33%), in light of its well-publicized exit from the mainstream smartphone market, doesn't need Icera. Fellow Fool Evan Niu recently made the case that NVIDIA should simply get out of the cellular baseband market entirely.
In fact, I recently stumbled upon another piece by Evan, in which he highlights the following quote from CEO Jen-Hsun Huang:
There is no standalone modem business anymore, and in ... [the] new 4G connected device marketplace, an integrated approach is necessary, and that's the reason why we bought Icera, and that's the reason why we're investing in LTE. So that's really one investment. You have to do it all together. or you don't do it at all.
In light of these statements, and the fact that NVIDIA has claimed that it won't develop another integrated apps processor and baseband product, should investors expect NVIDIA to shutter its baseband business?
Providing some context
Back when this statement was made, NVIDIA had made it clear that it was gunning for a pretty big chunk of the smartphone and tablet markets. Today, it's pretty different; it wants to go after "devices where performance and differentiation matter."
Further, NVIDIA is making a credible push into the automotive space, where, as fellow Fool Steve Symington points out, Tegra is giving cars the ability to "think like a supercomputer." NVIDIA's Chris Evenden pointed out at a recent presentation at a Wells Fargo conference that the company "still need[s] the baseband if only for automotive."
However, a point that I made recently is that if NVIDIA really needs a standalone baseband to succeed in automotive, it can just use a third-party chip from the likes of Qualcomm or Intel along with its Tegra. That means NVIDIA has to hand over the margin for those chips to one of these companies, but I have contended that the additional R&D would outweigh any potential cost of goods sold savings.
I'd like to explore this a little bit more.
Laying out some assumptions
At NVIDIA's recent analyst day, CFO Colette Kress showed a slide suggesting that NVIDIA's investment in Tegra came in at approximately $400 million. This is actually down from the $600 million figure that the company gave at GTC 2013.
To put this in perspective, Broadcom, which recently shuttered its own apps processor and cellular modem business, claimed annual research and development savings of about $700 million from the move.
It's not clear how much of Broadcom's cellular spending was on the modem versus all of the other pieces of IP found in a typical smartphone chip, but for the sake of this argument, let's suppose that a third of Broadcom's spending went toward the baseband and attendant RF technology, or about $233 million.
Now, according to IHS iSuppli, the Qualcomm MDM9615 baseband, WTR1605L transceiver, and RF front end runs Apple about $32 in the iPhone 5s. My guess is that the baseband and transceiver (which is, what I believe, NVIDIA provides as part of its cellular solutions) make up the bulk of this cost, so for the sake of this analysis, let's call the baseband and transceiver a $20 solution.
Finally, I'm going to make the assumption that for each unit of cellular baseband and transceiver that NVIDIA sells, it generates an incremental $15 (a discount to Qualcomm's premium solution) of revenue at 40% gross margin relative to the apps processor.
So, how many cellular modems does NVIDIA need to sell for this to make sense?
To recoup the approximately $233 million per year in baseband R&D that I estimated, NVIDIA would need to generate $582 million in incremental cellular revenue at the 40% gross margin level. Assuming $15 a piece for the baseband and RF transceiver part of the solution, this works out to about 38 million NVIDIA baseband and RF solutions in tablets, gaming devices, and automotive.
Now, in an email exchange with Nick Stam, senior director of technical marketing for NVIDIA, he claimed that they were expecting "25 million NVIDIA processors to ship to automakers within the next five years."
Further, given that NVIDIA seems to be playing at the high end of the tablet marker exclusively, I would be hesitant to expect the company to garner more than 10% of the entire tablet market with its Tegra processors. Based on a forecast found on Statista, the tablet market should grow to approximately 384 million units by 2018.
If NVIDIA grabs 10% of this with its Tegra processors, then this works out to about 39 million tablet units sold total.
Will it be able to do so?
According to a recent Intel presentation at its 2014 Developer Forum (which cited numbers from Strategy Analytics), approximately 32% of tablets shipped in 2018 will require either 3G or LTE cellular connectivity. Applying this percentage to my estimated NVIDIA tablet chip shipments, that gets us to 12.5 million potential cellular shipments.
To get to that 38 million cellular unit number estimated earlier, just about every automotive-oriented solution that NVIDIA sells would need to ship with a discrete cellular solution from NVIDIA.
Now, this analysis doesn't take into account the fact that NVIDIA is pushing into Google Chromebooks (Intel's CEO recently claimed that eventually, roughly half of Chromebooks shipped will sport cellular connections). There may also be other markets that NVIDIA can push cellular connectivity into that I don't see.
Foolish bottom line
I freely admit that if some of the assumptions here are tweaked, the breakeven point could be lowered. However, the key takeaway really seems to be that if NVIDIA shuttered its baseband business, it could realize a significant cost savings and better allocate that capital.
In fact, as noted, NVIDIA's Tegra R&D number did come down about $200 million from the number given in 2013. Perhaps the "shuttering" has already happened?
