Intel (INTC) needs to buy NVIDIA (NVDA), and it really ought not to wait any longer to do so. It is my belief that NVIDIA is truly on the cusp of some pretty game-changing things, and if Intel doesn't go ahead and buy NVIDIA now, it will evolve and grow into a competitor that - in my humble view - could be a real problem down the line. Without an X86 license in a world where X86 was the de-facto standard for mainstream compute, NVIDIA wasn't much of a threat outside of its Tesla business. But with ARM (ARMH) giving NVIDIA's engineers the freedom to build its own processors, the graphics giant has the potential to be an even fiercer competitor than AMD (AMD) was in the last decade in the HPC space, as well as the mobile SoC space.
NVIDIA's "Denver" Likely To Be Credible Threat In HPC Space
In the world of high performance computing, Intel's major competitor is - you guessed it - NVIDIA. Granted, while NVIDIA's HPC accelerators are largely attachments to Intel processors rather than standalone devices (but still direct competitors to Intel's own Xeon Phi accelerator line), I believe the capability to do a custom ARMv8 processor will expand NVIDIA's reach into this space.
It is my expectation that NVIDIA will actually integrate its own processor cores into upcoming GPUs, a move that will significantly improve the efficiency of the communication between the central processing units and the vector processor. I am expecting that Intel's next generation Xeon Phi co-processors will integrate "Goldmont" (14nm Atom "tock") processor cores alongside the vector processors for the same reason that NVIDIA is likely to do so with its Denver + GPU.
While I do not expect that the "Denver" processor core will make it into processors that go head-to-head with standalone "Xeon" processors for this space (ARMv8 + NEON is still quite limited compared to X86 + SSE 1/2/3/4 + AVX/AVX2/AVX3), and while I'm not expecting NVIDIA to actually go after the microserver space (yet), I think that longer term the company becomes a non-trivial threat to Intel in some parts of its datacenter group.
Graphics Dominates Smartphone/Tablet Workloads And NVIDIA Is Best-In-Class
Intel's own internal graphics efforts have been getting much better over the years. However, it would be a stretch to assume that architecturally Intel has caught up with either NVIDIA or AMD . While I do not know what Intel's "Gen 8" graphics cores slated to launch with "Broadwell" next year look like, I would assume that whatever NVIDIA has cooking up for next year (codename "Maxwell") should still be able to keep one step ahead architecturally, which should offset Intel's process advantage.
But here's a thought; what if Intel could build NVIDIA's world-class GPU designs on process technology that nobody else has? Not only would discrete gaming GPUs be a nice, high margin business that could soak up a good amount of future fab capacity, but the IP and expertise would ensure that Intel would have world-class SoCs as far as both CPU and GPU are concerned. Further, as gamers resonate well with the NVIDIA brand, this could become yet another selling point for Intel's SoCs - merging the well-liked NVIDIA/GeForce brands with the well liked Intel/Core/Atom brands. Many gamers today prefer NVIDIA discrete GPUs with Intel CPUs, so why not join the two on the same die?
At the end of the day, it's well recognized within Intel that GPUs will only get more important. And while I suspect that Intel's own internal efforts will eventually become "great", eliminating a competitor and consuming its arguably better technology now seems like a pretty good idea.
NVIDIA Is Dirt Cheap
NVIDIA trades at 7.28x EV/EBITDA, and I believe that this metric would look a lot better if the company just killed Tegra since that division is bleeding money. Even if Intel paid $25/share for the company (60% premium to most recent close), it would gain plenty of IP, a leadership position in discrete gaming GPUs (that would only widen as NVIDIA would move next generation GPUs to Intel's leading edge processes and make life even harder for AMD), eliminate the only real competitor in the HPC accelerator space, and arguably the world's best GPU IP. And, to boot, it would remove a potentially very dangerous competitor in the mobile space. On top of all of that, it would keep NVIDIA out of a larger player's hands.
This is highly speculative and probably unlikely to happen anytime soon (if ever), but I really do believe that it'd be a good idea for Intel to pony up and pick up NVIDIA at these levels. While Intel would need to lever up in order to do it (or do an all-stock deal near the low point of its share price), from a long term perspective, I can't really see why this would be a losing proposition. What do you think?