There are a lot of differences between the two biggest U.S. chipmakers, but as a pair of stories this week demonstrate, 2016 has provided a major common thread between them.
Each company has realized it's unlikely to see much (if any) growth in the coming years from the market that propelled its rise -- and one that needs to make big changes in response.
Intel (INTC) , recognizing PC CPUs will likely be a flat-to-declining business, has carried out big layoffs that hit its PC and mobile processor operations hard, and unveiled a 5-part strategy that aims to transform Intel "from a PC company to a company that powers the cloud and billions of smart, connected computing devices."
To back up its talk, Intel has launched a slew of new products meant to grow the company's consumer electronics exposure and increase its data center footprint. It has also formed autonomous driving-related alliances with BMW and Delphi and acquired server deep learning chip developer Nervana Systems and computer vision chip developer Movidius.
Qualcomm (QCOM) , meanwhile, is making the biggest acquisition in chip industry history to lower its reliance on mobile processor/modem and licensing businesses due to be pressured by weak smartphone growth, intensifying competition from other chipmakers and (down the line) 3G patent expirations. The company is spending $47 billion to buy Dutch chipmaker NXP Semiconductor (NXPI) , which among other things is the world's biggest supplier of automotive chips, smart card microcontrollers and NFC radios.
On Wednesday, Qualcomm took another step in cutting its mobile dependence by announcing it has begun sampling the Centriq 2400, its first offering for the nascent ARM server CPU market, to "key prospective customers." Google, reported in February to be interested in Qualcomm's server CPU efforts, could be among the firms sampling the product. Volume shipments aren't expected until the second half of 2017.
The Centriq 2400 packs 48 of Qualcomm's custom Falkor ARM CPU cores. Qualcomm is counting on the custom cores to help it take on Intel's dominant Xeon server CPU line, as well as its use of an advanced 10-nanometer manufacturing process -- possibly Taiwan Semiconductor's (TSM) , given Qualcomm's longstanding relationship with the chip foundry.
Helping Qualcomm's cause: Cloud giants and others are intrigued by the potential cost, density and/or power benefits ARM server chips could deliver, as well as the potential for ARM solutions (such as Cavium's ThunderX line) optimized for applications such as web servers, storage hardware and security appliances. And with Intel having slowed the pace at which it launches new manufacturing processes, the company no longer has a big process edge over TSM.
On the flip side, the ARM server market hasn't developed nearly as fast as many once expected. Cavium and AppliedMicro are its two main suppliers, and each remains a niche player in the broader server CPU market. Macom, which last month struck a $770 million deal to buy AppliedMicro, plans to unload the latter server CPU business.
A limited software ecosystem has stunted the ARM market's development, and so has Intel's aggressive efforts to meet the needs of enterprise and cloud server buyers. Samsung and Nvidia cancelled plans to go after the market in 2014, and Broadcom (AVGO) was just reported to have bowed out. AMD (AMD) is still participating, but it seems to be paying more to battling Intel in the data center via non-ARM (x86) CPUs, via its new Zen architecture.
While AMD and Intel remain fierce PC and server CPU rivals, they may be nearing a graphics detente. AMD rose 8.9% on Tuesday following a rumor that claimed the companies have inked a deal that will result in AMD Radeon GPU intellectual property being integrated with future Intel CPUs. This meshes with a March Bloomberg report stating Intel is talking with AMD about a graphics-licensing deal.
If Intel outsourced to AMD the development of the GPU cores integrated with its PC and (in a few cases) server CPUs, the benefits could be twofold. First, it would quite likely improve the competitive of Intel's integrated GPUs, which have lagged those of comparable AMD offerings even as the CPU performance of Intel's chips has remained superior. This could also make Intel a larger headache for Nvidia's (NVDA) low-end PC GPU business, albeit while having little impact on Nvidia's more lucrative high-end GPU business.
The second benefit: It would allow Intel, which just took a big ax to its PC CPU operations, to further pare the business' R&D spend. Intel has made it clear it wants to invest less in PCs and more in areas such as servers, consumer electronics, cars and embedded/IoT devices. Letting AMD handle its GPU R&D would be one more step in this direction.
Nonetheless, this is the kind of move that would've been hard to fathom several years ago, given the historical centrality of the PC market to Intel and the company's reputation for having a do-it-yourself mindset for the technologies going into its processors.
Then again, it would've also been hard to imagine Qualcomm making a big push into servers.
There's nothing like a lack of growth in a core business to upend a company's priorities.