Intel isn’t a company known for being a spendthrift, but it just spent more on Altera than the combined total for all previous acquisitions. Intel isn’t know for placing big bets on anything other than its own R&D and fabs. It’s acquisition strategy has been more like Apple’s, surgical buys and acqui-hires designed to fill defined technology gaps. Unlike HP, Intel, isn’t known for spending big money on new markets. As I write in this column, that history is one reason Intel’s $16.7 billion buyout of Altera, more than twice the size of Intel’s next largest deal ($7.6B for McAfee that also went against the grain in its size and market) is so significant. The deal is very strategic since Altera technology is a great fit with Intel’s rapidly growing data center business unit. Whether Altera is worth the heavy price is debatable, but in this deal, Intel takes a page from Facebook’s playbook, specifically the WhatsApp and Oculus acquisitions, by making a big bet on the future that is based more on vision and faith than products and financials.
As I detail in the column, it’s very hard to justify the Altera deal based on the numbers, which translate to 30-times Altera’s operating income and 8-times sales. This deal isn’t about financial engineering, but data center engineering. As I wrote last fall, Intel Is Already Inside Your Data Center, But Wants a Bigger, Better Spot, and again after Intel’s last earnings report, the company has aggressively moved beyond being a mere supplier of commodity server CPUs and motherboards and is building a complete portfolio designed for software-defined cloud data centers. The only justification for Intel’s move can be its perception of a secular technology shift from commodity processors to custom hardware purpose-built for specific applications.
The column walks through Intel’s rationale and the secular technology trends affecting its data center strategy, but the ultimate wisdom of the Altera acquisition will depend on how accurate Krzanich and his leadership team have assessed hardware trends for cloud data centers and connected devices.
Indeed, this deal may be an expensive admission that the days of counting on Moore’s Law process scaling to regularly yield significant performance improvements using general purpose CPUs are over. Rather, the path to faster applications entails customizing the hardware to application-specific algorithms, something that is difficult, costly and time-consuming using custom-designed silicon, but eminently feasible with FPGAs. I am looking forward to seeing how incorporates Altera technology and adjusts its long-term strategy and will explain what it all means here on MarkoInsights as events transpire.