While the timing of my post was fortuitous, today’s news suggests that Intel has been thinking about the data-center merchant silicon for some time. Acquisitions typically don’t come together overnight, and Intel doubtless has been taking careful note of the same trends many of us have witnessed.
Data Center on a Chip
In announcing the deal today, Intel has been straightforward about its motivations and objectives. As Intel officials explained to eWeek, Fulcrum’s chip technology will not only allow network-equipment vendors to satisfy demand for high-performance, low-latency 10GbE and 40GbE gear, but it also will put Intel in position to fulfill silicon requirements for all aspects of converged data centers. With that in mind, Intel has stated that it is working to integrate a portfolio of comprehensive data-center components — covering servers, storage, and networking — based on its Xeon processors.
With converged data centers all the rage at Cisco, HP, Dell, IBM, (and many other vendors besides), Intel wants to put itself in position to meet the burgeoning need.
Intel did not disclose financial details of the acquisition, which is expected to close in the third quarter, but analysts generally believe the deal will have only modest impact on Intel’s bottom line.
Strategically, though, the consensus is that it offers considerable upside. Intel apparently has told Deutsche Bank analysts that it now captures only about two percent of overall expenditures dedicated to data-center technology. Fulcrum is seen as a key ingredient in helping Intel substantially boost its data-center take.
Unlikely to Repeat Past Mistakes
The deal puts Intel into direct competition with other merchant-silicon vendors in the networking market, including Broadcom and Marvell. Perhaps a bigger concern, as pointed out by Insight64 analyst Nathan Brookwood, is that Intel failed in its previous acquisitions of network-chip suppliers. Those acquisitions, executed during the late 90s, included the $2.2-billion purchase of Level One.
Much has changed since then, of course — in the market in general as well as in Intel’s product portfolio — and Brookwood concedes that the Fulcrum buy seems a better fit strategically and technologically than Intel’s earlier forays into the networking space. Obviously, data-center convergence was not on the cards back then.
Aligned with March to Merchant Silicon, Rise of Cloud
To be sure, the acquisition is perfectly aligned with the networking community’s shift to merchant silicon and with the evolution of highly virtualized converged data centers, including cloud computing.
One vendor that’s enthusiastic about the deal is Arista Networks. In email correspondence after the deal was announced, Arista CEO Jayshree Ullal explained why she and her team are so excited at today’s news.
First off, Ullal noted that Arista is one of Fulcrum’s top customers. Intel’s acquisition of Fulcrum, Ullal said, “validates the enterprise-to-cloud networking migration.” What’s also validated, Ullal said, is merchant silicon, as opposed to “outdated clunky ASICs.” Now there are three major merchant chip vendors serving the networking industry: Intel, Broadcom, and Marvell.
Ullal also echoed others in saying that the deal is great for Intel because it moves the chip kingpin into networking/switch silicon and cloud computing. Finally, she said Fulcrum benefits because, with the full backing of Intel, it can leverage the parent company’s “processes and keep innovating now and beyond for big data, cloud, and virtualization.”
Even though, monetarily, there have been bigger acquisitions, today’s deal seems to have a strategic resonance that will be felt for a long time. Intel could play a significant role in expediting the already-quickening commoditization of networking hardware — in switches and in the converged data center — thereby putting even more pressure on networking and data-center vendors to compensate with the development and delivery of value-add software.