3 Comments

Summary:

Who would have thought that the cloud computing revolution would have turned Larry Ellison into a CEO who is hunting for chip acquisitions? But since Oracle’s vision for the cloud includes a highly customized box, why not own the brains that would run it?

2010-09-22_10-37-34_539

Who would have thought that the cloud computing revolution would have turned Oracle’s Larry Ellison into a CEO who’s hunting for chip acquisitions, as he himself said yesterday? However, Ellison is no dummy, and after Oracle World and the new Exalogic gear the company unveiled, it’s clear that Oracle is attempting to become the Apple of the enterprise sector. It’s building custom boxes, software and whatever else to deliver a user experience for the cloud that is simple to run and deploy, but proprietary as heck. No wonder the majority of the enterprise software geeks are slamming it.

I was surprised that Oracle actually released the latest version of the Sparc chip Sun was touting a while back, but it appears I shouldn’t have been. Since Oracle is pursuing this tightly integrated strategy, such differentiation at the silicon level makes sense. Look at Apple’s purchase of PA Semi and Intrinsity to goose its own hardware as it pushed the envelope battery life and performance for mobile computing. Oracle’s efforts to reduce energy consumption and deliver scalable performance for cloud computing may require the same commitment to custom chips, and Sun’s Sparc line may not be doing the job. So Ellison’s comments yesterday open the door for a variety of possibilities. He said:

“Our focus is to build our (intellectual property) portfolio. … You could see us buying chip companies,” Ellison said. “Silicon is very important, software IP is very important.”

As always happens when execs talk about buying chip companies, the stock of ARM, the British licensing company behind the low-power chips in cell phones and embedded products, saw its stock shoot up. I doubt Ellison is going to buy ARM, even though ARM has an $8.68 billion market cap and Oracle has $23.64 billion in cash and a hankering for controlling IP. ARM would be a tough sell simply because so many huge companies are dependent on its technology. It has over 600 licensees — some of which are huge like Apple, Samsung and Microsoft — putting its chips into servers, mobile phones and set-top boxes. It’s hard to imagine a company like Oracle buying ARM; it simply wouldn’t make sense for Oracle to keep the R&D dollars flowing and support a huge licensing business merely to gain a competitive advantage in the enterprise data center. Plus, the outrage from ARM licensees could halt a deal. A more likely bet would be to look to the ARM-licensees for buyout targets. Perhaps Smooth-Stone, which is building a low-power server using ARM-based chips and its own silicon to control the networking, or one of the high performance computing startups that have built a specialty chip for performance could work.

As for the other large chip players like Marvell, Broadcom or AMD, I’m not convinced those make any more sense. Yes, AMD’s in the toilet, but Oracle doesn’t need to buy AMD to get an x86 chip. If one wants to build some kind of super server, I suppose AMD’s planned GPU-CPU Fusion hybrid might have merit, but again, AMD’s still a big purchase for an x86-based chip. As for Broadcom, Marvell and the rest of the big chip vendors, it’s hard to see Oracle wanting a wireless radio play when it is building enterprise boxes. So my hunch is smaller HPC vendors, perhaps buying the computing line from a lesser-known company or even IBM’s Power PC line, or finding a smaller ARM licensee that has an interesting server technology plan.

Related GigaOM Pro Research (sub req’d):

    1. Mofo awesome post. Ellison says “Silicon is very important, software IP is very important.” A couple of reasons why Oracle buying ARM (if they were to go down the direction of buying a chip company would be an excellent idea):

      1) Silicon: ARM makes all its money syndicating their IP to the other chip mfgrs that have their own angles (Broadcom, Qualcomm, TI, Marvell, etc) and they are all incentivized to give Intel a run for its money. In the world of system on chips, this is incredibly important to ensure the market has competitive chip pricing but still keeps ARM working hard on R&D — I digress. By owning ARM, Oracle will be completely hedged on fundamental mobile IP and will have first mover access to understanding how core mobile R&D innovation today will affect the cloud tomorrow. And as Om pointed out a million times before, mobile growth is a really big deal for the cloud.

      2) Software IP: Intel buying McAffee and Infineon’s mobile chip group in the same two weeks, Red Bend buying VirtualLogix, RIM eating up QNX (an old pioneer in embedded software), Qualcomm heavily increasing its investing in developer tools all in the same year are part of a larger theme. With the rise of multiple connected devices and applications moving to the cloud, owning fundamental services (i.e. security, resource allocation, and having insight on QoS as per the above mentions) all point to the rise of important middleware. Desktop applications are being relegated to antiquity and the value of that entire layer in the stack in now shifting up to the cloud and down to the middleware. Apps on your TV, mobile, etc will largely be powered by data from the cloud and critical resource management from your favorite chip mfgr.

      3) Concurrency vs raw horsepower and energy consumption: While Intel has traditionally been good at increasing clockspeed, ARM was able to create environments where a few different apps could play nice together without wasting a lot of power. With the rise of the real-time web, and energy costs being more and more of a real issue, people are talking about ARM chips going into servers. If Ellison really means what he says about silicon and software IP being important, he can be a real catalyst for making the ARM server dream a reality.

      If anyone has white space on their website they would like to fill up, please let me know.

      Share
    2. It is fascinating to see this development. It is just over 40 years since the move by IBM to unbundle software (http://en.wikipedia.org/wiki/History_of_IBM#Unbundling_of_software_and_services_in_1969) which launched the software industry. The pendulum is moving in the opposite direction. Customers are probably looking for fully integrated solutions because of complexity of distributed computing. Highly specialized centers of competence such as server, network and software have created silos of expertise. Orchestrating the workflows in the different departments or finding the root cause of problems is complicated. No wonder customers see value in needing to deal with one vendor for integrated solutions.

      Share

    Comments have been disabled for this post