Many people wonder about Sun Microsystems. Sun used to be a major force in the computer world, with its Solaris boxes, introduction of Java, and its mantra “the network is the computer”, among others. It’s not insignificant by any means, but it seems to either be in graceful decline or perhaps biding its time before it reinvents itself once more.
Sun has a few key assets: Java, its hardware IP, its talent, and its cash (flow?). Its share price has languished these past few years. With a cash stockpile and valuable assets, rumours of Sun being bought out or pre-news exaggerations of partnerships are not surprising. Let’s pick Sun apart and posit some conjectures about who’d buy it.
Sun’s cash situation isn’t that exciting. It has two billion in the bank (mostly from Microsoft), a billion in debt, and runs at an annual net loss of a few hundred million (thankfully shrinking). If someone bought them and wasn’t interested in Sun’s revenue generating business (i.e. Solaris/SPARC) in the long term, the buyer could probably get some positive income, declining over the years. A company in the enterprise market with support staff could pull this one off. IBM strikes me as best suited, as they make money supporting declining technologies already (mainframes).
Sun assuredly has talent. Unfortunately, since Sun has been languishing so long, I think a lot of their best talent have already moved on, to the likes of Azul, Google, and others. As Scott McNealy put it himself, “We have littered the industry with ex-Sun employees.” However, those remaining aren’t dunces by any means, so there’s be some value there. There’s no reason to buy the company to get the people, though; Sun’s presence and share price isn’t great enough to resist direct appeals to its employee. Buyers would find it cheaper to “cherry-pick” the talent they want.
Sun’s IP is its the most interesting asset. Let’s separate Sun’s IP into the software and hardware. The software side includes Java, StarOffice, Java Desktop System, NetBeans, etc. The hardware IP is the network appliance technology that presumably is what Azul and Sun are fighting over. Solaris/SPARC, too, but we already discussed that.
Of the software IP mentioned so far, NetBeans, StarOffice, and JDS are probably hopeless causes on their own. With OpenOffice based on StarOffice and freely available, how will StarOffice do better than OpenOffice against the incumbent, Microsoft Office. Same deal with JDS v. RedHat/SUSE, and NetBeans v. Eclipse. Java is a very valuable asset, given its acceptance in enterprise software. Many systems and technologies are based on Java. Sun doesn’t seem to turn much of a profit from Java, except to run big Java apps on Solaris boxes.
Looking at Sun’s software IP, I think of IBM. Only a company with the wherewithal to write JVMs would be interested in owning Java and the language direction – the other commercial JVMs I’m aware of come from Oracle and BEA. IBM has Eclipse (NetBeans’ Matisse and the dev team could go there), gives up Windows internally (switch to StarOffice and prove it’s feasible), and have deeply committed to Linux in the long run (which currently is RedHat/SUSE for enterprises; with JDS, IBM could have their own). To a lesser degree, Oracle has similar potential synergies, but Oracle’s revenue engine is their database (a software product); would Oracle be okay with the free mentality pervasive in the other products? IBM, on the other hand, is a services and support company, and their newer software products all depend on Java.
IBM runs a risk for many of its products if Java was displaced, e.g. if Java lagged the pace of innovation of .NET for too long. IBM has already asked Sun to open source Java. IBM should be interested in Sun’s software IP, if only to better control its own destiny after committing to Java/Linux. If Java did go open source, though, that’d be better for IBM since they’d get the key asset for free.
The hardware IP is valuable, too, as the showdown with Azul indicates. For the hardware IP, we’re looking for a company who finds network attached appliances and big multi-CPU servers useful. This pretty much limits us to large enterprises and those who service them. Microsoft, Oracle, and IBM are the biggies.
Microsoft isn’t likely to try to displace its per-server Windows licenses. Oracle charges per CPU, but has not otherwise shown interest in hardware. Again it comes to IBM, which has already shown an interest in this direction with its purchase of DataPower. Wait, isn’t there someone else who has a network appliance and has tightly packed servers working on large data sets? Google.
While it sounds like I’m joining the bandwagon, as everything seems to be about Google nowadays, Sun’s hardware IP does have value to Google. Internal to Google there’s some value, but not enough to buy IP. However, if Google decides to approach the enterprise market again, and does so using the path started by the Google Search Appliance, then network appliance IP would be useful. This is particularly the case if Google takes their expertise in managing their massive computing/storage needs and commercializes it. Google’s not going to expose their IP, but how about building a few giant appliances that enterprises can distribute on their network, to have specialized processing and storage capabilities based on Google’s own?
I think the ideal case for the industry and Sun’s shareholders would be for Sun to split (like Palm did) into a hardware and a software entity. The software entity goes to or partners with IBM, and IBM advances the Java world with the same commitment and innovation shown for Eclipse and Apache – it’s self serving of course, as IBM displaces non-Java solutions for their own Java based ones. The hardware entity (which merges/partners/settles with Azul) shacks up with Google, who uses its software expertise with network appliance technologies to create black boxes that non-software businesses with lots of data (e.g. financials, health care) can use to achieve their Holy Grail: analysis and management of massive amounts of structured, unstructured, and semi-structured data, for a cost that is linear with scale.
The IBM/Sun scenario could happen today with benefits to IBM. C#/.NET, with Microsoft’s vast resources behind it, is advancing into more “f
rontier” areas than Java, and doing it faster. Some of those won’t pan out, but when the successful ventures start realizing gains (we’re talking at least a few years from here), Java is going to be left behind.
The Google/Sun scenario seems unlikely to be of value today, but it seems reasonable for some point in the future. I’m certain that many non-software businesses would love to get out of writing large amounts of software, and get into a model where the business-agnostic software is bought (easy cost/benefit, predictable timeline), and only business-specific software (e.g. custom analysis logic) is written. i.e. buy the software one is not an expert in (e.g. parallel programming, distributed caches, fault tolerant behaviour, massively scalable data processing) and write the specialized software for the areas one is an expert in (e.g. stock market prediction algorithms, statistical prediction models of cost).
Am I being too fanciful? Unrealistic? <shrug> It’s fun speculating. Something has to change with Sun, though; their graceful glide to the ground is going to pick up speed. Whether it turns into a plummet, a rebirth, a forceful buyout, or a distribution of assets to those who can best use them, it’ll be interesting to see.
