History repeats itself, it seems, because the defining technology battle of this decade is going to come straight from the 80s: it’s hardware versus software. [tweet] Every decade brings substantial advancements to both software and hardware, but in certain decades the strategic importance of one versus the other shifts dramatically in many segments. I’m using the term hardware loosely to include software wrapped in metal, which is still what companies such as Cisco and EMC live off of. Here is an extremely brief recent history of computing:
- 1950s: the decade of mainframes (go IBM!)
- 1960s: the decade of minis (go DEC!)
- 1970s: the decade of change (DEC ships VAX, Intel ships microcomputers)
- 1980s: the decade of the PC with the clone wars and the commoditization of hardware, assisted by a then little-known company run by a Harvard dropout by the name of Bill Gates
- 1990s: the decade of telecom/network hardware (Cisco goes public in 1990) and Internet software
- 2000s: the decade of storage appliances and smartphones on the hardware side and large-scale Internet software
The period between 1950 and 1980, and the business models of the dominant players, were about hardware. In the ’80s, for the first time, software stood on its own and started taking a significant portion of spending at the expense of hardware. In the ’90s there was more of everything: servers, routers, storage, and during Bubble 1.0 large enterprises wanted at least one of every type of Web-related software. During the millennium decade, hardware made big advances through smartphones and in the fast-growing storage business, while companies were able to spend less on software thanks to broader adoption of open-source technology. This decade will be defined by a reversal of this trend, one that will mimic the ’80s in terms of hardware commoditization.
Most hardware doesn’t matter because some hardware matters a lot. [tweet] Apple owns the top of the PC pyramid through its brilliance in hardware design and through the software leverage of OS X and iTunes. This forces all other PC manufacturers into a deadly, low-margin competition in the low and mid tiers.
Netbooks accelerate the race to the bottom. [tweet] In a short period of time, netbooks have become a big part of portable shipments. Pushed by subsidies from mobile operators wishing to lock users into multi-year plans, netbooks will become “smartphones with larger screens.” Netbooks are great for browser-based applications, which makes the netbook OS and hardware even less important. That’s good, because there isn’t much margin in a $300 netbook.
Virtual appliances replace physical appliances. [tweet] For many years, appliance vendors have extracted additional margin by slapping their logo on a commodity appliance. CIOs want none of this. Virtualization and advancements in distributed systems make it possible to run all kinds of enterprise applications and infrastructure services such as storage, networking, and security on commodity hardware. Commodity again means lower margins for hardware manufacturers, including companies such as EMC and Cisco, who have reacted by shifting their focus to service businesses and pure software packaging.
Cloud computing makes hardware less relevant. [tweet] This decade will be defined by a migration to cloud-based computing for everyone from consumers to the largest of enterprises. On the enterprise side, the move is driven by the desire to lower costs and add flexibility. On the consumer side, it’s driven by the need to manage data and applications across several devices (laptop, netbook, e-reader, mobile phone, etc.). Cloud-based architectures buck a multi-decade trend and emphasize service level agreements (SLAs) that come from software as opposed to hardware. Instead of powerful, expensive servers, high performance and availability come though horizontal scaling of unreliable, cheap servers combined with new distributed software architectures. On top of this, the very large cloud vendors will operate vast server farms which, increasingly, as Google does today, will deploy commodity custom servers. Even less margin for the major hardware players.
Software begins to dominate as the driver for mobile device purchases. [tweet] Historically, back in the now distant days when mobile phones were primarily used for calling, consumers chose the phone first and then went along with whatever software came with the device. RIM was the first to change this with the Blackberry. Then came Apple’s iPhone. In both cases, these were systems-hardware and software came together and were supported by additional desktop and server software, namely the Blackberry desktop client and Blackberry Enterprise Server and iTunes / iTunes Store. What is more important? The design of the Blackberry device or the fact that it’s the best mobile e-mail machine on the planet? Would you have bought the iPhone if it ran Motorola’s clamshell software? Google’s Android mobile OS pushes the divide further. I’m not suggesting that there won’t be really successful mobile hardware innovations. There will be a continuous streak of delightful innovation in devices. I’m simply arguing that, in this decade, the relative importance of mobile software and the third party ecosystem of software products and services will dominate.
Tablets are the obvious dark horse on the hardware side. If the human I/O problem is solved, we could see a radical shift in form factor that should exceed that of netbooks. I guess we won’t know for a few years. Even if Apple “does an iPhone” with its iSlate, it will be a long time before their volume meaningfully affects the landscape. Steve doesn’t like to sell things cheap.
There is one less obvious dark horse that hasn’t been named yet so let’s call it Rackware. [tweet] The “commodity custom servers” in data centers I mentioned above will look quite different from the typical servers that go on racks today. Google’s already do. In fact, they may combine CPU, memory, storage and I/O in very different and more variable ways than current servers for heat density, workload optimization and I/O virtualization reasons. Also, they may come not in server units but in rack units or other types of units (such as Google’s containers) that make deployment and management, including power and cooling, much cheaper and easier. It is foreseeable that a company could create sufficient new intellectual property in this area-both on the system and supply chain management side-to command premium margins for a period of time. Dell is a good example of this: one of the key differentiators they had in the early days was a supply chain patent that covered the just-in-time manufacturing of customized PCs.
The large scale value shift from hardware to software will have significant ramifications for innovation, venture capital, and investing. It will be an exciting decade.
