Attend Shahin Farshchi’s free webcast “5 Tools for Building Value Into Your Hardware Startups,” being held May 19 at 10 a.m. PT.
There are many compelling reasons to package new technology as a cloud service. Connected devices come in many forms: dongles, phones, tablets, televisions, cars, and even buildings. Intel is offering “connected buttons,” and others are introducing connected jewelry and accessories. Internet connectivity is also available through many channels: satellite, cellular, WiFi, bluetooth, and hybrid meshes. The plethora of powerful, beautiful connected devices, coupled with ubiquitous connectivity, creates an incredible channel for delivering novel services.
Hotmail, Salesforce, Workday, and many other software-as-a-service companies have fared well by offering their applications directly through Internet browsers. DropBox and Box, while creating tremendous media attention, have yet to prove they can offer storage services profitably on the cloud. Amazon doesn’t disclose the economics of its Amazon Web Services business in detail, though one would expect the opposite to be true if it were a lucrative business. ASICMiner and KNCMiner are leveraging their proprietary hashing chips to offer bitcoin mining as a service. Nervana is leveraging its proprietary chips as a service for deep learning. As more entrepreneurs attempt to harness the cloud as a powerful distribution channel for their novel technologies, here are a few factors to consider.
- Understand the economics of your business. Investors in service businesses look at potential return on invested capital (ROIC). ROIC is generally what the business nets in profits annually as a percentage of invested capital — the fixed cost required to offer the service. For example, the economic parameters of running a bitcoin mining service can be measured: hash rate, bitcoin price, and energy costs, against the cost associated with building and running the data centers. Though it is difficult to predict both hash rates and bitcoin prices, it is possible to model out different scenarios. The corresponding parameters for more nascent businesses, such as deep learning, are far more difficult to anticipate. For deep learning, applications range from image and speech recognition to natural language processing and autonomous vehicles. In lieu of available technology, it is difficult to predict what customers will pay, and for what. In rare cases, the data generated by the novel hardware could be of interest to a deep-pocketed strategic buyer, such as Google’s acquisition of Nest. In most cases, customers will be in the process of determining whether it makes sense for them to a) procure computation as a service (hence, use you), b) buy their own hardware, or c) rely on others to buy and maintain the requisite hardware. Some deep-learning hardware for mission-critical applications, such as autonomous driving, might reside locally in the car — though, I anticipate the bulk of AI/deep-learning algorithms to run in the cloud.
- Consider cost of capital. As a rule of thumb, capital is cheaper for less risky endeavors. If you find yourself needing millions to build out a technology that hasn’t been proven at scale in a nascent business, understand that you — and your early investors — will likely have to pay dearly to demonstrate the anticipated economics. Understand that traditional venture capitalists seek to fund companies that can either prove that a big business can be built around a new product that is relatively cheap to bring to market (hence, the popularity of software and consumer apps), or introduce a product that can address immense latent demand (i.e., clean energy or a cure for cancer). Be intellectually honest, and use proxies to determine the scale, time, and capital needed to prove out the economics of your service business. Then, ask yourself whether the likely dilution makes pursuing a service business a wise choice.
- Determine if Moore’s Law is your friend. The 18-month time constant associated with doubling computing power has been expected to grind to a halt in the next three years for the past 10 years. Immersion lithography, double patterning, extreme ultraviolet, novel transistor structures, and exotic materials have kept Moore’s Law chugging along. Compute power will continue to get cheaper and more energy efficient. Unfortunately, many semiconductor start-ups fall on the wrong side of Moore’s Law. As newcomers, they need to show customers silicon before standing a chance of becoming qualified as vendors. Furthermore, design cycles take longer, and volumes are lower, to the point where margins are rarely generated on initial designs, while established semiconductor companies sell their chips profitably. Since competitors will be offering their next-generation products on the more advanced nodes, the start-ups need to repeat the process of giving away equity to design their next chip. For many, this cycle repeats itself many times over — hence, the large capital requirements of most semiconductor companies. The same holds for a start-up cloud service company based on proprietary chips. As the next technology node comes around, competitors will be able to offer the same service with lower fixed and incremental costs. As the impact of your technology diminishes with time, so does your competitive advantage.
- Think beyond your initial service. Remember on-demand television? Many network providers invested heavily in infrastructure to support services that took many years to roll out, to the point where they may be rendered obsolete by Internet streaming. If the markets and economics of your service are still unclear, it might be a worthwhile exercise to evaluate other potential applications of the platform, or perhaps even take on a more vertical approach by implementing several applications to kickstart demand. The early Internet service providers entered the business offering connectivity, but made significant profits on value-add services, such as search, email, and marketplaces.
- Be creative about blending hardware and services. Apple’s iTunes service has been the envy of many start-up and established hardware companies — it did $18 billion in revenues in 2014 alone. Much of iTunes’ success can be attributed to Apple’s ability to procure practically all content that their customers would ever want to consume on their device. Though the instant gratification in exchange for parting with 99 cents was a no-brainer, it was the outcome of Apple’s persistent negotiations with record companies, while skeptics stood by and watched, that made the difference. Until recently, Apple was the only company with digital content distribution rights, with the iPod/iPhone being just a vehicle. On the other hand, DropCam’s Cloud Recording service obviates the need for local storage, but that hasn’t stopped competitors from offering similar services, effectively creating a race to the bottom. Similarly, wearable activity monitors such as the FitBit and Misfit offer free apps and cloud tracking in an attempt to push product and differentiate. The development of your hardware and service should go hand-in-hand toward creating a unique experience for your customer. Simply attaching a cloud service, or hoping someone will pay for the “data,” is wishful thinking, or at best, an inevitable race to the bottom.
Avoiding the long, arduous design and sales cycles associated with hardware integration and sales makes going direct to customers via cloud services appear attractive. However, entrepreneurs must apply the same principles and decision-making processes used toward making any infrastructure investment. Examples abound in energy, telecom, as well as existing cloud service companies. Do the math, understand the impact of technology trends, and be creative about the scope of your offerings. You’re far better off putting in the elbow grease today.
Cropped image on article and category pages by Nick Hughes on Flickr, used under a Creative Commons license.