The Future of Semi Startups

Semiconductors play a critical role in enabling the widespread deployment of cloud, data-center and mobile technologies. The rapid growth of compute, networking and storage in the cloud, faster mobile and core bandwidths, sophisticated consumer devices, stronger security at decreasing cost and power are all enabled by improvements in chips.

The semiconductor industry is delivering these improvements using rapid evolution of architectures and standards, increased chip integration, higher I/O and DRAM bandwidth rates, latest process nodes going to 20nm and smaller, 3D transistor technologies for lower power consumption along with breakthrough multi-chip packaging capabilities.

At the same time, investment in new revolutionary semiconductor startups has plummeted in recent years (source: GSA). This does not bode well for long-term, disruptive innovation in the industry.

There are a few reasons for this trend:

1. The cost of developing new silicon in latest process technology nodes from scratch is considerably larger compared to 7-8 years ago! Developing a competitive product requires a tremendous amount of functionality in one chip at the 28nm or 20nm process node (due to transistors being much smaller vs earlier nodes). This increases cost of intellectual property (IP) – whether it is developed in-house or licensed. Further, the cost of a mask-set for a new chip design in 28nm is in the $2M to $3M range, and consequently the penalty for any bugs/errors is significantly greater. In fact, the cost-per-transistor, which has declined consistently in every prior process node reduction, is expected to be same or higher as the industry transitions to 3D transistors at the 20nm process node.
2. The time to develop a new product has increased due to the greater amount of IP required. In addition, the turn-around time is longer through the fab when using smaller geometries. Once silicon products become available, OEM customers then design systems using them and the revenue ramp really starts once the OEM customers start shipping. So time to revenue from development start for silicon products has continued to increase. This is in contrast to time between new software releases delivered through the cloud (e.g. Workday or Salesforce.com), which have reduced dramatically.
3. Venture investors have a range of alternatives in areas such as consumer internet, cloud services, enterprise and infrastructure software and systems, where the investment and risk required to achieve profitability and liquidity is substantially lower. IPOs and acquisitions such as Workday, Nicira, Fusion-IO and Data Domain have delivered superior attractive returns on investment compared to silicon investments. In contrast, there have been few IPOs with similar profile in the semiconductor space over the last 5-6 years.

So, what does this mean for entrepreneurs, investors and the semiconductor industry going forward?

In examining the overall semiconductor industry, there are segments where new startups have an opportunity to develop breakthrough products, compete against large incumbents and deliver superior returns to entrepreneurs and investors. The characteristics of these segments include: new technology innovation (which is obvious), lack of need for latest process technology nodes and inability for incumbents to easily integrate functionality into existing or future products. Examples of these segments include:

• Analog products such as RF power-amplifiers and PHY technologies addressing newer applications/interfaces. Analog IP of significant size is hard to integrate into digital silicon dominated chips as it impacts the yields and is expensive. Analog IP also doesn’t shrink meaningfully in size (or cost) in going to the next process technology node. Additionally, in certain product areas such as 3G/4G RF power-amplifiers, existing silicon products utilize Gallium Arsenide process, which has higher cost and provides new entrants an opportunity with lower-cost bulk CMOS process technology. However, this transition is not trivial and requires solving several technology challenges.

•  Sensor technologies: Almost every consumer product these days includes a variety of sensors such as touch sensors, motion sensors, pressure sensors, audio sensors & imaging sensors. Innovation in MEMS gyroscope technologies and a shift towards CMOS have opened opportunities for new entrants. In a number of cases, the leading process technology node is not required which keeps the required investment at an acceptable level. Moreover, incumbents delivering processors or wireless technology find it impractical or unfeasible to integrate sensors into their products.

To create startup opportunities in digital silicon chips that aim to deliver revolutionary innovation, we can learn from other industries. For example, in the software industry, a ground swell of open-source enabled startups to innovate in established segments such as OSs (to compete vs Microsoft) or databases (to compete vs Oracle). In the semiconductor industry, this will involve open-sourcing IP, developing open-source or low-cost CAD tools and provide equitable access to the fabs. Universities, government funding and individual initiatives of key engineers and leaders have to be the key drivers to make this happen. Established companies also need to confront this situation and look for creative solutions to develop revolutionary digital technologies outside of their existing roadmaps, processes and organizations. A thriving startup ecosystem across all segments is vital for a healthy, innovative, long-term future of the semiconductor industry!

My First Blog: Tech Industry Trends

After a fantastic tenure at Cavium, I recently left to work on new things. I joined NEA as en executive-in-residence and it has been an awesome experience.

It has been refreshing and intellectually stimulating to step back and review all the exciting happenings in the technology space. After a fair amount of reading, looking at 100+ business ideas and meeting with over 35 early-stage companies, I see a new level of enthusiasm that mirrors the early days of the internet. Almost all segments of the computer industry are going through meaningful change, which presents opportunities to create enormous value. Here is a summary of the high-level trends as I see them (with a bias towards infrastructure areas).

Compute, networking and storage are moving to the cloud. They are getting aggregated, shared across multiple tenants, virtualized and are delivering unprecedented elasticity. This trend is no secret, but it is amazing to see the pervasiveness of cloud xaaS startups (ie SW, IT, platform, backup... as-a-service). Moore's law is doing well and networking is catching up with 100Gb Ethernet and optical technologies implemented in creative new topologies. Customers like the idea of virtual software appliances versus purpose-built networking hardware to achieve flexibility and elasticity. Powerful multi-core processors in servers are enabling this virtual appliance concept. In the networking space, software defined networking (SDN) is in the early stages of driving a dramatic infrastructure change which will deliver unparalleled flexibility, control and cost benefits. SDN has also created new challenges/opportunities in the areas of security, services, manageability and interoperability that need to be solved. Several segments of the SDN market such as switches, virtualization, controllers and services are already in late development stages at several incumbents and new players. However widespread SDN adoption still has many hurdles to overcome and I talk more about them in my related blog on this topic.

Both servers and clients are being reshaped significantly. In the server space, there is increased convergence. Companies are developing integrated server-switch (Pluribus) or server-storage solutions to reduce cost and simplify management. Some are even more ambitious in their integration plans. Scale-out and low latency are the mantras in large data-centers. Flash memory and associated modern filesystems have become standard enablers. In the client space within the enterprise, BYOx (bring-your-own phone, tablet, application etc.) is a big wave and presents areas for productivity improvement and at the same time creates security and management challenges. In the consumer client device space, the media is enamored with the war between Android and iOS and the decline of other platforms, which makes for great news, but isn't really an area for new startups.

Enterprise client-server applications are being redefined. Software apps that ran in browsers on PC's or using proprietary methods are shifting towards mobile and wireless. The internet which started out serving static pages using HTTP with short-lived TCP connections and a simple request-response model has grown up to be the predominant link for a broad range of client/server applications. Breakthrough innovations and standards such as HTML5 and Websocket technology from companies like Kaazing enable symmetric, long-lived connections between clients and servers. This enables both push and pull of data, efficient use of the internet bandwidth, lower latency and is emerging as the proper way to connect mobile devices with legacy and new enterprise applications.

In the consumer, communication and SMB space, there is innovation in a number of areas. The mobile handset and tablet devices are impacting a number of segments including PCs, MP3 players, cameras, navigation devices and surveillance/security. Video is on the verge of exploding, as we will increasingly use video sources in phones, tablets, cars, medicine etc. Companies like Blue Jeans Networks and Altia are providing new connectivity and panoramic capabilities that will drive significantly greater video collaboration in the enterprise, home and online education markets. Vendors such as Zscaler, Aryaka and Entreda are providing IT services in the cloud for small/medium businesses (SMB) reducing capex and administration costs while delivering enterprise class capabilities. New community sites like Spiceworks provide unbiased and rapid IT support for SMBs from peers.

Analytics and big-data are big! After the success of Splunk and Guavus, there are a number of companies focused on vertical markets ranging from hospitals, retail to machine data analysis. In the future, companies will be able to monitor their resource usage and customer dynamics with a level of accuracy and insight that has never existed before, enabling better decisions and faster responsiveness.

The last mile is predominantly going to be 4G Mobile or 802.11n Wifi evolving to LTE-A and 802.11ac. WAN optimization protocols and techniques that were developed for the last mile for wired access networks dont apply well to mobile. Consequently, multiple startups are busy with innovation for mobile traffic optimization.

Security is sexy again with Palo Alto Networks' success on wall street and the upcoming IPO of Fireeye. The cloud requires rethinking security completely as the concept of perimeter security is outdated with multiple tenants and increasing east-west traffic within the data-center. Mandiant's recent report highlighting attacks on US corporations and government using advanced persistent threats (APT) have energized the startup ecosystem to develop innovative & stronger security solutions, as signature-based and honeypot based solutions are not good enough.

There is even innovation in venture investing especially in the area of consumer software. New firms are focusing on early stage consumer companies, where initial investments are low and the number of investments is significantly higher, as it is harder to predict adoption rates. Angel investors can now invest directly into very early stage startups at places such as the Angel List.

This is the time to dream big dreams and create lasting, impactful change!

Rajiv