Venturing Into a New Era of AI

In 2016, Fairview published its first report on venture capital investment opportunities emerging as a result of advances in machine learning and artificial intelligence (“AI”) technology. We have been significant investors in the category by building exposure through our partnerships and direct-investment activity. By almost any measure, our experience has been successful. Entrepreneurs have built innovative and highly valuable companies across the infrastructure, platform, and application layers of the ecosystem. The past decade was one in which AI truly came of age.

As we enter a new decade, AI has evolved to become table stakes for almost any technology company. Data sets around products and services, and models that support personalization, improvement, automation, and intelligent insights are now a must-have in enterprise and consumer applications. The increasing ubiquity and advancement of the technology will be sustained by the continuation of transformative trends in areas such as digital data, cloud computing, edge computing, and quantum computing. Advances in AI are expected to continue relatively unabated through the COVID-19 pandemic. In fact, AI has played a major role in the response to the pandemic. It has been used to model cases and spread, to diagnose and treat patients, and to aid in the search for a vaccine.

The next generation of AI technology and its application remains one of the most important singular market opportunities for venture capital. While there are clear limits to what can be accomplished with today’s technology, the technology is tangible, and research, development, and applications continue to accelerate rapidly. As the AI market has evolved, the skill sets, networks, and due diligence requirements investors must employ have changed. The ability to navigate hyperbole is critical. The investment opportunity set for limited partners has also changed. There are now more active funds in the category than ever before, including a new crop of specialized managers. Perceptive investors who can identify where the true opportunities lie within the continued innovation and application growth in the space will be advantaged.

Developments in AI

Since Fairview last reported on the AI market, there have been major advances throughout the ecosystem.  We have witnessed an increase in speed and accuracy and a decrease in costs. For example, the time to train a machine learning algorithm on the ImageNet dataset, a database commonly used for visual object recognition software research, fell from about three hours in 2017 to 88 seconds by the end of 2019. Further, the costs to conduct that training fell from thousands of dollars to tens of dollars. Until 2012, AI results closely tracked Moore’s Law – compute, the amount of computation used in the largest AI training runs, was doubling approximatley every two years. Post-2012, AI compute has been doubling every 3.4 months, a net increase of 300,000x. [1]

From a talent perspective, AI has now become the most popular specialization for computer science graduates in North America, with over 20% of graduating computer science Ph.D.’s specializing in AI. Over 60% of those graduates are entering industry, up from 20% in the last decade.1 Despite this increase, there is strong evidence that the supply of top-tier AI talent does not meet the demand; but this may change soon, as interest in the field continues to increase. For example, applicants to study artificial intelligence in UC Berkeley’s electrical engineering and computer sciences doctoral program numbered 341 a decade ago but by last year, had surged to 2,700.[2]

As AI has become more mainstream, technology layers have become better defined and are more robust. AI infrastructure is more widely accessible and increasingly optimized around the management, storage, and compute of data, the key inputs for AI applications. AI enabling technology has emerged in the form of tools that support the development of AI-based applications, particularly around content intelligence, computer vision and natural language understanding. Developments continue to be made on the infrastructure and enabling technology fronts with companies such as Amazon, Google, IBM and Microsoft serving as the largest players, but also complemented by numerous focused startups as well many open source tools. We have also seen the emergence of Artificial Intelligence as a Service (AIaaS) offered by larger companies and newer startups in the space, providing companies off the shelf AI solutions to experiment, test or build products. The benefits are similar to SaaS and IaaS – transparent costs through pay as you use structures, reduced development time, and increased flexibility.

With AI infrastructure and tools more readily available to entrepreneurs, the scope of AI applications has increased. Applying AI represents the broadest opportunity set and the largest markets for investors. While autonomous driving has received the most investment historically, growth in recent years has come across horizontal and vertical applications such as robotic process automation (“RPA”), supply chain management, and industrial automation. These new or historically untapped markets are now much more accessible for AI software providers, creating growth opportunities for AI applications, particularly for repetitive tasks and predictable roles. Notably, a common theme is that, instead of replacing humans, the most effective automation aims to augment and improve human performance.

Venture Capital Investment in AI

Venture capital investment in AI companies, particularly those at the application layer, is difficult to measure accurately since so many startups are now incorporating AI into their products or services. Industry data sources do not fully capture the full impact AI is having on the venture market, and primarily capture AI infrastructure, AI enabling and AI-first application companies. However, the data can be a good indicator of the directionality of the effect AI is having on investment activity and an indicator of the key trends for investments in areas that serve as the foundation for growth in AI applications.

Broadly, AI has unquestionably given rise to a major new cohort of investments. Investment in AI-related companies has accelerated, particularly over the past five years, over which time the total dollars invested has increased at a compounded annual growth rate of nearly 80%. Deal activity is also up, with over 1,000 new AI-related venture capital deals in each of the past two years. AI investment is also expanding relative to the venture capital market. In 2019, AI investment accounted for 13% of all venture capital deals and 18% of all venture capital dollars invested, up from 4% and 9%, respectively, in 2014.

The average deal size and valuations for AI companies fluctuate, as the market is still relatively nascent. However, evident from the data is that AI investments are receiving more capital than the typical venture capital investment. Average deal sizes for AI companies tend to be higher for two primary reasons: (1) the companies tend to be slightly more capital intensive, especially at the infrastructure and enabling layers, and (2) more AI investment is occurring at later stages.

Later stage investment in AI companies has been noteworthy. Even though there is significant new company formation in the space, most of the capital has been flowing to mature AI companies. Emerging winners in this category have attracted capital because of their ability to capture substantial value as the infrastructure and enabling layers become more established. Early exit activity has supported this trend, with several large acquisitions in the space. For example, Intel recently acquired AI chipmaker Habana Labs for $2 billion in late 2019. Habana Labs had raised capital from firms such as Battery Ventures and Bessemer Venture Partners. The top acquirers of AI companies have been large tech companies such as Apple, Google, Microsoft, Facebook and Intel, primarily at the infrastructure and enabling layers.[3]

Aggregate acquisition volume though has been higher for application companies, with retail, healthcare, cybersecurity and finance companies the top acquirers of verticalized companies providing advanced AI-driven analytics, management and personalization services. 

Investing in AI – the LP Perspective

AI is an increasingly important component of venture capital portfolios, and growth in the category means that greater exposure will likely come naturally to most active investors. However, as with any aspect of venture capital investing, being proactive and navigating the market with a prepared mind is critical to long-term success. There are important nuances in AI technology and business models that venture capitalists need to consider when evaluating companies, which often requires specific skillsets and networks. Limited partners must become adept at evaluating these skillsets and networks when investing in venture capital funds, and may need to adjust their approach to portfolio construction to develop optimal exposure.  

Most AI companies, particularly at the application layer, resemble software companies and feature user interfaces and other elements built with traditional code. The unique and most valuable components are the algorithms and trained data models that drive the applications – the “AI.” Data models can be complex to maintain and may even require customization for each customer, which can increase costs. As a result, AI business models are different and can be heavier on services than traditional software companies. A company’s ability to productize the service element can become a major advantage. Speed to value is also important since it can be beneficial to have customers quickly realize the value add of a service relative to a growing set competitors. The service element may also make it harder for companies to expand across verticals.

Infrastructure costs are typically greater for AI companies relative to traditional software. These variable costs are more difficult to understand and predict for growing companies and different AI companies can feature different infrastructure requirements. Additionally, data is a very valuable input for AI companies – access to high-quality data to train models is critical. However, contrary to early beliefs, proprietary data may not always serve as a long-term competitive advantage, as data becomes more commoditized and training becomes more advanced. The importance of data will vary depending upon the company’s overall strategy and market dynamics. Finally, the engineering talent required to build AI models, despite the growing interest in the field, can be hard to find and is expensive. Finding, attracting and retaining qualified engineers to build true AI companies is difficult and the companies with the best talent have significant advantages.

Venture capitalists must be able to assess the aforementioned factors when making investments in the category. Technical knowledge is helpful in assessing the viability and distinctiveness of data models and the training data itself. Operational experience helps in understanding new business model elements and scalability. Often, firms will do heavy thematic research in these areas and may hire former engineers or entrepreneurs. Sourcing and assessing the right teams often requires an expansion of networks for most venture capitalists – the AI category often features many first-time entrepreneurs, experts often come from universities and have PhDs, and talent is spread globally. Some firms will add team members with strong networks in these areas. Firms that have been thoughtful in their approach to AI investing will be advantaged in sifting through much of the hyperbole that currently exists in the market and will be better positioned to assist their portfolio companies as they grow.

When investing in venture capital firms pursuing AI investments, limited partners should consider the relevancy of the firms’ experience, the technical and operational expertise resident at the firm, and its networks. The level and type of exposure, in terms of stage and AI technology layers, should also be considered. The approaches firms take can vary significantly and limited partners have several options for developing exposure to the category:

Multi-Strategy Firms: The most active firms in terms of the number of deals and dollars invested in the AI category have been large multi-strategy firms that invest across stages and strategies. These firms have developed investment theses, and sometimes teams, around AI or have viewed AI as an extension of enterprise investment activity. Since their fund sizes are often larger and investments scope broader, these firms have invested across AI technology layers and have been the most active in funding infrastructure and AI enabling companies. Examples include Battery Ventures, Bessemer Venture Partners, and Kleiner Perkins.

Enterprise Firms: Investing in AI has been a natural extension for most venture capital firms investing with an enterprise focus. These firms are already well-positioned to invest in the category and often take highly thematic approaches. For example, Emergence Capital Partners, one of the leading enterprise venture capital firms, believes the future of the enterprise will be defined by Coaching Networks – AI learning loops that continuously identify specific, hyper-effective strategies and distribute the best performing advice to the rest of the network. In their thesis, AI is not a threat that replaces employees but is an asset that helps employees unlock their full potential.

AI-Focused Firms: In recent years, we have witnessed a rapid increase in the number of new firms specializing in AI, particularly applied AI at the seed stage. Many of these firms lack experience but the best to emerge are likely to do well as the category is large enough to support specialized firms. One of the more mature firms in the category is Zetta Venture Partners, which has a focus on AI-first companies with B2B business models.

Fairview has built portfolios that include firms that fall into each of the above categories. We have also co-invested directly into multiple AI companies. Two recent examples include a Series A investment in a company that drives management insights through AI and a Series B investment in a robotic process automation (RPA) company.

Most limited partners will have exposure to multi-strategy firms investing in the category, but we believe that focus is valuable when it comes to AI and that the market opportunity is significant enough to warrant the inclusion of specialized managers. As companies in the category scale, we have also found more co-investment opportunities emerge, particularly with companies that have quickly proven their technology and rapidly built a customer base.

AI technology is accelerating behind a maturing ecosystem and its applications will be far-reaching (perhaps even affecting the future of the venture capital business itself), and the end markets are major.  Some of the most significant venture-backed companies in the future are likely to be AI companies. Remaining informed and building strategic exposure to the category will be vital for limited partners with sophisticated venture capital portfolios.

[1] Raymond Perrault, Yoav Shoham, Erik Brynjolfsson, Jack Clark, John Etchemendy, Barbara Grosz, Terah Lyons, James Manyika, Saurabh Mishra, and Juan Carlos Niebles, “The AI Index 2019 Annual Report”, AI Index Steering Committee, Human-Centered AI Institute, Stanford University, Stanford, CA, December 2019

[2]Saphir, Ann. “As companies embrace AI, it's a job-seeker's market.” Reuters, October 15, 2018 Web. March 5, 2020.

[3] CB Insights 2019 AI Trends Report