Value capture from tech player
Generative AI companies are leveraging several strategic mechanisms to capture value from incumbents and reshape the traditional business landscapes. One of these mechanisms is becoming the technological supplier for incumbents.
Given the highly specialized nature of Generative AI, companies in this space have invested significant resources into developing, training, and refining their AI models. This, in turn, has allowed them to generate advanced technologies that are often too costly or complex for incumbents to produce in-house. Consequently, many incumbent firms find it more cost-effective and efficient to source this technology from AI firms rather than develop it themselves.
By becoming a technology provider, Generative AI companies can capture a significant portion of the value chain. They can charge premiums for their advanced technologies, ensuring a steady stream of revenue. Additionally, this role can provide them with access to valuable data and insights, further enhancing their capabilities and competitive advantage.
On the other side of the spectrum, Generative AI companies can offer new, sophisticated ways to deliver incumbents’ offerings to the market. Take the case of generative AI in the retail sector, for example. Traditionally, retail incumbents have depended on physical stores, e-commerce websites, and recently, mobile applications to distribute their products. But with the advent of Generative AI, new avenues for distribution are emerging. OpenAI has for example opened for plugins development. Companies can develop tools which make their offer visible on ChatGPT. For example, a user on ChatGPT could ask the system for recipes and thanks to the Instacart plugin order automatically the necessary ingredient from the retail platform. OpenAI becomes a sort of distributor for Instactart offer. More prospectively, a Generative AI company could create highly realistic virtual shopping experiences, transforming the way consumers shop. By distributing retail products through these advanced platforms, AI firms can enhance customer experiences, drive greater engagement, and ultimately increase sales for incumbent retailers.
Moreover, by becoming a distributor, AI firms can establish direct relationships with end-users, allowing them to receive feedback, fine-tune their offerings, and build brand loyalty. Over time, these relationships can help AI companies drive the adoption of their technologies, leading to a greater market share and more sustainable revenue streams.
Lastly, the dual strategy of becoming both a supplier and distributor can allow AI companies to capture a larger portion of the value chain. This can potentially increase their profits, strengthen their market position, and give them a greater influence over the direction of their industry.
A battle within the tech stack
The landscape of Generative AI is proving to be a fertile and dynamic battleground, with a wide array of players vying for supremacy. The generative AI ecosystem comprises diverse players, each with a unique role. Data owners, for example Reddit for text and Getty Image for pictures, hold vast troves of user data that provide the raw material for AI training. Model providers like OpenAI, the creator of GPT-4, design and train AI models that transform this data into insights and actions. Cloud solutions such as Amazon AWS, Microsoft Azure, and Google Cloud provide the infrastructure necessary to run these AI models at scale. Finally, applications serve as the interface between AI and end-users, converting AI outputs into actionable recommendations or automating tasks.
In this intricate ecosystem, value capture often sparks contention. While data owners might argue their data’s crucial role warrants a substantial portion of the value generated (see how Reddit claims for a fairer share of value and Getty sued Stablity AI for using its data without permission), model providers could counterclaim their sophisticated algorithms are the real value drivers. Cloud solutions and applications, being the enablers of access and usability, also present strong cases for significant value capture.
Though they could appear separated, the 4 layers of the stack (data, models, cloud and application) are connected and hence deciding on a positioning on these 4 layers is a strategic decision. For example, data and model work jointly, the more data, the better the model. It’s been described as a data flywheel effect: more user engagement means more data which means better performance of the model, then more user growth, hence more data. Some companies would decide to leverage this effect and position at the two layers. Tech giants are positioning on the 4 layers. Google recently announced Generative AI would be introduced into its advertising business.
For the moment, the value is mainly captured by infrastructure and cloud providers as they monetize their services to model developers and applications. Surprisingly, the companies which create value through models haven’t captured it yet and applications companies struggle with retention and differentiation and spend 20-40% of their revenue on inference and fine-tuning.
Entry strategy by startups
Startups are carving out their own space in this field by focusing on niche applications, leveraging agility, and driving innovation. For example, startups like Grammarly leverage AI for real-time grammar checking and writing enhancement. Their market entry strategies often revolve around delivering unique value propositions, swift adaptability, and creating symbiotic relationships with larger players. In particular, partnering with a cloud company enables them to access better this critical resource (Google invested in Anthropic). By focusing on specialized applications, they capture value by addressing gaps overlooked by tech giants and creating unique customer experiences.
Investor Gordon Ritter suggests that specialised software with deep data and narrow context will create the most value in the Generative AI era. According to this view, startups should focus on building in specific domains with a narrow context or choose to major in Generative AI’s emerging technical capabilities and hunt for a function or vertical problem that benefits from their insights.
Beyond value proposition, user experience and scope of data used, startups build their market entry strategy choosing a position in the closed-source / open-source continuum. Some rely on closed approach monetising through API (for example OpenAI with ChatGPT), others rely on open approach and have service oriented revenues (for example Stability AI for Stable diffusion model). Choosing between closed and open source is mainly a tradeoff between diffusion and appropriability. The more open, the higher the diffusion and the lower the appropriability. Open source models catch up rapidly in terms of performance and are much less costly. A $100 open-source model with 13 billion parameters is competing with a $10 million Google high-end model with 540 billion parameters. A growth of open source models would mean a much smaller moat for Google and OpenAI who would be competed by much cheaper alternatives. The impact of the distribution choice (open source vs API) will be important for the businesses using these digital services. Using open source requires additional efforts (and costs) on the user side compared with API but offers more independence. API services will be preferred for convenience and open source for independence. According to how costly will be the API services, some users may find strategically interesting to support open source initiatives (the same way Amazon, Facebook and Apple supported OpenStreetMap when Google maps changed the terms for using the API).
A battle between tech giants
The rivalry between Google and Microsoft serves as a prime example of the struggle for dominance in AI. Microsoft’s acquisition of Nuance Communications, a cloud and AI software leader, as well as its investment in OpenAI highlight its aggressive push into AI and cloud solutions. In contrast, Google’s investments in AI-first initiatives, like its DeepMind division and the development of Tensorflow, an open-source machine learning platform, underscore its commitment to leading this space.
These tech giants are locked in a fierce battle for value capture, each leveraging its unique capabilities. Microsoft banks on its enterprise relationships and comprehensive software ecosystem, while Google leverages its vast data resources and leadership in machine learning research. Amazon could focus on hosting on its cloud other companies’ models. With the high entry barriers to entering the market, the concentration risk is significant.
How to prepare
Incumbents are under attack on two main fronts. First, Generative AI companies compete directly with them with a renewed value proposition or operating model. Second, they are facing value capture on their value chain by tech companies which become providers of Generative AI solutions and as they act as distributors of products and services.
To navigate into this new battlefield they might consider the following actions.
1. Invest in Understanding Generative AI: The first step for incumbents is to develop a thorough understanding of generative AI and its potential impact on their industry. This can involve hiring AI experts, training existing staff, or consulting with AI firms. Leaders need to understand the capabilities and limitations of AI, as well as the opportunities it can offer and the threats it can pose.
2. Align Generative AI with Business Strategy: AI should not be seen as just a tool, but as a strategic asset that can support and enhance the business’s overall strategy. Incumbents should identify how AI can help them achieve their business objectives, whether it’s improving operational efficiency, enhancing customer experience, or creating new products or services. This would also include identifying when they prefer not to use Generative AI to build a distinctive positioning.
3. Develop or Acquire AI Capabilities: Incumbents need to build or acquire the necessary capabilities to implement and manage AI. This can be done internally, through partnerships, or via acquisitions. Companies like Salesforce and Microsoft have chosen to acquire AI startups to fast-track their AI capabilities. In finance and mobility we see partnerships being announced between incumbents and Generative AI specialists. Make/Buy/Partner decisions are strategic ones and each represents some tradeoffs.
4. Build Ethical AI Practices: With the increasing usage of AI, issues around data privacy, security, and ethical use are coming to the forefront. Incumbents need to build robust ethical AI practices and ensure they are adhering to relevant regulations. This can help prevent potential legal issues and protect the company’s reputation.
5. Foster an AI-Ready Culture: Incorporating AI into a business requires more than just technical changes – it requires cultural shifts as well. Incumbents should strive to foster a culture that embraces innovation, is comfortable with experimentation, and is prepared for the changes that AI can bring.
6. Transform the processes to become AI-fit: To reap the benefits of automation through Generative AI, processes and organisation must be transformed. It may concern structure, processes, job description or training.
Preparing for this new context includes first being strategic in deciding on how to use (and not to use) Generative AI ; second implementing actions to access capabilities and third engaging in a cultural and organizational transformation.