Massive value can be created by leveraging data: reducing costs, generating revenue on the existing business model, and launching new business models. However, for this value to be captured, one element is key: digital infrastructure. Not only is it key for data users but it also concentrates significant value of the data economy. When looking at the value capture, it’s the infrastructure, stupid!
Data-driven value-creation opportunities are numerous
With data about their users, their machines and their distributors or suppliers, companies can explore many different value-creation opportunities.
To generate more revenue, companies can capture emerging demand through social network analysis or they can increase sales through recommendations or customization. They can also develop data services and data monetisation through data brokers.
Opportunities to reduce costs are sometimes even bigger. Productivity can be significantly improved when tasks are automated thanks to data analysis. For example, automatically detecting products missing on the shelves at Walmart. Similarly, analysing massive historical data on equipment functioning can help predict and then reduce the cost of failures.
The use cases have been described in many reports and research and the estimates of value creation are in the Trillion $ order of magnitude. It looks as if the companies who benefit from the value created by data and AI are the companies using them for improving their operations. It’s partly true and to identify which are the companies benefitting the most from the value created by data and AI, having a look at the value chain components could prove useful.
The 4 components of the data value chain
The data value chain can be split into 4 components:
- Generation: activities and assets necessary to capture and record data (structured, semi-structured and unstructured). E.g.: Web applications, ERP, IoT and connected devices, Social media, …
- Collection: activities and assets to collect, validate and store data. E.g.: cleansing, reduction, integration, storage infrastructure and models, security, …
- Analysis: activities and assets to analyse and generate insights. E.g.:Semantic analysis, models (Predictive, Descriptive, Prescriptive), visualisation (graph, maps, 3D, …), …
- Exchange: activities to expose outputs internally and externally. E.g.: decision-making, trading.
A lot of the use cases and estimates from consulting companies’ reports focus on the Exchange component, as the examples we used above illustrate. In the same direction, academic research has also proven the contribution of data/AI assets to value creation. To deliver the value, the other three components are necessary. The Generation component, i.e. the data the company has access to, is usually considered the most distinctive part and the market for data trading has soared. However, the ability of the companies specialising in this component to capture back value is limited. First companies in need of data can develop their own way to collect data and second because competition between data providers is pretty fierce. For the same type of reason, companies at the Analysis component see their bargaining ability to capture value limited by the need they have to access reliable infrastructure.
The infrastructure providers capture a significant share of the value created in the data economy
Companies in the Collection component are quite well-positioned to capture significant value. One reason is that it’s an asset-heavy business. Scaling capabilities in chip manufacturing or data centers requires significant capital which reduces the possibility for new entrants to join the market. The other reason is that demand is growing fast both from the end users (companies needing capabilities to host and run their analytics at scale) and from the companies at the Analysis component, for example OpenAI, which needs reliable infrastructure to have its models running.
To illustrate that, cloud providers are experiencing a boom both in revenue and profits. Amazon Web Services, the cloud activity of Amazon, has provided 76% of Amazon’s operating income over the past decade. Route planning, pricing analytics, research and development run on demanding Machine Learning models supported by cloud providers.
Another illustration is the current frenzy about chip manufacturers. Arm shares soar almost 50% when the SoftBank-backed company raised its earnings outlook as it posted revenue of $824mn for the three months to the end of December 2023, up 14 per cent year-on-year and surpassing analysts’ expectations. The juggernaut Nvidia saw its shares rise five-fold in little more than a year. This value increase is supported by exceptional revenue growth triggered by demand from Generative AI specialists.
To maximise their value capture, companies at the Collection component develop partnerships in two directions.
First, with Generative AI companies. In the recent deals inked, cloud and chips providers poured capital into OpenAI and Anthropic for example which agreed to use technology such as chips and cloud computing services from the companies that invested in it. Part of the amount invested returns in forms of revenue. When Google invested in Anthropic, Anthropic also agreed to buy computing power through Google’s cloud computing service, which it uses to train and serve its technologies. Microsoft has poured $12 billion into OpenAI which has spent most of the money on Microsoft’s cloud services.
The second direction of partnerships is with non-digital native companies which have matured their digital transformation. With the spread of digital technologies, infrastructure-related assets and capabilities are critical now in automotive or equipments manufacturing for example and companies such as Renault and Schneider Electric have signed significant partnerships with Google, Amazon or Microsoft to access their cloud and computing technologies.
Implications
Growing market, growing competition in the Collection component. With the market growth expected, supported both by AI-specialised companies and incumbents, companies at the Collection component will probably act collaboratively to support any initiative which increases the demand for their infrastructure (for example investing in companies which run on their infrastructure). At the same time, they will probably make it collectively harder for new entrants to compete with massive investments in capacity which will increase the barrier to entry. Last, the competition between them will be quite intense both on the client’s side and the supplier’s side.
Opportunities at the Analysis component. As data generation continues to grow on a steep curve and infrastructure to analyse them scales, opportunities to develop and run models are bigger. Additionally, as we described, companies at the Collection component are likely to support all initiatives increasing the need for their infrastructure. Last, promises of value creation at the Exchange part from specialised Analyst support demand.
Value evasion in the Exchange component. The value creation opportunities at the last component are likely to grow even bigger. Better technology, cheaper access, easier integration will support the scale of solutions helping companies to reduce costs and develop revenue. However, a significant part of the value created there will be transferred back to Analysis companies (the Business segment of OpenAI is much more dynamic and lucrative than the consumer one) and to Collection companies for accessing to their infrastructure.
It’s particularly interesting to notice that we are observing here a change in value capture from digital players. For the last twenty years, digital players have exerted control at the downstream part of value chains, engaging directly with end-customers (search engines, marketplaces, applications,…). They challenged incumbents adding a layer between them and their customers.
What we observe now is a capture from the upstream parts of value chains, digital companies becoming suppliers of incumbents for a critical component of their operations. Incumbents have obviously some cards to play and they are used to deal with suppliers and exert control over them. We will see how they react.
Photo de Jordan Harrison sur Unsplash
