The Microsoft–Mistral AI partnership will bring both benefits and challenges for telecoms operators

Mobile World Congress (MWC) 2024 gave vendors across the generative AI (GenAI) value chain an opportunity to present their latest developments and strategies (Figure 1). Microsoft’s announcement of its partnership with open-source AI specialist, Mistral AI, was particularly noteworthy. This partnership will be beneficial for operators that are early adopters of GenAI as well as those that are looking to scale their GenAI deployments to support more use cases. However, there are concerns that it will dampen operators’ GenAI ambitions. In this article, we summarise Analysys Mason’s views on the Microsoft–Mistral AI partnership and discuss its implications for the telecoms industry.

Figure 1: Examples of the GenAI-related announcements made at MWC 2024

Source: Analysys Mason

Microsoft is partnering with open-source AI solution provider, Mistral AI, to become the GenAI platform provider of choice

2024 is expected to be the year when operators scale out their GenAI deployments. They will need to make critical decisions regarding which GenAI platforms to adopt in order to simplify access to foundation models (FMs) and enable model deployment and management in ways that align with internal policies and existing industry relationships. Microsoft’s announcement of its partnership with Mistral AI at MWC 2024 demonstrates its desire to become the GenAI platform provider of choice.

Microsoft will host Mistral AI’s recently launched proprietary large language model (LLM), Mistral Large, on its Azure AI platform. Mistral Large supports English, French, German, Italian and Spanish. It can be accessed via the Model Catalog in Azure Machine Learning studio and Azure AI Studio.

The inclusion of Mistral Large will increase the number of commercial proprietary FMs that are available on Microsoft’s platform. Prior to this announcement, OpenAI’s models were the only available commercial proprietary models on the Azure AI platform.

The partnership will also enable Mistral AI to use the telecoms industry’s most popular GenAI platform (based on the number of publicly announced GenAI-related deals captured in Analysys Mason’s Communications service provider data and AI/analytics activity tracker) as a channel to promote Mistral Large. Being present on popular GenAI platforms is a key growth strategy for Mistral AI as it develops its commercial offerings. Mistral AI can also use Microsoft’s infrastructure to develop and grow its open-source and commercial offerings and adapt its model deployment environment to match customers’ preferences (particularly for those customers that already store and manage data on Microsoft Azure). As such, the partnership will create opportunities for both companies to grow, in terms of revenue and the number of customers.

The Microsoft–Mistral AI partnership will give operators access to more LLMs and better vendor support

Microsoft’s operator customers will gain several benefits from the partnership, including access to other commercial GenAI models besides OpenAI’s models. The reliance on a single commercial FM provider places restrictions on the type of use cases that operators can develop, especially for those use cases that are language-specific. It is generally agreed that some models will run particular categories of use cases more effectively than others, so providing operators with a choice of commercial models enables them to select the option that is most suitable for the use case that they want to address.

Microsoft’s operator customers can also potentially increase the pace at which they develop their GenAI use cases. The use of open-source GenAI models comes with cost benefits, but also brings challenges that can slow down implementation. For example, limited access to the deep AI expertise needed to fine tune these models to address unique GenAI needs can affect how quickly operators can develop their GenAI use cases. The partnership will give Microsoft’s operator customers access to additional support and other services offered by Mistral AI, which in turn may accelerate use case development.

This partnership could adversely impact operators’ GenAI opportunities if they do not take action

The Microsoft–Mistral AI deal raises some concerns that operators must address. Mistral Large can only be accessed via two GenAI platforms: Microsoft Azure and Mistral AI’s GenAI platform (La Plateforme).^1,2 Consequently, operators that use both GenAI platforms offered by other providers and the Mistral Large LLM will have a complex GenAI environment to manage. This may also make the governance of LLMs challenging. Only operators with the expertise and finances to develop, run and manage their own GenAI platforms can navigate this complexity.

This partnership could also dampen operators’ GenAI ambitions, especially those based in Europe. For example, it will raise alarm bells for Europe-based operators such as Deutsche Telekom that are building sovereign AI cloud environments and competing GenAI solutions to meet the needs of European enterprises. Indeed, Microsoft’s partnership with a Europe-based AI solution provider that offers support for multiple local languages and allows users to host models outside of the Azure environment will strengthen its ability to fulfil the GenAI needs of its European customers and prospects. Microsoft has made several other announcements linked to expanding its AI and cloud infrastructure and platform offerings in Europe over the next 2 years. It will invest USD3.5 million in Germany and USD2.1 billion in Spain to develop AI and cloud capabilities.

Operators must rethink their enterprise GenAI propositions to counter these negative effects. They should develop offerings that combine their sovereign AI cloud infrastructure solutions with connectivity services (such as private 5G networks), as exemplified by Korea Telecom and AWS and Microsoft’s GenAI platform capabilities (including its growing diverse portfolio of LLMs). This will allow operators to better position themselves as strategic GenAI solution providers to their target enterprise customers.

¹ This is a similar set up to that between Microsoft and OpenAI (OpenAI is the only other AI company that has a distribution agreement with Microsoft).

² Note that operators can access Mistral AI’s open-source models via GCP, AWS, Snowflake and NVIDIA.