91̽»¨

Depending on their position on the generative AI value chain, companies will garner interest from different sets of third-party investors. But the businesses that will succeed in raising capital in this emerging area will be those who develop and execute their strategies while keeping these potential investors’ interests in mind.

Capital embracing generative AI…in certain pockets

There are shining examples of capital being allocated to , and wealth being created in, the initial stages of the generative AI value chain, the ‘producing’ end, including Microsoft’s $10bn+ investment in OpenAI and the 236% share price rise of chip manufacturer Nvidia, for 2023 to end August. Further examples are those providing data warehousing and platforming space, like Databricks which is planning a funding round that would value the business at $43bn (it reported annual revenues of $1bn in its latest year to January 2023, up 60% from the previous year), and ARM’s recent IPO which was five times over-subscribed.

That end of the value chain is capital intensive and until recently very speculative – an area primed for extremely well-capitalised firms that can afford to make those big bets, like Microsoft, Google and Softbank, with the latter being the backers of ARM.

However, as generative AI technology matures, we can expect capital to begin flowing into the ‘adoption’ end of the value chain. This will include companies building applications and offering services, helping firms employ the technology, as well as firms that are outcompeting others through its adoption. This end is less capital intensive, less speculative, as they will be built on maturing models, and likely to be populated by a wide range of providers by type and size.

Critically, we believe that the largest opportunity lies in applications. This is where generative AI engages with businesses, making them faster, better and stronger, for a fee. A swarm of applications will be developed – we see them appearing already on our social media feeds – for a whole host of end markets. The extension further along the value chain is occurring fast, meaning it won’t take long for meaningful critical mass to appear there. ChatGPT is one of example of an application, built off OpenAI’s models, as is Bing Chat.

Currently, 70% of those actively using or experimenting with AI are using freely available models and related applications, while only 46% had purchased third-party solutions and 43% are using internally built solutions (respondents could select multiple options).

Nevertheless, once companies can build the skills needed to adopt this technology and the applications, redesign processes and embark on change management – they will probably need to buy services to support them. We expect spend and the related investment opportunity in this area to rise significantly as companies recognise both the opportunity that these apps represent, and the threat posed by competitors that have adopted them.

The generative AI value chain – who gets what?

As you can see below, in a summary of the generative AI value chain and the related interest areas of third-party equity capital providers, private equity is most likely to invest at the ‘adoption’ end of the value chain, while VC and corporate venture towards the ‘producing’ end. ‘Transformed adopters,’ however, will be attractive to all investors, especially while the technology gives them a competitive advantage.

Let’s take a further look into how capital may be divided.

Capital markets will focus on infrastructure and models to start

Though not currently active, capital markets will initially focus on infrastructure and models. When the markets open, we expect the interest to lie within infrastructure, and move into the adoption enabling end of the value chain (applications), and in those that have best adopted the technology and are able to out-perform better than their competitors.

There are, however, likely to be some exceptions, including the IPO of ARM, and KKR-backed Hitachi Kokusai Electric (chip equipment maker).

Private equity will focus on adoption

The risk profile of PE suggests this capital is unlikely to invest in model development, but there’s evidence it will be attracted to the stable annual recurring revenues of hosting and platforms (mostly the domain of the larger PE investors who can provide the capital for their big-ticket infrastructure projects), and on the adoption end of the value chain, namely on applications and services.

In fact, software and commercial services are the two most active sectors for PE by deal count, so this is not new ground for private equity. PE is drawn to the high margins, recurring revenue profile and high growth potential typically associated with applications and services so companies that can set themselves up in this way will be able to draw in this capital pool, which is large at $1trn of ‘dry powder’ globally.

Outside of the value chain, we expect PE to be heavily drawn to businesses that can leverage the technology to disrupt industry incumbents, improve decision making, increase efficiency, optimise operations and enhance their resilience.

Corporate venturing focused on applications

Coupling high levels of cash reserves and an increasing propensity to invest in R&D among larger corporate with venturing arms, we expect corporates venturing to increasingly make investments in new models and application development, for their own use or to commercialise them for third parties.

Venture capital focused on mode

VC activity has been down over the last 18 months, with a cooling of interest for cryptocurrency and blockchain. However, AI and machine learning has remained a bright spot, with investment in this area in H1 2023 up by 42% compared to H1 2022 (Pitchbook Data Inc).

The investments to date have mostly been in model developers, but we expect activity to grow in applications as the related target markets begin adopting the solutions. Example model developers include France-based Mistral that raised $113m in seed funding in June 2023 which was Europe’s largest ever seed round. It valued the fledgling business at $260m. It was four weeks old at the time, but worth mentioning that the founders were ex Google Deepmind and Meta.

VC’s high risk/high return investment profile makes it well-suited to an emerging area like this. However, VC fund raising activity is under pressure, (60% down in H1 2023 from H1 2022 on a global basis according to data from Preqin) so companies looking to engage with these investors may find them to be very selective.

Raising capital can be a daunting process, especially in an environment of high cost of debt and economic headwinds. But for those businesses able to match the opportunity of generative AI with their expertise, they will find an audience of investors that will lean into them. Who those investors will be will depend on the capital requirements, the degree of certainty, reliability of revenue, margins and growth prospects. It might be difficult in this nascent stage of the technology, but the more companies can set themselves up to tick those boxes, the more interest they will gain and the more likely they are to secure an investment partner that’s ideally suited to them.