M&A within GenAI space

It is often said that many acquisitions fail – discontinued products, disbanded acquired teams (often due to cultural incompatibility, or poor integration), and overpayment.

The reasons for acquisitions are typically access to IP/technology (e.g. Infinite Band by Nvidia), expanding user bases (WhatsApp by Meta), or acquiring talent (DeepMind by Google). Acquisitions are pursued for expansion or diversification in market or product offerings, or sometimes to eliminate competition.

Big companies – the potential acquirers – have always faced the choice of developing internally or acquiring smaller firms. But is it any different for GenAI? Possibly. GenAI-based products replace older versions, which were often built at great expense, creating resistance from their original creators. Some GenAI applications, such as AI-based companions – some of which fall under the adult industry – do not align with any established products. While GenAI demos often impress, they are not adopted, due to e.g. sporadic but significant errors, known as "hallucinations."

So, what is happening in M&A within the GenAI space?

The main issue is the high uncertainty of valuations, which are based on promises and bug assumptions: hallucinations will go away soon; and users will integrate AI into their workflows and routines.

Secondly, the space has become crowded; for any given idea, there are multiple startups pursuing it, and each startup’s valuation assumes that they will capture the majority of the market.

Additionally, the access of open-source AI models and cloud-based inference-ready GPUs has lowered the barriers to entry, meaning many startups have shallow or no moats. This only adds to the uncertainty about who, if anyone, will emerge as a winner.

Perhaps potential acquirers are adopting a wait-and-see strategy while simultaneously reimplementing the most promising ideas themselves. They can afford to be late to the game if their user base is large enough.