"It’s a very fragmented space, especially when you consider the full gamut of AI types, which extends far beyond the headlines’ focus on generative AI tools. In the GenAI arena, as the rapid arms race in model development continues, we’re seeing a wide range of industry-targeted or use case-specific offerings come online that build on top of a model offered by one of the leading developers (often with flexibility, by design, to substitute an alternative model if the need arises). I expect that trend to continue. At the portfolio company level, we’re also helping businesses of all kinds grapple with whether and how to incorporate AI to remain competitive and drive strong valuations in a sale process, while remaining nimble enough to adapt to further tech advances. On the due diligence front, we’ve seen increased scrutiny on whether companies are practicing good GenAI hygiene and putting appropriate policies and guardrails in place regarding confidentiality, IP rights, privacy, cybersecurity, bias and legal compliance, including as private equity and strategic acquirers continue to frequently rely on representation and warranty insurance and underwriters become increasingly attuned to the potential implications of AI use."

"Foundation model developers have understandably received a lot of attention from investment professionals. However, in the medium to long term, I expect there to be only a small handful of meaningful players in the foundation model market and only a small number of businesses with an opportunity to invest in them. I am therefore seeing attention shift to three other target types. First, we have intermediaries who add a valuable proprietary element, such as industry-specific knowledge, to a foundation model via fine-tuning or retrieval augmented generation. Second, we have non-GenAI developers, such as those offering discriminative AI models that can categorize expenses and bills. These models can be more cost efficient than generative models for certain tasks and can still plug into generative models for composite outputs. And third, we have ‘picks and shovels’ and ‘complementary targets’ — such as providers of data centers, data labelling, and security against GenAI inversion attacks and model poisoning. Targets like these will ultimately sustain the AI industry's expansion, offering stable and significant value beyond the spotlight of the latest foundation model breakthroughs."

"Proof of concept in AI is high risk compared to established technology platforms. While discussions about AI in M&A transactions are increasing, these tend to be in niche areas. Companies are more focused on adopting AI for portfolio use rather than acquiring the technology itself. Where I think we may see activity in areas that are adjacent to AI. We are seeing a notable increase in investment in data centers to support front-end AI engines. Large cap sponsors see this as a growth area needing more infrastructure due to the rise in machine learning and AI adoption. While technology investments face scrutiny and valuation gaps, large corporations lead the market, potentially driving more direct investments in AI platforms – a kind of virtuous circle."