There is a particular moment in the life of a platform technology when the question shifts from whether to engage to how much does it cost. For enterprise AI, that moment arrived sometime in the second half of 2024 – quietly, without the fanfare that had accompanied every earlier announcement in the cycle – and the market has been adjusting to its implications ever since. What changed was not the technology itself. The large language models that underpinned the wave of enterprise AI products in 2023 were, in capability terms, already capable of genuine production work. What changed significantly was organisational confidence. The first cohort of serious enterprise deployments, of actual workflow integrations tied to measurable business outcomes, produced results that procurement committees could defend. And once that happened, the budgets followed. We are now in the phase that comes after adoption stops being a decision and becomes a competitive obligation. The CIOs who were cautious in 2023 and 2024 are not cautious anymore. They have fallen behind.
The practical consequence of this shift has been felt most acutely not in the data infrastructure beneath the AI model layer. Enterprise AI is, in the end, only as useful as the data it can access, and the data landscape inside most large organisations is, to put it charitably, not ready.
We are looking at data quality, data lineage, access governance, observability and the sheer problem of making structured and unstructured data legible to AI systems at the speed those systems require. These are engineering problems of genuine depth, and they do not resolve themselves. The companies that have been working on them for five or six years, often without much attention from the investment community, are now extraordinarily well positioned. Their customers have, in effect, arrived.
This is one of the more interesting asymmetries in the current market. The AI model companies have attracted enormous attention and capital. The infrastructure companies that make those models usable inside real enterprise environments have attracted considerably less of both, and are now, in our assessment, among the most defensible businesses in the technology sector.
Not every company benefiting from the current deployment wave is building something that endures. This distinction matters enormously in pre-IPO investing, where holding periods span the full arc from early enterprise adoption to public-market debut. The companies that Hollands Team finds the most compelling share a specific characteristic: their products become more valuable and more structurally capable as the customer's data and workflow patterns accumulate within the platform. This is compounding in the most literal sense. The switching cost is the degradation of a system that has been trained, over years, on a specific organisation's operating reality.
This quality is found most reliably in data infrastructure and in enterprise software built around continuous learning loops. The products where the value is not static at the point of sale but grows through use. It is less common, and more fragile, in standalone AI applications where the underlying model can be replaced by a competitor's equivalent.
The deployment wave will produce winners and it will produce casualties. The distinction between them is, in our view, a function of whether the product's position within the enterprise compounds or merely persists.
The pre-IPO technology market in 2025 and 2026 has been characterised by a significant bifurcation. Companies with the characteristics described above – recurring revenue, strong net retention, a data or workflow moat and credible progress toward profitability – are attracting valuations that reflect institutional confidence in their public-market trajectories. A second group, comprising businesses with AI positioning but without the underlying structural advantages, is finding the private market considerably less forgiving than it was two years ago.
This bifurcation is, in our view at Hollands Team, the correct response to the evidence. The question for investors is whether they have the analytical framework to sit consistently on the right side of it. And also whether they entered the relevant positions at a point in the company's development where the current valuation still leaves room for the returns that justify the illiquidity. We believe the most interesting opportunities in the current market are in companies that passed through their fastest growth phase in 2022 and 2023, are now approaching breakeven or profitability, and are in late-stage private rounds ahead of what will, in the next eighteen to thirty-six months, be a materially more active IPO window than the one investors have endured since 2022.
The infrastructure of the next enterprise era is being priced now, in private markets, by a relatively small number of investors with the patience and the framework to evaluate it properly. The public markets will form their own view in due course.