AI Investment H1 2026: Rounds + Valuations

The H1 2026 AI investment retrospective tells a different story than the headline numbers suggest. Capital is concentrating, not diversifying — and the concentration is happening one layer up the stack from where most observers were looking. Agentic infrastructure attracted the marginal venture dollar that, twelve months earlier, would have gone to foundation-model labs.

That rotation matters because it changes which categories defend valuation through H2 and which compress. The four trend lines in this report are the most reliable forward indicators we have for the back half of the year, and the geographic and sector composition of H1 capital flows tells operators where the procurement budgets land next.

This retrospective covers what shipped in capital terms across H1 2026 — headline rounds by sector, valuation medians, the four sector concentrations that defined the period, geographic distribution, the four governing trend lines, and a scenario-bracketed projection for H2. Read it as planning input for positioning, vendor selection, and competitive analysis.

Venture capital allocation is a forward indicator for the capability landscape twelve to eighteen months out. Companies funded in H1 2026 are the categories operators will be procuring from across H2 and into H1 2027. The retrospective is therefore not a backward-looking artefact — it is the procurement roadmap, derived from the allocation choices of a group with strong information asymmetry on commercial traction.

What changed in H1 was the composition of the marginal dollar. In 2024 and early 2025, the marginal venture dollar funded foundation models — the bet that scaling laws would produce continued differentiation. In H1 2026, the marginal dollar funded agentic infrastructure — the bet that orchestration, memory, evaluation, and tool integration would be where margin accumulates as the underlying models commoditise. That rotation is the most important signal in the period and the one operators should read closely when sequencing vendor selection.

A second compositional change deserves attention: capital began distinguishing between agentic infrastructure (horizontal tooling) and vertical AI (industry-specific applications). Both attracted capital in H1, but at different multiples and with different implied exit narratives. Operators evaluating vendors should understand which side a given company sits on, because the failure modes differ.

The fifty-plus rounds we tracked cluster into four sectors with distinct capital signatures — round-size profile, stage mix, and implied exit narrative. Below, each sector is summarised with its most representative deal pattern. We name patterns, not specific companies, because the procurement read is more durable than any single deal headline.

The pattern across sectors is consistent on one point: capital preferred companies with a clear, defensible insertion into an existing workflow over companies competing on raw model capability. That preference is itself a rotation — earlier cycles rewarded capability-led positioning. H1 2026 rewarded distribution-led and workflow-led positioning, with capability assumed as table stakes.

Valuation medians across H1 2026 told a two-sided story. Early stages — Seed and Series A — held firm or modestly expanded. Growth stages — Series B and later — showed visible compression against 2025 comparables. The split is the clearest empirical signal of capital discipline returning to the category.

The chart below tracks reported post-money valuation medians by stage across H1 2026 versus H1 2025 baseline. We index against the 2025 baseline because absolute figures vary widely by sector within stage; the directional change is the readable signal.

The compression at Series B and later is not uniform — vertical-AI companies and the strongest coding-agent rounds resisted it, while broad-horizontal infrastructure plays bore the brunt. The interpretation we favour: late-stage capital is differentiating on unit economics and revenue durability rather than category membership. The era of category-wide multiple expansion ended in H1.

For operators, the practical read is straightforward. Vendors raising at the upper end of stage medians in H1 are signalling commercial traction that survived a tighter price-discovery process. Vendors raising at the lower end may still be excellent choices, but the capital signal alone no longer carries the confidence it did twelve months earlier.

The same four sectors that anchored the headline rounds above account for the bulk of capital allocated in H1. They behave differently as procurement categories, however, and the table below summarises the practical differences operators should understand when sequencing vendor selection.

One sector-boundary note matters for interpretation. Coding agents and search-and-retrieval overlap meaningfully — a company offering repository-aware retrieval to a coding workflow legitimately sits in both. We classified by the company's primary revenue motion rather than its technical surface, which means our breakdown will differ marginally from a strict technical taxonomy. Operators should view this taxonomy as a procurement orientation, not a technical schema.

The geographic distribution of H1 2026 AI capital tells a story of measurable but partial dispersion. The Bay Area still hosts the largest single concentration of rounds by count, but its share contracted versus the 2025 baseline. European and APAC ecosystems absorbed the difference.

The procurement read on geographic dispersion is practical. If your organisation operates across multiple regions, the H1 pattern argues for evaluating regional vendors in their home markets rather than defaulting to Bay-Area-headquartered providers. That argument has been theoretically true for years; it became practically true in H1 2026 as regional vendor quality crossed the threshold required for serious procurement consideration.

Underneath the sector and stage breakdowns, four governing trend lines explain the bulk of H1 allocation behaviour. Each is worth naming explicitly because each carries implications into H2 and beyond.

The four trends interact in a way that matters. Trend 1 (agentic infrastructure rotation) and Trend 3 (vertical AI rise) together describe a stack that is consolidating at the infrastructure layer and specialising at the application layer. Trend 2 (growth-stage discipline) is the corrective mechanism enforcing it. Trend 4 (regional rebalance) is the geography in which the first three play out. Taken together, the four lines describe a maturing category, not a cooling one.

For operators, the planning move is to position procurement around Trends 1 and 3 — favour agentic infrastructure for horizontal capability and vertical AI for domain-specific workflows — while reading Trend 2 as confirmation that the vendors raising at the upper end of stage medians are the ones that survived a tighter price-discovery process. Trend 4 simply argues for regional vendor evaluation as a default discipline rather than an exception. Our AI digital transformation engagements begin from exactly this kind of capital-flow reading, calibrated to a client's procurement and competitive context.

Forecasting venture allocation six months out is exercise in scenario discipline, not prediction. The projection below sets out a base case anchored on H1 trend extension, with a downside case tied to macro deal-finance availability and an upside case tied to accelerated exit activity unlocking secondary capital.

Base case. The four governing trends carry into H2 with modest amplification — agentic infrastructure consolidation continues, vertical AI gains additional share, growth-stage discipline holds or modestly tightens, and regional rebalancing continues at the H1 pace. Disclosed round count grows at roughly the H1 trajectory, with capital concentration further favouring distribution-led and workflow-led plays over capability-led ones.

Downside case. Macro deal-finance availability tightens — credit markets compress, public-market exit windows narrow, or a high- profile late-stage down-round triggers broader recalibration. Round count holds but average size compresses, growth-stage multiples extend their H1 compression by another 10-20%, and agentic infrastructure consolidation accelerates as smaller players struggle to raise.

Upside case. Accelerated exit activity — strategic acquisitions of agentic-tooling companies, secondary transactions at growth stage, and a meaningful AI-native IPO — unlocks recycled capital. Round count expands beyond the H1 trajectory, vertical AI attracts disproportionate share of the recycled capital, and regional rebalancing accelerates as the unlocked capital seeks geographic diversification.

We will publish the next investment retrospective in mid-November 2026 with H2 data and a recalibration against this projection. The discipline matters more than the specific call — we will disclose hit and miss openly so the framework can be evaluated. For operators positioning procurement now, the right read is the base case, with the downside case as a stress test for vendor selection and the upside case as an opportunity signal if exit activity materially accelerates.