🌏 中文版
Funding Details
| Item | Value |
|---|---|
| Company | Nebuly (Italy, Turin / United States, New York) |
| Round | Series A |
| Amount | $21M |
| Lead | Planven |
| Participants | Exor Ventures, Tenacity Capital, TGVG Capital, Fastweb+Vodafone (via Founders Factory), CDP Venture Capital SGR (Digital Transition Fund, backed by the EU's NextGenerationEU program), existing investors and angels |
| Valuation | Not disclosed |
| Total raised | Not fully disclosed (prior round was a 2023 seed, amount undisclosed) |
| Founded | 2022 |
| Employees | ~40, plans to expand to 60 after this round |
What the company does
Nebuly measures what enterprise AI agents are actually worth — not how often they're used, but how much they impact employee productivity and customer revenue, translated into numbers a board can act on.
The product covers two tracks. For internal AI tools, it measures time saved, new capabilities employees gain, how different teams use AI, and whether AI is being applied to high-value work. For customer-facing AI assistants, it measures AI-influenced revenue and identifies churn risk and upsell opportunities. The platform integrates with AI systems enterprises already run, including tools built on Claude, ChatGPT, and Microsoft Copilot, and can run on-premises or in a customer's private cloud so enterprise data never has to leave for an external model API. Co-founder and CEO Francesco Signorato puts the problem bluntly: companies are making hiring and headcount decisions based on assumptions about AI, not evidence.
Customer contracts start at $100,000 a year, targeting large enterprises. The company operates out of Turin and New York, and plans to expand its US and European sales teams with the new funding.
What this round signals
What it means for the agent ecosystem
Nebuly's pitch names a phenomenon directly — Signorato calls it "tokenmaxxing," where companies treat AI usage itself (user counts, prompts, responses) as the goal, without being able to answer what that usage actually bought them. The signal here is that measuring agent value is breaking out from being a feature inside observability tooling into a standalone product category — Nebuly's angle doesn't fully overlap with watchlist peers like LangSmith, Arize AI, or Braintrust: it's betting on "value," not "behavior tracing" or "eval scores."
What the investors are betting on
CDP Venture Capital SGR's participation, channeled through the EU's NextGenerationEU-backed Digital Transition Fund, gives this round a degree of European industrial-policy backing rather than purely market-driven capital. Planven's rationale as lead investor isn't spelled out in much public detail, but the participant list — Fastweb+Vodafone via Founders Factory, Exor Ventures — skews toward Italian and European domestic capital rather than Silicon Valley funds. That's a notably different pattern from Mirendil, LangChain, and Rox, all led almost entirely by US VCs on the same day, reflecting how local capital carries more weight in this more compliance-oriented corner of enterprise AI governance and measurement.
Numbers worth watching
- At $21M, this Series A is one to two orders of magnitude smaller than the same-day Mirendil ($1B), LangChain ($300M), and Rox ($100M) rounds — a reminder that "measuring agent value" remains a much smaller early-stage niche than "building agents" or "governing agents"
- Customer contracts starting at $100,000 a year are on the high end for an early-stage company in this space, indicating Nebuly positioned itself for large enterprise customers from the start rather than chasing volume with a low-cost self-serve plan
- The team's planned expansion from ~40 to 60 people (50% growth) is far more modest than LangChain's same-day plan to nearly double headcount from 435 to 850, underscoring that Nebuly is still in an early, smaller-scale expansion phase rather than a scaling push
Watchlist status
Nebuly is not yet tracked in the watchlist. We'd recommend adding it to section B6 (Agent observability / evaluation): even though its entry point is business-value measurement rather than traditional behavior tracing or model evaluation, its customer base overlaps heavily with B6 peers like Arize AI, Braintrust, and LangSmith — worth watching whether it expands into behavior tracing and becomes a direct competitor to those existing B6 companies.
Today's takeaway
The most interesting part of this round isn't the amount — it's the appearance of EU NextGenerationEU funding among the participants. That signals that "measuring what enterprise AI is actually worth" has become a question of industrial policy in Europe, not just a business problem, standing in clear contrast to the same-day US deals led entirely by market-driven VC capital.