HKU Jockey Club Enterprise Sustainability Global Research Institute
World-Class Hub for Sustainability
Sean Shun Cao | G. Andrew Karolyi | William W. Xiong | Hui Xu
Biodiversity entrepreneurship
June 12, 2026
Can large language models (LLMs) help with entrepreneurship research? In the following paper, the authors use an open-source LLaMA model and fine-tune its parameters to enable the model to identify climate-solution (biodiversity) startups. We then provide a systematic exploration of these innovative startups—who they are, how they are financed, the challenges and opportunities they face, and how they promote themselves on social media. See more below.
Key Takeaways
- Research Questions: How do biodiversity-focused startups raise capital? Which investors finance them? And can public communication help reduce the information frictions that limit access to conventional venture funding?
- Data and Method: The authors fine-tune Llama-3.1-8B on a manually verified Crunchbase training sample and identify 630 biodiversity-linked startups in PitchBook from 2010–2023. They match 605 of these firms to comparable generic startups by industry, region, founding year, and employment, then examine financing deals, investor composition, and Twitter activity.
- Findings
- Biodiversity startups receive a comparable number of financing deals but raise much less capital. Average total fundraising is $9.8 million, compared with $22.5 million for matched generic startups, and the average deal size is $3.2 million, compared with $6.6 million.
- Financing relies on a hybrid investor base. Biodiversity startups have 2.20 times greater odds of receiving impact-investor funding and 1.26 times greater odds of receiving public capital than matched generic startups, after controlling for company and deal characteristics.
- Twitter activity is associated with broader investor access. Having an active Twitter account is linked to 35.8% higher total funds raised, and greater posting intensity is associated with more financing deals for biodiversity ventures.
- Implication: Biodiversity finance requires institutional support beyond standard venture capital. Impact investors, public-sector capital, syndication, and credible communication channels help make ecological value more legible to private investors.
Source Publication:
Cao, S. S., Karolyi, G. A., Xiong, W. W., & Xu, H. (2026). “Biodiversity Entrepreneurship.” Review of Finance, 30(1), 43–86.
Background
Biodiversity finance confronts a basic problem in private markets. Ecosystems support food production, pharmaceuticals, water regulation, soil quality, natural resources, and supply-chain resilience, but much of this value is difficult to assign to a single firm’s cash flows. A startup that restores coral reefs, protects forests, monitors wildlife, or builds biodiversity data infrastructure may create broad economic and ecological benefits long before those benefits appear as scalable revenues.
This disconnect between creating ecological value and generating near-term scalable profits makes biodiversity different from many parts of climate finance. A renewable-energy project can often generate revenue by selling power. A biodiversity venture frequently creates value by preserving or restoring natural capital whose benefits are distributed across firms, communities, and future periods. The resulting financing problem is not simply that biodiversity matters for society; it is that investors must assess projects whose benefits are hard to verify, hard to contract on, and often hard to monetize through conventional venture-capital models.
This paper examines this problem through biodiversity startups, the private ventures whose core activities are linked to biodiversity conservation, restoration, monitoring, or nature-based services. The authors ask how such firms can be identified, how their financing compares with otherwise similar startups, which investors support them, and whether social media helps them reach investors beyond the specialized impact-finance community.
Data and Methodology
The authors construct a new dataset because biodiversity entrepreneurship is difficult to observe in existing private-market data. They begin with a manually verified sample of 173 biodiversity startups from Crunchbase, using biodiversity-related keywords and website checks. They then fine-tune Llama-3.1-8B to distinguish biodiversity ventures from broader sustainability startups and generic startups. The model achieves about 96% accuracy and identifies 630 biodiversity-linked startups in PitchBook, where the authors can observe company profiles, financing histories, and investor characteristics.
To benchmark financing outcomes, the authors match 605 biodiversity startups to 605 generic startups operating in the same industry and region, with similar founding years and employment levels. This matching design allows the paper to compare biodiversity and non-biodiversity ventures with broadly similar age, size, sector, and geography. The deal-level analysis covers 2,483 financing deals and distinguishes between impact investors and public or philanthropic capital. The paper also studies Twitter activity, both through cross-sectional variation in whether startups maintain active accounts and through posting intensity among firms with accounts. A further test uses Twitter’s 2017 character-limit expansion as a shock that increased the platform’s ability to convey richer information.
Findings and Discussion
Biodiversity entrepreneurship is a heterogeneous sector rather than a single business category. The authors classify ventures into five ecological categories: marine, forest, agricultural, animal, and multifunctional biodiversity. Multifunctional ventures account for the largest share of firms, 26.5%, and the largest share of total funding, 43%. Forest-related startups represent 25.7% of the sample and receive 33.7% of total funding. Marine ventures have a similar firm count but receive only 10.6% of total funding. The paper also distinguishes direct conservation ventures from indirect approaches such as biodiversity data platforms, education, consulting, and public-awareness tools. Indirect ventures make up 53.8% of the sample and receive 55.2% of total financing.
The financing gap appears in the size of funding rather than the number of transactions. Biodiversity startups secure 2.39 financing deals on average, close to the 2.17 deals received by matched generic startups. The amount raised is much lower. Average total fundraising is $9.8 million for biodiversity startups, compared with $22.5 million for matched generic startups. Average deal size is $3.2 million, roughly half the $6.6 million raised by the matched group. The gap remains after matching on industry, region, founding year, and employment, indicating biodiversity ventures face financing frictions beyond standard startup characteristics.
Investor composition helps explain how these ventures obtain funding despite smaller deal sizes. Biodiversity startups draw more heavily on values-oriented capital. Relative to matched generic startups, they have 2.20 times greater odds of receiving impact-investor funding and 1.26 times greater odds of receiving public capital. In the broader PitchBook universe, about 5% of startup deals involve impact investors, compared with about 15% of biodiversity startup deals. Public and philanthropic investors also appear more frequently in biodiversity financing.
This investor base combines conventional venture capital with accelerators, incubators, government agencies, sovereign wealth funds, public institutions, and niche impact funds. Biodiversity-investing funds are more likely to prefer minority stakes and syndication, suggesting an investment model built around risk sharing and coalition formation. Impact investors are associated with larger deal sizes for biodiversity startups, which helps narrow the financing gap. Public capital plays a different role. It appears more often in early-stage support, grants, accelerators, and de-risking arrangements, where deal sizes tend to be smaller but may help firms reach later rounds of private financing.
Social media provides another channel for reducing financing frictions. Twitter presence is associated with higher fundraising and more financing deals for biodiversity startups. Having an active account is linked to 35.8% higher total funds raised. Among firms with accounts, a 1% increase in the number of tweets is associated with a 0.157% increase in the total number of financing deals. The relationship is especially pronounced for deal frequency: posting intensity is more strongly associated with additional deals for biodiversity startups than for matched generic startups.
The mechanism is informational. Mission-aligned investors may already understand biodiversity objectives and have experience evaluating nature-focused business models. Profit-oriented investors may face greater uncertainty about scientific credibility, commercial scalability, ecological measurement, and revenue potential. Twitter offers a low-cost channel for communicating progress, explaining technology, signaling credibility, and reaching investors outside the specialized biodiversity-finance community. The paper’s test around Twitter’s 2017 character-limit expansion provides suggestive evidence that richer communication increased deal sizes for biodiversity startups with Twitter accounts, whereas pre-treatment trends appear statistically weak.
Implications
The findings point to a central challenge in nature finance: biodiversity ventures can create economic value that is real but difficult for early-stage investors to measure and price. Standard venture capital may undersupply funding when ecological benefits are diffuse, verification costs are high, and revenue models are still developing. Impact investors and public-sector actors help address this gap by supplying capital that is more patient, more mission aligned, or better suited to early-stage uncertainty.
For policymakers, the evidence supports blended-finance programs that target the investment process rather than broad biodiversity commitments alone. Public and philanthropic capital can fund proof-of-concept projects, measurement systems, pilot programs, accelerators, and early-stage experimentation. These interventions can create the track records and information infrastructure that private investors need before committing larger amounts of capital.
For entrepreneurs, the paper shows scientific quality and ecological relevance are only part of the financing problem. Biodiversity startups also need to make their value proposition intelligible to investors: what the venture measures, how its ecological output can be verified, where revenue comes from, how the model scales, and why the technology or service can survive in private markets. Strategic communication is therefore part of capital formation, especially when investors lack specialized knowledge of biodiversity.
For investors, biodiversity entrepreneurship represents a growing private-market segment with persistent underfunding. The opportunity depends on the ability to evaluate ecological outcomes, commercial models, and institutional support together. The ventures most likely to attract capital are those that can translate nature-positive activities into credible investment signals. In that sense, biodiversity finance is both a capital-allocation problem and an information problem: closing the gap requires investors who can price ecological value and requires institutions that make such value easier to observe.