Applying portfolio theory to bio startups
This business model involves a “hub” / parent company that handles non-scientific work that owns and manages a portfolio of single asset companies. It centralized overhead operations and treats assets almost as a hedge fund would.
BridgeBio, Roivant Sciences, Biohaven Pharmaceuticals, ElevateBio, Gossamer Bio, Nimbus Therapeutics, and PureTech. Flagship Pioneering, Third Rock, and Curie Bio are also somewhat relevant.
The premise stems from portfolio theory which states that independent, uncorrelated return streams lower risk without lowering expected returns. Diversification is the one free lunch in finance.
The portfolio’s risk being lower means that the parent company doesn’t have to perpetually raise money via highly dilutive equity raises but can raise relatively cheap debt. Indeed, the first people to try the idea in biomedical funding (BridgeBio) successfully raised 60% (!) of their funding through some form of debt, namely $75M of senior secured debt at 9% interest and nearly $1.3B of convertible notes.
In search of uncorrelated assets, hub-and-spoke companies typically either diversify across disease areas or across rare genetic diseases each with its unique mechanism of action. Treatments can also be pursued through different modalities.
Ideally, the model also allows for:
You can construct an equity-only transaction, an option to acquire or option to license, a traditional licensing deal, or intricate multi-asset development collaborations. That is not to say any of those deal types cannot be achieved with a traditional structure, but the subsidiary structure allows existing IP, arising IP, non-compete clauses, and the preservation of platform IP vs. asset IP to be easily identified and pre-packaged for a variety of transactions in a subsidiary structure. For example, in an M&A type transaction, everything relevant to the program can be lifted out, including CRO and CMO relationships, supply chain contracts, academic relationships, IP licenses, financial statements, governance and quality processes, etc., without disrupting staff, platform, or the underlying infrastructure.
Implementing the strategy effectively necessitates truly excellent operational efficiency like BridgeBio:
The two broad categories of asset-centric platform financial structures:
LLC “holding company” model. We’ve embraced this approach with Nimbus Discovery. Its an LLC “parent” company with individual C-corporations (Inc’s) as wholly own subsidiaries. The platform and team are housed in a ‘management’ subsidiary, and as each program is generated it is moved out and housed in its own entity. When these programs mature to partnering stage, our aim is to have Pharma acquire them cleanly for the IP and data package in an earnout approach (e.g., upfront, milestones, downstreams). The value from the sale of the equity flows through the LLC as capital gain back to shareholders. And we’ll keep working on new programs over time. Standalone LLC model. As I understand it, Adimab, Ablexis, and Resolve Therapeutics (and one of our AVDC companies, not yet announced) have been structured in this manner. In this model, licensing deals or technology access payments for specific assets can be channeled through to shareholders directly as LLCs are “pass-through” entities. Most start this way at their inception, but Adimab actually converted itself after its Series D from a C-corp into an LLC (not a small feat, nor a small legal bill; Errik Anderson deserves a ton of credit). Typically, institutional VCs will need to set up special entities that sit between a venture fund and the portfolio company to protect some of their LPs from tax issues. But when the favorable economics of the LLC model are shown to LPs, the choice to go with an LLC becomes quite rational.
The business model on a go-forward basis
Pioneering efforts aimed at thesis like BridgeBio, Nimbus, and Roivant rely on a supply of distressed or left-for-dead assets. But, now that the thesis has been proven out, there’s a good number of players crowding the space, bidding for a fixed pool of clinical assets. So the underpricing of clinical assets has likely shrunk materially.
Some early-stage startups over the last ten years have attempted to turn in-licensing asset selection into a computational problem. While they’ve fallen flat so far, we’re interested in this approach.
Maybe the hub-and-spoke business model becomes even more practical as pre-clinical discovery and novel modalities becomes ever-more commoditized and industrialized. In silico approaches, more predictive disease models, etc. are all on their own exponentials while AI should help with faster clinical trial enrollment and we’re hyper-focused on short cuts to the clinical trial process itself.
In that future, 1) the risk of early stage assets can be assessed more precisely, 2) the capital intensity of translation is lower per attempt, and 3) in-house de novo development may not have to be the company’s core competency.
For companies building upon this model, it’s worth noting that licenses from academic institutions appear to be systematically cheaper than those from fellow commercial firms.
For that future, one should design an AI system that estimates value of drug candidates globally and cheapest possible path to human data.
Luba Greenwood (@luba_greenwood) on X
The Entrepreneur’s Guide to a Biotech Startup Guide To Biotech Startup
Company Histories
Hub-and-Spoke Biotechs: BridgeBioHow to Build This Generation's Big Pharma CompanyWhat does Ramaswamy's Roivant do?Hub-and-Spoke Biotechs: Nimbus TherapeuticsRoyalty Pharma
Curious cases of financial engineering in biotech
