---
title: "Formation Bio's $600M Drug Fund: AI, NewCos and LP Economics"
description: "Formation Bio's $600M fund has reportedly licensed two undisclosed drugs. How its NewCos, AI tools, clinical deals and LP economics work—and what remains unproven."
date: 2026-11-03T10:31:11.000Z
source: https://venturecapitaltracker.com/formation-bio-600m-drug-development-accelerator-fund-2026
---

# Formation Bio's $600M Drug Fund: AI, NewCos and LP Economics

> Formation Bio's $600M fund has reportedly licensed two undisclosed drugs. How its NewCos, AI tools, clinical deals and LP economics work—and what remains unproven.

Formation Bio announced **$600 million in capital commitments** for the **Formation Bio Drug Development Accelerator Fund** on **October 28, 2026**. The New York AI-native pharmaceutical developer plans to acquire or license **10–15 Phase 1 and Phase 2 drug programs over the next three to five years**, putting each program in a dedicated company supported by Formation's clinical and technology platform. The crucial distinction: **this is an investment fund for individual medicines, not a new $600 million equity round or valuation for Formation Bio itself.** [The original manager announcement](https://www.formation.bio/press/formation-bio-announces-usd600-million-debut-fund-to-expand-drug-development-capacity) specifies commitments, not a new Series E.

There is also more progress than the press release reveals. In an [October 28 interview with Forbes](https://tollbit.forbes.com/sites/innovationrx/2026/10/28/why-an-ai-biotech-startup-launched-a-600-million-fund-to-buy-stalled-drugs/), co-founder and CEO **Benjamine Liu said the new fund had already licensed two drug candidates** but declined to name them. That is independently reported interview information, not a list of two identifiable assets or proof of clinical success. The fund gives Formation a new route to finance clinical programs; the harder question is whether it can select the right drugs, move them into trials faster and earn enough from successful programs to compensate for failures.

## $600M fund: essential facts and the missing terms

| Question | Verified position as of November 3, 2026 |
| --- | --- |
| Announcement | October 28, 2026; company-issued announcement |
| Fund | Formation Bio Drug Development Accelerator Fund, debut vehicle |
| Capital | **$600M of fund commitments**; not a startup round, valuation or confirmed amount deployed |
| GP | Formation Bio through a wholly owned general-partner subsidiary |
| LPs | Research foundations, institutional endowments, hospitals and philanthropic investors; no individual LPs or allocations publicly named |
| GP co-investment | Formation says it commits its own money, but does not give the amount |
| Strategy | Acquire or in-license Phase 1 and Phase 2 therapeutic assets; operate asset-specific NewCos |
| Deployment target | 10–15 programs in 3–5 years; clinical milestones often targeted 2–4 years after acquisition |
| Early deployment | CEO told Forbes **two programs already licensed**; their identities and economics are undisclosed |
| Still undisclosed | Fund minimum subscription, management fee, profit share/carry percentage, asset budgets, liquidity period and whether this was a final close |

These details come principally from [Formation's release](https://www.formation.bio/press/formation-bio-announces-usd600-million-debut-fund-to-expand-drug-development-capacity), [Liu's fund explanation](https://www.formation.bio/blog/introducing-the-usd600m-formation-bio-drug-development-accelerator-fund) and [Forbes' direct interview](https://tollbit.forbes.com/sites/innovationrx/2026/10/28/why-an-ai-biotech-startup-launched-a-600-million-fund-to-buy-stalled-drugs/). Some third-party databases categorize the announcement as a **final close**, but the manager's published release does not explicitly say that; VCT therefore records the **commitments announcement**, not an independently verified final close. [Example of the differing tracker classification](https://signals.getkrest.app/fund/formation-bio-drug-development-accelerator-fund).

A back-of-envelope $600 million divided by 10–15 programs suggests **$40 million–$60 million per drug**. That is only a gross arithmetic average, **not a disclosed check size**. Capital may be reserved for follow-on development, manufacturing, trials or operating needs, and Formation says allocations will depend on each asset's circumstances.

## Who is buying what—and why this is unlike an ordinary VC fund

The fund's customers, counterparties and beneficiaries play different roles. A pharmaceutical company or platform biotech may own a technically promising molecule it cannot prioritize; it can license or sell rights to an asset-specific company. Formation evaluates the molecule and supplies development staff, software and operating systems. The fund's institutional LPs supply risk capital and hope to share in the value of later licensing, sale or commercialization. Patients and clinicians are the eventual beneficiaries only if trials demonstrate a treatment works; there is no retail consumer product here.

Formation describes a **hub-and-spoke** structure. Each drug is held in its own NewCo, with a focused general manager and team, while scientific diligence, clinical operations, regulatory, safety, manufacturing coordination and AI systems are provided by a central organization. The original company can negotiate upfront cash, later milestone payments, royalties, or an equity interest in the NewCo. None is guaranteed on every deal. [Formation's detailed NewCo explanation](https://www.formation.bio/blog/formation-bio-newco-model) describes why the separate entity can remain independently financeable and transactable.

This arrangement is designed to avoid the cost and delay of setting up a fresh full-service biotech around every molecule. But **separating ownership does not separate the clinical science**: a failed efficacy trial can still impair the value of the entire asset. Nor does the existence of a fund by itself establish that the central platform is less costly than alternatives.

### An illustrative founder-to-patient walkthrough

The sequence below combines the company's documented operating model with an **illustrative** transaction. It is not a claim that any of the fund's unnamed new licenses has completed every step.

1. **The originator decides what to out-license.** A biotech has early human or IND-ready evidence for a candidate but prioritizes its main pipeline. Formation investigates biological mechanism, existing patient data, differentiation, patents, manufacturing and the commercial case. A deprioritized program is not necessarily a bad one—but the reason it was deprioritized matters.
2. **Researchers test the investment thesis.** Scientific and business-development staff examine target biology, disease indications, safety, dose, competitors and trial feasibility. Human experts decide whether the uncertainties justify capital at risk.
3. **The parties sign the rights deal.** The originator may take upfront cash and retain milestones, royalties or negotiated equity. Geography and field-of-use rights can remain with the original owner. The amount paid up front is not equivalent to the headline potential milestone value.
4. **A dedicated NewCo and operating pod begin work.** A general manager coordinates a focused asset team, with central teams providing trial design, medical and regulatory support, data infrastructure and operational tooling. The fund and Formation's balance sheet may both participate, subject to the deal's actual terms.
5. **A clinical trial produces evidence.** Investigators recruit participants and evaluate prespecified safety and efficacy outcomes. AI can help plan and administer studies, but cannot replace controlled clinical evidence or regulator review.
6. **A value-inflection decision follows.** Depending on results and financing needs, the company may advance a later trial, partner with pharma, sell or license rights—or end the program. A proposed exit is not an achieved return.

The **buyer** on the original rights transaction, the **sponsor** listed in clinical-trial records and the **future commercial partner** may all be different entities. That distinction is particularly important when tracking ownership, data readouts and who actually receives proceeds.

## A closer product tour: Atlas, Delphi, Forge, Apollo, ARK and Phyla

Formation is not selling a general-purpose chatbot to drugmakers. Its [published technology architecture](https://www.formation.bio/technology) describes tools used internally across a drug program's life cycle:

| Platform | What an operator actually does with it | Critical limitation |
| --- | --- | --- |
| **Atlas — asset hunting** | Searches and screens trial-stage programs; maps mechanisms, existing trials, regulatory signals and competitive treatments | Finding an overlooked molecule does not demonstrate a viable drug |
| **Delphi — predictive intelligence** | Scores probability-of-technical-success components, alternative indications and commercial or biological assumptions | A forecast is not a proven clinical success probability |
| **Forge — trial design** | Investigates trial populations, endpoints, sites, precedents, enrollment feasibility, schedules and costs | A proposed protocol still requires scientific, ethical and regulatory scrutiny |
| **Apollo — trial execution** | Coordinates study startup, enrollment signals, compliance issues, timelines and operational records | Operational speed is not itself evidence of treatment efficacy |
| **ARK — internal orchestration** | Connects applications and data sources for governed AI agents and reusable staff workflows | Tool access must respect permission, validation and audit controls |
| **Phyla — cited drug diligence** | Resolves biological identifiers and compares evidence across databases, including negative findings and missing sources | Correct citations do not prove a target or molecule will work in patients |

For a concrete first-visit example, a researcher can use ARK to ask whether a particular gene is a plausible target for a disease. Phyla retrieves underlying target, disease and competing-drug evidence, reconciles incompatible identifiers, and returns claims with citations. A drug hunter then tests that research against the actual molecule, trial evidence and commercial landscape. **This is a documented company workflow, not hands-on testing performed by VCT.** [Phyla's technical report](https://www.formation.bio/blog/phyla) explains its data and provenance design.

Formation says its shared data platform covers more than **720,000 clinical trials, 150 million real-world patients represented in datasets, 4.6 million scientific papers and 290,000 transactions**. These are **company-reported coverage figures**, not independently audited data quality or proof that its models outperform clinicians.

### The available AI benchmarks—and what they do not establish

The most concrete published technical assessment is Phyla's **internal test**, reported October 19, 2026. Formation says an evaluation on **seven target–disease pairs** found **92%** of Phyla's verifiable research claims fully supported by cited sources, compared with **80%** for Claude Science and **67%** for GPT-Rosalind. It also reports 24% faster simple fact retrieval and 39% fewer output tokens in a separate task. Importantly, these were **company-selected evaluations of research evidence**, not independent multicenter studies, patient outcomes, reduced trial costs or regulatory approvals. [Read the methods and caveats in the company's technical article](https://www.formation.bio/blog/phyla).

The human-workflow component is more tangible: Formation says **91 employees used Phyla in its first six months** and 63% of users returned in more than one month. That suggests adoption among its staff, not proof of a clinical advantage. Its older **Muse** recruitment tool, introduced with [Sanofi and OpenAI in November 2024](https://www.formation.bio/press/introducing-muse), illustrates a separate use case: creating recruitment intelligence, outreach materials and prescreening questions. The companies announced planned deployments; a launch statement should not be treated as a controlled efficacy result.

## Three real asset cases illustrate how the model earns—or risks—money

Formation's legacy pipeline and transactions show what an asset-focused strategy looks like in practice. **None of the three cases below should be assumed to belong to the new $600M fund.** Forbes reported two fund-backed licenses without naming them; pre-existing Formation programs are separate evidence.

| Example | What actually happened | What it proves—and does not prove |
| --- | --- | --- |
| **Bleecker Bio / BLKR201** | Formation licensed the CNS-penetrant TYK2 inhibitor from Lynk in December 2025. It announced first-participant dosing in a Phase 1 healthy-volunteer study in June 2026, about five months after deal close according to the company. | Demonstrates an identifiable rights-to-trial-start sequence, **not** successful efficacy, FDA approval or independently established acceleration |
| **Libertas Bio / gusacitinib** | Formation subsidiary Libertas licensed the drug to Sanofi in June 2025 under a transaction potentially worth **up to €545M**, including upfront and conditional milestones, plus potential future royalties. | Shows a possible asset-level pharmaceutical partnering route; the ceiling is **not cash received**, guaranteed proceeds or profit |
| **High Line Bio / sprifermin** | Formation acquired rights to the Merck KGaA-originated knee-osteoarthritis candidate in 2022. Historical studies showed cartilage effects, while questions about pain and knee-replacement outcomes remained. | Demonstrates the need to distinguish a measurable biomarker or structural change from a patient-relevant clinical benefit |

For **BLKR201**, the [December licensing announcement](https://www.formation.bio/press/formation-bio-licenses-phase-1-ready-cns-penetrant-tyk2-inhibitor-from-lynk-pharmaceuticals) explains the economics: Lynk received a minority stake in Bleecker Bio, an upfront payment, eligibility for additional milestones in a package reaching **up to $605M**, and potential tiered royalties. The headline is a **maximum conditional deal package**, not the purchase price paid immediately. Formation's [June 9, 2026 trial update](https://www.formation.bio/press/formation-bio-doses-first-participant-in-phase-1-trial-of-blkr201-a-cns-penetrant-tyk2-inhibitor) said a first-in-human study had started dosing and expected an early-2027 readout. A study start can be verified separately from whether the drug ultimately works.

For **gusacitinib**, the [manager-issued Sanofi licensing release](https://www.prnewswire.com/news-releases/libertas-bio-a-formation-bio-subsidiary-license-of-gusacitinib-a-dual-jaksyk-inhibitor-to-sanofi-302487403.html) reports potential payments to Formation's Libertas Bio subsidiary and royalties contingent on future sales. This is the opposite side of the drug-asset trade: having acquired an investigational program earlier, Formation can seek a larger pharmaceutical partner. Yet the published economics do not show how much has actually been received, the development costs incurred or the net return on invested capital.

The **sprifermin** story shows why an AI prediction requires caution. In an [April 2026 Forbes profile](https://tollbit.forbes.com/sites/amyfeldman/2026/04/16/this-sam-altman-backed-18-billion-startup-bets-ai-can-get-drugs-through-clinical-trials-faster-formation-bio/), Formation described modeling osteoarthritis progression with approximately 23,000 patient records and 48,000 MRI scans to investigate cartilage growth and later knee-replacement risk. Such analyses can sharpen a trial hypothesis, but cannot settle whether an investigational medicine improves pain or prevents surgery. The [current company pipeline](https://www.formation.bio/licensing-and-pipeline) lists sprifermin at Phase 2, BLKR201 at Phase 1 and another program, KMR301, at preclinical stage. **That public pipeline is not a disclosed asset register for the 2026 fund.**

## Formation Bio valuation, previous rounds and why $600M cannot be added to them

Formation (formerly TrialSpark) was founded by **Benjamine Liu and Linhao Zhang**. Its [June 26, 2024 corporate Series D announcement](https://www.prnewswire.com/news-releases/formation-bio-secures-372mm-series-d-funding-to-grow-drug-pipeline-and-continue-expanding-ai-driven-drug-development-platform-302182347.html) reported **$372M of equity financing**, led by Andreessen Horowitz, with Sanofi, Sequoia, Thrive Capital and other investors participating. Formation did not state an exact Series D valuation in that announcement. [a16z investor Scott Kupor's investment rationale](https://a16z.com/announcement/investing-in-formation-bio/) centered on a technology-native pharmaceutical company focused on **development and trials rather than molecule discovery alone**, citing Formation's history across more than 300 supported clinical trials.

In April and October 2026, **Forbes reported** roughly **$615M in cumulative company funding and a $1.8B valuation**. Those are Forbes-reported private-company figures, not a newly disclosed price for the fund, not necessarily a new valuation dated October 2026, and not a fresh financing round. Other databases show different historical totals, so VCT should not present an exact audited cumulative total. The fund's **$600M of LP commitments belongs in a separate fund record**, not in the startup's cumulative venture-funding field. [Forbes background](https://tollbit.forbes.com/sites/amyfeldman/2026/04/16/this-sam-altman-backed-18-billion-startup-bets-ai-can-get-drugs-through-clinical-trials-faster-formation-bio/).

## How does this compare with Roivant, BridgeBio and conventional biotech financing?

| Route | Main capital and operating model | Where it differs from Formation's 2026 fund |
| --- | --- | --- |
| **Formation's fund/NewCos** | Dedicated LP capital finances drug-asset rights and development within a shared AI and clinical organization | Explicit fund-backed individual-asset strategy; LP and related-party economics remain private |
| **Roivant / Vants** | Group of drug-development companies with focused programs and shared strategic capabilities | Similar portfolio-company logic; ownership, stage and transaction terms vary |
| **BridgeBio** | Decentralized companies or focused teams for genetic medicines, supported by shared clinical, regulatory and commercialization expertise | A mature therapeutic developer and commercial-stage company; not an identical debut private LP vehicle |
| **Traditional VC-backed biotech spinout** | Investors fund a new equity company that builds or contracts an operating team | Founder may retain independent control, but financing and staffing must be assembled independently |
| **CRO services or pharma in-licensing** | Sponsors buy trial operations, or a large drugmaker licenses a candidate | The CRO route usually does not itself provide at-risk acquisition capital; licensing can transfer more control and upside |

[Roivant's description](https://www.roivant.com/about/company) and [BridgeBio's operating model](https://bridgebio.com/about) show that Formation did **not invent** the hub-and-spoke structure. Its differentiator is the proposed combination of externally committed **asset-level fund capital**, centralized operating pods and an AI-heavy research and development system. A different [a16z–Eli Lilly venture fund](https://a16z.com/announcement-andreessen-horowitz-partners-with-lilly-to-launch-first-of-its-kind-biotech-ecosystem-venture-fund/) announced in 2025 planned up to $500M of investments in **biotech companies**: relevant industry financing context, but not the same thing as Formation's drug-program fund.

## The LP diligence questions a $600M headline cannot answer

**What is the actual risk-return profile?** Phase 1–2 assets have meaningful clinical failure risk. The fund announcement does not disclose target net returns, estimated loss rates, per-asset budgets, follow-on reserves or realized liquidity. Formation itself says its portfolio approach assumes ordinary industry probabilities of failure rather than making every candidate a likely success.

**Who receives what when a medicine is sold?** Original licensors may own equity and receive milestones or royalties; the fund owns an investment interest; Formation may contribute capital, provide services and earn GP profit participation. The private fund terms would need to specify fees, valuations, related-party transactions, conflicts processes and the waterfall allocating proceeds. **No standard 2-and-20 assumption is justified by the public material.**

**Are underdeveloped drugs a bargain or an adverse-selection trap?** Pharma may deprioritize programs for perfectly rational portfolio reasons, but some candidates could have unresolved biological, toxicological or commercial problems. The test is whether diligence finds evidence overlooked by others—not the quantity of assets acquired.

**Does AI create repeatable clinical value?** A defensible claim would compare time from rights acquisition to trial initiation, enrollment performance, actual spend per protocol, quality findings and outcomes across appropriately matched studies. Formation offers promising technical examples, but not yet a public independently controlled cohort showing greater approval rates or verified portfolio-level returns.

**What is actually funded today?** The CEO's report of two newly licensed assets is a useful starting fact. Their identities, rights, stages, allocation amounts and whether either corresponds to the parent company's published pipeline are not disclosed. The official announcement also does not establish a legally verified final-close date.

**Bottom line:** Formation's $600M vehicle is a serious experiment in **financing drug-development capacity separately from startup equity**. Its potential advantage is not that AI guarantees better medicines, but that better sourcing, reusable operations and dedicated asset capital might let it investigate more candidates without rebuilding a company each time. The meaningful scorecard is **clinical evidence and realized asset economics**, not a larger number in a fundraising headline. Explore VCT's [Formation Bio startup profile](/startup/formation-bio) and [Drug Development Accelerator Fund profile](/fund/formation-bio-drug-development-accelerator-fund) to keep the company, fund and event records distinct.

**By:** [Venture Capital Tracker](https://venturecapitaltracker.com/editorial-policy)
**Last updated:** November 3, 2026

**Editorial note:** AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.