The $2 Trillion Pharma Race: Where Startups and Venture Capital Fit

In the third quarter of 2026, Eli Lilly generated $24.4 billion in revenue. Two medicines, Mounjaro and Zepbound, contributed $16.1 billion of that total. Both use tirzepatide. One successful drug franchise now accounts for roughly two-thirds of a pharmaceutical giant's quarterly sales.

That is the economic prize biotechnology startups are trying to capture—and the reason established drugmakers keep investing in, licensing from, and acquiring smaller companies. Yet the market is more complicated than a comparison between pharmaceutical sales and software revenue: medicines take years of testing, most candidates fail, and patent protection eventually expires.

The investment story is therefore not simply that AI might invent more drugs. It is who gets paid at each stage of turning biological research into an approved medicine.

A $2 trillion market with a replacement problem

Evaluate's June 2026 World Preview forecasts worldwide prescription-drug sales above $2 trillion by 2032, growing more than 7% annually from 2025. It also estimates that over $500 billion in sales could be exposed to generic competition during 2026–2032. Exposure is not the same as revenue that will certainly disappear: the impact depends on patents, geography, competitors, pricing and the ability to move patients to newer products.

Lilly shows both sides of the equation. Mounjaro generated $10.7 billion in third-quarter revenue and Zepbound $5.4 billion, according to the company's October 29 financial release. But Lilly also reported that volume growth was partly offset by lower realized prices. A large patient market is not an unlimited pricing opportunity.

Drugmakers must continually replace aging franchises with differentiated treatments. Obesity and metabolic disease, oncology, and immunology offer substantial commercial demand, but each presents a different hurdle. In obesity, a challenger needs to improve on increasingly effective existing medicines. In cancer, an attractive early laboratory result must translate into meaningful patient benefit. In immunology, extending one asset across several diseases can expand revenue, but each new use needs clinical evidence.

This creates the space for startups: a small team can pursue a new biological target, delivery technology or molecule that an incumbent's existing portfolio has not prioritized.

Follow the money: venture rounds are only one financing channel

The latest completed quarterly dataset shows how much pharmaceutical investment happens outside conventional startup funding. J.P. Morgan's Q3 2026 biopharma review, using DealForma data, reports activity for the nine months ended September 30:

These figures must not be added together as though they describe one pool of startup investment. Only about 6% of announced licensing deal value represented upfront cash, according to J.P. Morgan. A startup may advertise a multibillion-dollar collaboration while most of the amount remains contingent on successful development, regulatory approvals or commercial sales.

For investors, the relevant questions are: How much cash changes hands now? Who owns the drug rights? Which party pays for future trials? Which milestones are achievable? Who collects royalties if a medicine succeeds?

Those questions determine whether a partnership creates immediate financial strength or mainly offers a route to future financing.

Four types of startup are competing for pharmaceutical value

The boundaries blur. A company can sell software to pharmaceutical customers and simultaneously develop its own pipeline. This creates different returns and risks: tool vendors may monetize sooner but capture less upside per drug; asset owners may capture more value but need far more capital before clinical success becomes clear.

Isomorphic Labs: capital for the end-to-end drug engine

On May 12, 2026, Isomorphic Labs announced a $2.1 billion Series B led by Thrive Capital, with Alphabet, GV, MGX, Temasek, CapitalG and the UK Sovereign AI Fund among participants. The funds are intended to scale its AI drug-design engine and progress its own therapeutic programs toward the clinic.

The financing demonstrates the capital intensity of combining AI research, biological validation and a proprietary drug pipeline. It does not demonstrate that the company's medicines have been approved.

Chai and Boltz: selling the research layer to pharma

Chai Discovery announced a $400 million Series C in July 2026 and disclosed collaborations with pharmaceutical companies including Pfizer, Eli Lilly, Novartis and Bristol Myers Squibb. Boltz launched as a public-benefit corporation in January 2026 with a $28 million seed round and a multiyear Pfizer collaboration.

Their approaches are not identical, but both point to an investable software and research-infrastructure layer. The near-term test is whether scientists actually use these systems repeatedly, and whether customers renew contracts because the tools improve decisions rather than simply produce impressive molecular images.

Enveda, Pathos and BigHat: moving from models to clinical programs

Enveda closed a $311 million Series E in September to advance drug candidates discovered through its natural-chemistry platform. Pathos AI raised $237 million Series E in October, led by Bristol Myers Squibb, at a company-disclosed approximately $3 billion post-money valuation; it is developing a portfolio of oncology assets. BigHat raised $75 million Series C to advance its therapeutic pipeline and experimental protein-design platform.

All three illustrate the distinction between a good technical model and a drug-development company. Once assets enter human trials, clinical endpoints, safety, manufacturing and cash runway increasingly drive value.

For funding details, sources and business-model comparisons, see our AI drug-discovery companies guide. For a chronological view across traditional and AI-enabled biotechs, see the 2026 biotech funding tracker.

The clinical bottleneck may be a larger business than discovery itself

Better computational models can increase the number of drug candidates worth testing. Yet more promising candidates do not automatically mean more approved medicines: studies still need patient recruitment, dosing, manufacturing, safety monitoring, evidence review and regulatory evaluation.

Formation Bio has taken an unusually explicit approach to this constraint. Its October 28 announcement describes a $600 million Drug Development Accelerator Fund intended to finance 10–15 drug programs over three to five years. The fund acquires or licenses clinical-stage assets into separate entities that use a shared development team and technology stack.

That $600 million is fund commitments, not a new Formation Bio equity financing or a $600 million company valuation. The premise is that a shared operating system could reduce the cost of developing multiple assets. Whether it actually improves clinical success and investor returns remains to be demonstrated. Our Formation Bio analysis explains the separate asset companies, disclosed terms and missing economics.

The opportunity reaches beyond AI-first companies. Contract research, clinical data, manufacturing capacity and laboratory automation may all benefit if discovery produces more viable candidates. But a wave of more candidates could also make expensive testing facilities the industry's next binding constraint.

Specialist venture capital is financing that risk

Investors are approaching biotech at multiple stages. Forbion announced €2.3 billion raised across two life-sciences funds in October 2026. Jeito Capital's second fund exceeded €1 billion earlier in the year and focuses on biopharma development. These are fund sizes, not investments in a single startup.

Their appeal to limited partners depends on specialist judgment: deciding which scientific uncertainty deserves capital, which clinical milestone can create value, and when to license or sell an asset instead of funding another large trial.

Strategic pharma investors operate differently. When Lilly or Bristol Myers Squibb invests in a startup, it may gain financial exposure and an early relationship with a potential future partner. That does not mean a subsequent acquisition or licensing deal is guaranteed.

For Europe's founders, the involvement of large cross-border funds matters because clinical-stage biotechs can require follow-on investment far larger than a normal software startup. See VCT's existing coverage of Forbion's two funds and Jeito's European biopharma strategy.

China and the global competition for drug assets

Evaluate reports that China has rapidly increased its importance as a source of drug assets licensed by international pharmaceutical companies. Its 2026 World Preview expects Chinese-origin deals to represent more than two-thirds of global licensing deal value for the year, compared with less than 5% five years earlier. This is an industry forecast, not proof that two-thirds of all completed agreements or upfront payments originate in China.

For a US or European startup, the implication is practical: comparison sets are increasingly global. A drug can be scientifically novel locally and still compete against another program already generating clinical evidence elsewhere. Geography also changes intellectual-property, regulatory and commercialization negotiations.

Can AI actually improve the probability of approval?

There is a critical evidentiary ladder:

1. A model predicts a promising molecule. Useful for prioritizing experiments; not evidence of efficacy.
2. Laboratory studies confirm the molecule behaves as predicted. Stronger scientific evidence; still not a patient outcome.
3. Early clinical trials assess safety and initial efficacy signals. Encouraging results need replication and adequate trial design.
4. Larger trials establish clinically meaningful benefit. Only then is the commercial thesis substantially de-risked.
5. Regulatory approval and distribution follow. Approval is not equivalent to successful launch, reimbursement or durable revenue.

Insilico Medicine announced that the first patient had been dosed in a Phase III study of rentosertib in September 2026. This is meaningful advancement beyond software benchmarks—but the Phase III study still has to establish safety and efficacy; trial initiation is not proof that the drug works.

Iambic Therapeutics priced a $150 million Nasdaq IPO in October 2026, and Generate Biomedicines priced a $400 million Nasdaq IPO in February. Public-market access is another sign that AI-associated drug development has progressed beyond private venture portfolios. It is not a clinical success metric.

The US FDA and European Medicines Agency have also published principles for responsible AI use in drug development. Models must have clearly defined purposes, reliable validation, appropriate oversight and risk management. Regulatory evidence cannot be replaced by a claim of faster discovery.

Where the next investment returns may come from

Three outcomes are plausible, and the eventual mix remains uncertain.

In one, the most valuable AI companies are specialized software and data providers that sell into every large drugmaker. In another, full-stack startups own medicines and capture clinical-asset economics rather than software fees. In a third, clinical execution and physical experimentation become the highest-value bottlenecks as the supply of candidate molecules rises.

The best evidence will be operational: repeat pharmaceutical customers, disclosed upfront contract economics, reproducible laboratory performance, well-designed clinical trials, and eventual approvals.

The industry's $2 trillion sales outlook explains why the opportunity is large. The ability to turn discovery into validated, differentiated medicine will determine who captures it.

Methodology: Market forecasts are attributed to Evaluate; nine-month financing totals are from J.P. Morgan/DealForma through September 30, 2026; company funding figures come from issuer releases. Announced funding, fund commitments, licensing maxima, IPO proceeds and drug revenue are distinct categories. This analysis is a dated editorial assessment, not a projection of guaranteed investor returns.