On May 12, 2026, Bristol Myers Squibb committed up to $15.2 billion to Hengrui Pharma across 13 programs, none of them yet in human trials. Guaranteed near-term cash was $950 million, an upfront-to-total ratio of 4.1%. Six months earlier Novartis bought Avidity Biosciences, a US company with three late-stage programs, for about $12 billion, effectively all of it upfront. Two deals of similar headline size, seven months apart, with biotech valuation logic running in opposite directions. The gap is not a quirk of two transactions; it is how US boards price China-origin assets.
The asymmetry is deliberate, and it is wrong in a specific way. Boards reach for one number, the China discount, when they are pricing three unrelated risks at once. The first is governance: state-owned-entity exposure, variable interest entity (VIE) structures, listing venue, and the overhang from BIOSECURE, the COINS Act, and the Biotech Investment National Security Act (BINSA). The second is data: China's human genetic resources (HGR) regime and the FDA's posture on China-only clinical evidence. The third is manufacturing: contract-manufacturing cost, bioreactor capacity, and the concentration of antibody-drug conjugate (ADC) work. A blended haircut treats all three as one risk. One is partly theatrical, one is verifiable and cheap to close, and the third, read correctly, is a premium.
Three risks, one blended haircut
The scale is not in dispute. Jefferies analyst Cui Cui put the structural gap in plain figures in July 2025: upfront payments on China licensing deals run 60-70% below comparable Western deals, and total deal sizes 40-50% below. Locust Walk's 2025 review showed China-based sellers taking 47% of global licensing value, up 22 points from 2024, on an average deal of $2.1 billion against $800 million for US transactions. Cross-border out-licensing to US and European pharma reached $136 billion in 2025, up from under $5 billion in 2020.
The discount is real, but it is closing. Mark Lansdell of Evaluate tracked average China upfronts rising 230%, from $52 million in 2022 to $172 million in early 2026, with the 2026 figure 22% above the 2025 average. "It's not a bargain basement anymore," he said. "China-based companies are now bringing their upfront payments into line with what you would expect to see for deals with companies headquartered elsewhere." The market is repricing faster than most boards are.
The governance discount: regulated and theatrical
Part of the governance discount is grounded. BIOSECURE became law on December 18, 2025 as Section 851 of the FY2026 defense authorization act, barring federal agencies from contracting with designated biotechnology companies of concern. BINSA, introduced on June 2, 2026 by Representatives John Moolenaar and Debbie Dingell, would fold biotech into the COINS Act's outbound-investment screen, reaching licensing deals and joint ventures. Council of Institutional Investors research from August 2025 documents VIE-linked discounts of up to 30% on US-listed Chinese firms. That portion is real, escalating, and properly priced into any structure that leans on a named entity or on technology transfer at scale.
The rest is theatrical. A September 10, 2025 New York Times report on a draft executive order restricting US access to Chinese assets cost Zai Lab 9% and BeOne 10.5% in a single session, on an order the White House later said it was not pursuing. David Risinger of Leerink Partners named the flaw: "Restricting access to cost-effective and swift development opportunities in China could hinder the progress of US biopharma drug development." When the BMS-Hengrui deal landed, Moolenaar wrote to the Treasury calling it dangerous, and BioWorld ran the headline "BMS-Hengrui deal strikes panic in Washington." The noise is loud, but on the current statute it does not block transactions built around named-entity exposure. The discipline is to price the regulated risk once and refuse to pay for the theatrical risk twice.
The data discount: verifiable and cheap to close
China's HGR Rules, updated in May 2024 to move authority from the science ministry to the National Health Commission, govern the export of genetic materials and genomic data. The 2023 implementing rules limited HGR to hereditary data, gene and genome data, and excluded clinical, imaging, protein, and metabolic data. The regime restricts a narrow class of post-deal data flows. It does not invalidate the clinical record a board is buying.
The FDA is the harder constraint, and it runs through practice, not statute. The agency's rejection of Lilly and Innovent's ORIENT-11 data established that China-only Phase 3 evidence is generally insufficient at approval. House appropriators attached report language to the FY2027 FDA bill on April 29, 2026, directing the agency to refuse covered clinical data from China and three other nations in IND filings; J.P. Morgan noted that report language carries no force of law as drafted. The real constraint is a bridging-study cost, and recent deals price it as a line item. Takeda's October 2025 Innovent deal put Takeda at 60% of US development cost and 60% of US profit, with US manufacturing committed. AstraZeneca structured its eight CSPC obesity and diabetes programs so that it runs all development from the IND forward; the lead asset entered Phase 1 under that structure, sidestepping the China-only data question. A flat "China data" haircut prices a cost that two of the year's largest deals chose to itemize instead.
The manufacturing premium boards refuse to credit
The third category is the one boards read backward. WuXi Biologics supported 945 integrated client projects as of December 31, 2025, including 74 in Phase 3 and 25 in commercial manufacturing, with a 100% pass rate across 22 FDA and EMA inspections and a DNA-to-IND clock of six months against an industry norm of 10-12. It added 209 projects in 2025, weighted toward US clients. China carried more than 50% idle bioreactor capacity in 2024 while US and European utilization sat below 30%, with biologics manufacturing at a third of US cost. WuXi XDC holds 100,000 of the 580,000 liters of global ADC bioconjugation capacity, and China accounts for about 90% of global ADC out-licensing. This is a strategic-supply position, not a line to discount.
Boards pay for it twice when they miss that. Pfizer moved 3SBio's lead asset to manufacturing sites in North Carolina and Kansas; Takeda committed to US manufacturing for the Innovent programs. Both moves are defensible on national-security grounds, and both surrender the per-batch arbitrage that made the asset cheap. The haircut applied for "Chinese manufacturing" is often paid once in the headline discount and again in the cost of moving the work out. The cross-border mechanics that produce these structures sit inside the 2026 APAC outlicensing playbook.
Comparable transactions, read by upfront
Five China-origin deals across 2024-Q2 2026 form a clean comparator set. AbbVie and RemeGen, January 2026: $650 million upfront, $5.6 billion headline. Pfizer and 3SBio, May 2025: $1.25 billion upfront plus $100 million in equity, $6.15 billion headline. Hengrui and GSK, July 2025: $500 million upfront, $12.5 billion across 12 programs. Hengrui and BMS, May 2026: $600 million upfront, $15.2 billion across 13 pre-human programs. CSPC and AstraZeneca, January 2026: $1.2 billion upfront, $18.5 billion across eight programs. Novartis and Avidity, the US comparator, closed near $12 billion for three late-stage programs at effectively 100% upfront.
Read the set by upfront, not headline. The upfront-to-total ratio runs 4-22%; the headline figures are biobucks, long-dated and conditional. PharmaVoice found that only 40 of 92 disclosed China deals in 2025 revealed an upfront at all. Upfront per program, adjusted for stage, is the cleanest input a board has, and it prices Pfizer's $1.25 billion bet on one Phase 2-3 asset much closer to a Western norm than to a discount.
A three-part biotech valuation framework
Each risk gets its own band, scored on its own. The governance band runs from 0% (ex-China rights, no named-entity exposure, no Chinese manufacturing in the value chain) to 15% (a BIOSECURE-adjacent counterparty, a VIE-listed seller, technology transfer at scale). Political-cycle noise is priced once, in the structure, not layered on a second time.
The data band runs from 0% (a US-led IND from preclinical) to 10% (China-only Phase 2 with a disclosed, costed bridging plan). The band closes the moment the bridging cost becomes a line item rather than a guess. Takeda's 60/40 split is one way to close it; AstraZeneca's preclinical-stage IND is another.
The manufacturing position is a premium, not a band. Where the structure captures Chinese capacity at signing and keeps the option to move on a regulatory or political trigger, it earns 5-15% over a US-origin comparable at the same stage. Where it forces immediate transfer, the premium reverts to zero and the acquirer carries the cost. Revised value is the governance and data bands applied to the upfront, with the manufacturing premium credited in the open. One number, three named parts, each defensible to an audit committee.
Seven moves before the finance committee votes
Seven moves before the finance committee votes on a China-origin asset.
One. Make the BD head report the upfront-to-total ratio, not the headline. Headline is biobucks; upfront is the priced commitment.
Two. Require the governance band itemized, with named-entity exposure, VIE structure, and BIOSECURE status as separate lines. Refuse a blended figure.
Three. Require a written bridging-data plan, with cost, timeline, and US-patient thresholds, before any data-band discount is applied. No plan, no discount.
Four. Require a stated manufacturing position: capture, transfer, or hybrid. Credit the premium where the structure holds Chinese capacity; charge the cost where it does not.
Five. Apply Albert Bourla's 3SBio standard above $250 million upfront. On that deal Bourla said: "We didn't do due diligence in a data room. We sent people on-site." On-site clinical diligence is the cheapest way to close the data band.
Six. Track the convergence. Evaluate's 230% rise in average upfronts since 2022 is the clearest sign the discount window is closing; a 2027 deal struck at 2025 multiples will look expensive within 18 months.
Seven. Book a board review at the deal's first anniversary against the original three bands, with the BD head required to reconcile any drift.
The bear case is real. It is also incomplete.
The bear case rests on statute, not sentiment. BIOSECURE is law, BINSA is moving, and the FY2027 report language signals an appropriations mood that is hardening. Moolenaar put it bluntly after the BMS-Hengrui deal: "When Pfizer and Bristol Myers partner with Chinese biotech companies, their deals imperil the future of American drug innovation." That is genuine risk, and a genuine argument for holding option value in an under-priced asset as a hedge against a deal coming undone.
It is also incomplete. Bourla's on-site diligence on 3SBio, Takeda's 60/40 cost share, and AstraZeneca's preclinical-stage IND all show the China discount giving way to priced, itemized risk. BMS research chief Robert Plenge framed the logic with no discount in sight: "Tapping capabilities across geographies could cut the time to early clinical insights and support informed decisions." That is a sourcing decision, not a discount. For a US board the question is not whether to apply a China discount. It is whether "discount" is the right name for the work the board has declined to do. The next 18 months are the cleanest stretch left to do it deliberately.