Akeso: The Bispecific That Made Merck Blink

I. Introduction & Episode Roadmap

On the morning of May 30, 2024, a press release went out from a company headquartered in Zhongshan, a mid-sized manufacturing city in Guangdong province that most Western investors could not have found on a map. It ran a few hundred words. It said that a Phase 3 trial had met its primary endpoint.

By the end of the trading day, Merck & Co. — the owner of Keytruda, the best-selling prescription drug in the history of the pharmaceutical industry — had shed billions of dollars in market value, and a small Nasdaq-listed shell of a biotech called Summit Therapeutics had traded up as much as 272% intraday.¹

The company behind the press release was 康方生物 Akeso, Inc. (HKEX: 9926.HK). The drug was ivonescimab. And the claim was the one thing nobody in global oncology had managed in the decade since checkpoint inhibitors rewrote cancer care: in a randomized head-to-head trial, Akeso's molecule had beaten pembrolizumab — Keytruda — on progression-free survival, and not by a rounding error.²

That single readout is the reason this company is worth an episode. But it is also the reason the company is so easy to get wrong.

Here is the shape of the business as it stands today. Akeso finished FY2025 with commercial product sales of RMB 3,033.1 million, up 51.5% year over year — real revenue from two approved, reimbursed drugs sold into the Chinese market, not a story stock with a PowerPoint.³ It banked a $500 million upfront payment in 2023 from Summit Therapeutics for ex-China rights to ivonescimab, in a deal worth up to $5 billion including milestones — one of the largest out-licensing transactions a Chinese biotech had ever signed at the time.⁴ Its founder, Dr. 夏瑜 Michelle Xia, went from an unsalaried startup founder in 2012 to a billionaire on paper.⁵ And the stock has repeatedly moved 30% to 50% in a matter of days on individual data prints, in both directions.

The central question this piece will test is deliberately narrow, because the sloppy version of the Akeso story — "Chinese biotech kills Keytruda" — is the version that has cost investors money in both directions.

The narrow question is this: is ivonescimab a genuine pipeline-in-a-product that resets global standard-of-care in lung cancer, or is it a strong China dataset that Western regulators and Western trials have not yet fully blessed? And critically: does Akeso's own overall survival data — the number that determines whether patients actually live longer, not merely whether their scans look better for longer — support the stronger claim, or does it narrow it?

That last distinction is not academic. It is the fault line the entire investment case runs along, and Akeso's own trials have landed on both sides of it.

The route from here: how four returnee scientists built an antibody platform from nothing in a provincial city; how cadonilimab became the world's first approved dual-checkpoint bispecific and quietly turned into the cash engine funding everything else; what the ivonescimab data actually shows once you separate the headline from the survival curves; the Summit deal and the governance questions attached to Akeso's chosen global partner; the money, including the one profitable year that was not really profitable; the founder-CEO and how she handled a very public miss; the competitive field, which now includes a rival mechanism backed by a bigger checkbook than Akeso's; and finally the bull and bear cases, and the small number of things worth actually watching.

Start where the company started: with no money, no platform, and no reason for anyone to pay attention.

↑ Back to Top

II. Origins: Building an Antibody Platform From Scratch (2012–2020)

In March 2012, four scientists incorporated a company in Zhongshan, Guangdong. Between them they had decades of experience at Western pharmaceutical companies and universities. Between them they also had less than $3 million in venture funding, and for a stretch of the early years the founders went without salaries.⁵

The lead founder was Dr. 夏瑜 Michelle Xia. Her path was, in the specific way that Chinese biotech in the 2010s was built, entirely typical and entirely improbable at the same time. Undergraduate biochemistry at 中山大学 Sun Yat-sen University in Guangzhou. A PhD from Newcastle University in the United Kingdom. Then cancer research in the United States before returning to China — one of the 海归 "sea turtles," the wave of Chinese-born scientists who trained abroad and came home in the 2000s and 2010s to build companies.⁵

She did not do it alone, and the composition of the founding team is more revealing than the usual founder-myth telling allows. Dr. Li Baiyong had spent twelve years at Pfizer working on cancer immunotherapy. Dr. Wang Zhongmin — "Max Wang" — brought protein expression and structural biology, the unglamorous discipline that determines whether a cleverly designed antibody can actually be produced at scale. Dr. Zhang Peng rounded out the group.⁵ Between the four of them they covered discovery biology, immunology, protein engineering, and process development — which is to say, they had assembled in one room the exact set of capabilities required to originate a biologic, rather than a single charismatic scientist plus a hiring plan. For a company that would later stake everything on molecular geometry that nobody else had built, that founding composition was not incidental.

What made her cohort different from the previous generation of Chinese pharma is what they chose to build. China's drug industry had been, for most of its modern history, a generics and API business: make the molecule someone else invented, make it cheaply, sell it domestically. The returnees wanted to originate. That ambition was cheap to state and brutally expensive to execute, because originating a biologic requires three separate hard capabilities — discovery, engineering, and manufacturing — and each of them takes years and capital to stand up.

Akeso's defining early decision was to build all three in-house rather than rent them.

That sounds like a platitude until you look at what the alternative looked like at the time. The path of least resistance for a Chinese biotech in 2013 was to in-license a molecule someone else had discovered, hand manufacturing to a contract organization, and run trials for the China market. It was faster, cheaper, and lower risk. It also structurally capped what you could ever become: a distributor of other people's science.

Akeso instead built its own antibody discovery engine, its own bispecific engineering capability, and its own manufacturing — years before most domestic peers.⁵ The cost was time and cash the company barely had. The payoff was that when the company later wanted to build a tetravalent bispecific antibody with a specific binding geometry, it did not have to ask anyone's permission or wait in anyone's queue. Everything in this story after 2020 rests on that decision.

The first outside validation arrived earlier than most people realize, and from an ironic direction. In 2015, Akeso out-licensed a cancer antibody, quavonlimab, to Merck & Co. — the same company whose flagship drug it would later attack head-on.⁵ In dollar terms the deal was small. In signalling terms it was enormous: a global top-five pharmaceutical company had looked at the science coming out of a three-year-old Guangdong startup and decided it was worth transacting on. Seven years before the Summit deal, someone in New Jersey had already concluded Akeso could do real antibody engineering.

That is the honest way to read the 2015 deal. It is not evidence that Akeso was destined to succeed. Plenty of small licensing deals with big pharma go nowhere, and this one did not make Akeso a household name. It is evidence that the platform was real — a bounded, specific claim, and the only one the deal actually supports.

The company reached public markets on April 24, 2020, listing on the Hong Kong Stock Exchange under stock code 9926 and raising approximately HK$2.44 billion.⁶ The shares rose roughly 50% on debut, taking the company to a valuation near HK$18.5 billion.⁷

The timing was not luck so much as a window that Akeso stepped through along with everyone else. Hong Kong's Chapter 18A regime, introduced in 2018, let pre-revenue biotechs list, and a cohort went public in quick succession: 百济神州 BeiGene, 信达生物 Innovent, 君实生物 Junshi. Capital was abundant, the pandemic had made biotech the most fashionable sector on earth, and Chinese innovator drug companies were being repriced as a category rather than as individual assets.

For investors, that cohort framing matters more than it first appears, and it is worth holding onto for later. Every one of those companies raised money on roughly the same promise — that a Chinese originator could build globally competitive drugs. Six years on, they have delivered very different outcomes, and the differences are instructive about what actually separates a good China biotech from a great one. Akeso's answer to that question was a molecule that had not yet been approved anywhere.

↑ Back to Top

III. Cadonilimab: The World's First Approved Immunotherapy Bispecific

The drug that made Akeso famous is not the drug that pays Akeso's bills.

On June 29, 2022, China's National Medical Products Administration approved cadonilimab — brand name 开坦尼, internal code AK104 — for relapsed or metastatic cervical cancer.⁸ Two firsts landed in that single decision. It was the world's first approved dual immune-checkpoint bispecific antibody, hitting PD-1 and CTLA-4 with one molecule. And it was the first bispecific antibody originated in China ever to reach approval.⁸

To understand why that mattered commercially rather than just scientifically, you need one piece of biology, and it is simpler than the acronyms suggest.

Cancer cells survive partly by switching off the immune cells that would otherwise kill them. Checkpoint inhibitors are drugs that flip those switches back on. PD-1 and CTLA-4 are two different switches, working at different points in the immune response — think of PD-1 as releasing the brakes on immune cells that have already reached the tumor, and CTLA-4 as widening the pipeline of immune cells that get activated in the first place. Blocking both together works better than blocking either alone. Oncologists have known this for years.

The problem is that hitting both switches with two separate drugs also doubles the toxicity. Combination checkpoint therapy is genuinely effective and genuinely rough on patients; a meaningful fraction cannot tolerate it. Akeso's bet with cadonilimab was that engineering both functions into a single molecule — with binding properties tuned so that the drug concentrates its activity in tumor tissue rather than throughout the body — could capture more of the combination's efficacy with less of its collateral damage.

That is the template. Hold onto it, because Akeso ran the identical playbook a second time with ivonescimab, and the second run is what the market now pays for.

What happened next is the part of the Akeso story that gets least attention and probably deserves the most, because it answers a question that separates biotechs that matter from biotechs that don't: can this company actually sell a drug, or can it only discover one?

Cadonilimab generated RMB 546.3 million in its first six months on the market.⁹ By the first half of 2024, revenue had reached roughly RMB 939 million.¹⁰ By the first half of 2025, commercial sales were RMB 1,401.6 million, up 49.2% year over year.¹¹ Cumulatively, roughly 120,000 patients have been treated with the drug.³

Those are not milestone payments or grant income. Those are prescriptions, written by Chinese oncologists, dispensed to Chinese patients, reimbursed by the Chinese state.

That last clause is the one that does the work. In the Chinese pharmaceutical market, the variable that determines whether a drug becomes a real business is inclusion on the 国家医保药品目录 National Reimbursement Drug List — the NRDL. Getting on the list requires accepting steep, negotiated price cuts. Staying off it means most patients simply cannot afford the drug. Every approved Akeso product and indication is now NRDL-listed.³

That is the trade at the heart of Chinese pharma economics, and it is worth being clear-eyed about: Akeso's China volumes are real and growing fast, but they are being earned at prices set across a negotiating table with a single, dominant, price-setting buyer. Volume growth of 50% a year is genuinely impressive. It is also the growth rate you would expect when a drug trades price for access in a market with enormous unmet need. This is a strong commercial engine, not a pricing-power business.

Indication expansion has been the growth mechanism. A second approval came in October 2024 for first-line gastric and gastroesophageal junction adenocarcinoma. A third followed in June 2025 for first-line cervical cancer in all-comers — meaning irrespective of PD-L1 expression, which materially widens the eligible patient pool. Registrational work in hepatocellular carcinoma continues, and a Phase 3 trial pitting cadonilimab head-to-head against nivolumab in first-line gastric cancer is underway.³

Now the limitation, and it is a significant one. Cadonilimab is not a global asset. It has no approval outside China and no major Western partner. A US Phase 3 trial in HER2-negative gastric cancer has FDA clearance but is early.³ For all its scientific distinction — and being first in the world at anything in oncology is not nothing — cadonilimab today is a China commercial engine, full stop.

That distinction is the single most useful frame for reading Akeso's financials. Cadonilimab plus ivonescimab together produced FY2025 product revenue of RMB 3,033.1 million.³ That line — commercial product sales into China — is the only durable revenue Akeso has. Licensing upfronts are lumpy, one-time, and dependent on deal-signing cadence, and we will see in Section VI exactly how badly that distinction can distort a headline.

What "first in the world" is actually worth

There is a claim implicit in the cadonilimab story that deserves to be tested rather than assumed, because it recurs throughout the Akeso narrative: that being first to approve a novel mechanism confers a durable advantage.

Test it against Akeso's own record with this exact drug. Cadonilimab was approved in June 2022. It was, and as of today remains, the world's first and only approved bispecific of its class.³ Four years of unchallenged mechanistic exclusivity is a long run by oncology standards. And what did that exclusivity produce? Roughly RMB 1.4 billion of half-year sales in a single national market, no ex-China approval, no major Western partner, and a US Phase 3 program still in its early stages.¹¹³

That is not a failure — it is a good drug building a good franchise. But it is a direct, same-company, same-management demonstration that first-in-world status does not automatically convert into global commercial scale. The conversion requires a partner, a multiregional trial package, and a regulator's signature, and cadonilimab has so far obtained none of the three outside China.

Hold that record next to the ivonescimab thesis. The bull case for ivonescimab rests substantially on it being first in a mechanism that everyone now wants. Akeso has already been first in a mechanism once, with a drug of genuine scientific distinction, and four years later the global value of that primacy remains largely unrealized. That does not reject the ivonescimab claim — ivonescimab has the partner, the global trial, and the FDA filing that cadonilimab lacks, which is precisely the difference. But it narrows it: the operative variable is not the scientific first, it is the global development and regulatory execution stacked on top of it. Watch the second thing, not the first.

So cadonilimab funds the company. Ivonescimab is what the company is valued on. Which brings us to the trial that started all of this, and to the far messier question of what it actually proved.

↑ Back to Top

IV. Ivonescimab: The Keytruda Challenger, and What the Data Actually Shows

The molecule, in plain language

Ivonescimab — 依沃西单抗, code AK112 — is a tetravalent bispecific antibody that binds PD-1 and VEGF.²

Unpack that. PD-1 you have already met: the immune brake that checkpoint inhibitors release. VEGF is a different target entirely — it is the signal tumors use to grow their own blood supply. Anti-VEGF drugs like bevacizumab have been standard oncology tools for two decades, starving tumors of blood vessels. Combining PD-1 blockade with VEGF blockade is an old idea with a solid rationale: the abnormal blood vessels tumors build also physically obstruct immune cells from getting in. Cut the vessels back, and the released immune cells can actually reach the target.

The problem is that VEGF blockade throughout the body causes real toxicity — hypertension, bleeding, impaired wound healing. So the engineering question is not "should you block both?" It is "how do you block both without poisoning the patient?"

Akeso's answer is a geometry trick. "Tetravalent" means four binding arms — two for each target. The molecule is designed for cooperative binding: when it latches onto PD-1 on an immune cell, its shape changes in a way that makes it grip VEGF more tightly. The practical consequence is that VEGF blockade becomes concentrated where PD-1-expressing immune cells are dense, which is disproportionately inside the tumor.²

A useful analogy: rather than spraying weedkiller across an entire field, the molecule is engineered so the herbicide activates most strongly where the weeds actually are. Whether that engineering claim translates into a durable clinical advantage is precisely what the trials are meant to answer — and the answer so far is genuinely mixed.

HARMONi-2: the readout that moved markets

HARMONi-2 was a China-only Phase 3 trial testing ivonescimab as a single agent against pembrolizumab as a single agent in first-line PD-L1-positive non-small cell lung cancer. Head-to-head. No chemotherapy in either arm. The cleanest possible comparison.

Topline results were announced on May 30, 2024, presented in full at the World Conference on Lung Cancer that September, and published in The Lancet in March 2025.¹² Median progression-free survival was 11.14 months for ivonescimab versus 5.82 months for pembrolizumab — a hazard ratio of 0.51 with p<0.0001. Objective response rate was 50% versus 39%. The benefit held across both squamous and non-squamous histology and across PD-L1 expression subgroups.²

A hazard ratio of 0.51 means, roughly, that at any given moment a patient on ivonescimab was about half as likely to have their disease progress. In oncology, where a hazard ratio of 0.80 gets a drug approved, 0.51 in a head-to-head against the category leader is a startling number. It was the first randomized Phase 3 win by any drug over pembrolizumab monotherapy — a molecule that had been the reference standard for a decade.

The market reaction is worth documenting because it tells you how the industry read the result, and how violently this stock trades. Summit Therapeutics, holder of ex-China rights, rose as much as 272% intraday.¹ Akeso itself had already crashed roughly 43% in the days before the announcement on leak and sentiment fears, then rallied 37.5% to 50% once the data was confirmed.¹²

Pause on that sequence. Down 43%, then up 40-plus percent, on the same trial, within days, driven by rumor and then by fact. Whatever else this stock is, it is a vehicle in which enormous binary risk is priced and repriced constantly.

The falsification test: what happened when they measured survival

Now the part that the celebratory coverage handled poorly, and that belongs right here rather than buried in a risk section, because it directly tests the strongest version of the bull claim.

Progression-free survival measures how long before a scan shows the tumor growing again. Overall survival measures how long the patient lives. They are correlated but not the same, and oncology history is littered with drugs that improved the first without improving the second. Regulators, and increasingly payers, care most about the second.

In April 2025, an unplanned interim analysis of overall survival in HARMONi-2 was disclosed. Ivonescimab reduced the risk of death by 22.3% versus pembrolizumab — and did not reach statistical significance.¹³ Summit shares fell approximately 36% in a single session.¹⁴

That is the central disconfirming fact in the entire Akeso story, and it needs to be held at full weight. The trial that made the company famous cut the risk of progression roughly in half but, at that interim look, could not demonstrate a statistically significant survival advantage over the drug it was beating.

Akeso's CEO responded publicly. Xia argued that the market reaction had "deviated from the core of the issue," and that HARMONi-2 had been designed primarily as a China registration study — powered for progression-free survival, sized for a regulatory filing — rather than as the definitive global proof of a survival benefit.¹³

Assess that response on its merits, because how management explains a miss is itself data. The argument is technically coherent: a trial powered for one endpoint at a specified event count will frequently be underpowered for a different endpoint at an early interim, and an unplanned interim analysis is statistically the weakest kind of look at survival data. It is also a specific, falsifiable claim rather than a deflection — it names the mechanism and implies a test. That is a materially better pattern than the standard playbook failure mode of blaming macro conditions or giving a non-answer.

But two things are simultaneously true, and the bullish coverage of this episode consistently reported only the first. The explanation was reasonable. It also did not satisfy the market, and it did not change the fact that the survival number was not there. A later interim cut did report overall survival reaching significance — which supports management's framing. It does not retroactively make the April 2025 miss a non-event, because investors who bought the "already proven Keytruda killer" narrative had to sit through a 36% drawdown to find out.

And it was not a one-off. HARMONi-3, testing ivonescimab in combination in squamous NSCLC, also missed at an interim analysis, surprising analysts and pressuring the stock again. Two interim misses on the endpoint that matters most is a pattern, not noise. The correct analytical posture is to weight immature overall survival data with real caution and wait for the mature cuts.

HARMONi: does the China data replicate outside China?

The global Phase 3 trial, named simply HARMONi, was run by Summit across 114 centers spanning Asia, Europe, and North America, in patients with EGFR-mutated NSCLC whose disease had progressed after tyrosine kinase inhibitor therapy. The primary analysis was published in The Lancet Oncology in 2026: progression-free survival of 6.8 versus 4.4 months, hazard ratio 0.52, p<0.0001. Overall survival trended positive but did not reach significance at the primary cut.¹⁵

The most interesting number in that program is not in the headline. It is in the Western subgroup. At 9.2 months of follow-up, the overall survival hazard ratio in Western patients was 0.98 — statistically indistinguishable from no benefit at all. By 23.2 months of follow-up, that same subgroup hazard ratio had improved to 0.76, converging toward what the China data had shown.¹⁵

That trajectory is the single most decision-relevant piece of evidence in this entire story, and it cuts genuinely both ways. The bull reading: survival curves in immunotherapy characteristically separate late, the early Western number was immature, and the drift from 0.98 to 0.76 is exactly what you would expect if the China result does replicate globally. The bear reading: at the first honest look, the drug showed literally no survival benefit in Western patients, and subgroup analyses that improve with follow-up are also exactly what you get from small numbers and multiplicity. Both readings are defensible on current evidence. This is still developing, not settled.

The precedent nobody at Akeso wants cited

There is a specific historical case that has to sit next to any claim that strong China data will carry a drug through the FDA, because it is the closest available analogue and it went badly.

In 2022, Eli Lilly and 信达生物 Innovent brought sintilimab (信迪利单抗, marketed in China as Tyvyt) to the FDA. An Oncologic Drugs Advisory Committee panel voted 14 to 1 that the China-only trial data — run against a comparator the panel considered outdated — was not generalizable to US patients. The application was rejected. Lilly subsequently walked away from ex-China rights entirely.[^16]

That was not an isolated regulatory quirk. Research published in The Lancet Oncology in 2025 found that oncology multiregional trial success rates run structurally lower in the US than in China, and to date no oncology biologics licence application resting solely on China data has cleared the FDA without a supplementary global trial.¹⁵

This is the most important disconfirming data point available for the "ivonescimab is a proven Keytruda-killer" framing, and it does specific analytical work: it narrows the claim rather than destroying it. Akeso and Summit did not make Innovent's mistake — HARMONi is a genuine multiregional trial with Western enrollment, which is precisely the supplementary global evidence the FDA has demanded before. But the honest statement of the current position is "strong China data, maturing global data, unresolved Western survival benefit," not "already validated worldwide." Anyone holding the stronger version is holding a claim the evidence does not yet support.

Where regulation actually stands, as of today

In China, ivonescimab was approved by the NMPA in May 2024 for second-line-and-beyond EGFR-mutated NSCLC, with a positive overall survival trend, and has since added indications; both approved lung cancer indications are NRDL-listed, and a supplemental application in first-line squamous NSCLC is under review.³

In the United States, the FDA granted ivonescimab both Breakthrough Therapy and Fast Track designation. Summit's biologics licence application, in second-line-and-beyond EGFR-mutated NSCLC post-TKI, was accepted for filing on January 29, 2026, with a PDUFA action date of November 14, 2026.¹⁶

That date, roughly ten weeks from now, is the forward event that resolves the near-term thesis. Not a conference abstract, not a press release, not a subgroup analysis. An FDA decision.

The broader development program has scaled aggressively: Akeso disclosed 15 Phase 3 trials for ivonescimab, including five global studies and seven head-to-head comparisons against approved PD-1/PD-L1 therapies, with roughly 70,000 patients treated with the drug to date.³ HARMONi-6, testing ivonescimab in first-line squamous NSCLC against tislelizumab-based therapy, delivered roughly a 34% reduction in death risk and was selected for a plenary presentation at ASCO — the strongest survival package the program has produced.³ And HARMONi-7 is the trial management has flagged as the one that will settle the global survival question outright.

That is the KPI. Not the next press release. The maturing overall survival hazard ratio.

None of this scale would exist without a partner willing to fund it — which is where a Californian former Scientology-linked entrepreneur with a $21 billion track record and a lawsuit enters the story.

↑ Back to Top

V. The Summit Therapeutics Deal: Betting $5B on a Bispecific

In late 2022, Summit Therapeutics was not much of a company. It had no approved products, a modest market capitalization, and an antibiotic program that had not gone anywhere. What it had was Robert Duggan.

Duggan is one of the more unusual figures in modern biotech. He is not a scientist. His pattern is to take control of a company, concentrate it around a single asset he believes the market has mispriced, and fund it personally. He did exactly that at Pharmacyclics, which he ran through the development of Imbruvica and sold to AbbVie in 2015 for $21 billion. That outcome bought him permission — from himself, and from a following of retail investors — to run the same play again.

On December 6, 2022, Summit announced it was in-licensing ex-China rights to ivonescimab from Akeso: the United States, Canada, Europe, and Japan, for $500 million upfront and up to $4.5 billion in milestone payments, a headline of up to $5 billion, plus low-double-digit royalties on sales.⁴ Akeso retained China and the rest of the world. The transaction closed in January 2023.¹⁷ In June 2024 the two companies expanded the arrangement to cover Latin America, the Middle East, and Africa for a further $15 million.³

Read the structure carefully, because it tells you what Akeso chose. It sold geography, not the asset. It kept its home market — where it has an approved drug, NRDL access, and a commercial organization — and monetized the territories where building a sales force would have cost a decade and billions it did not have. For a company that had just barely reached profitable scale in China, that is a rational allocation of scarce capability.

It also created a dependency Akeso does not control, and that dependency has since acquired complications.

The governance problem inside the partner

Summit's side of the deal had to be financed. Duggan financed a large share of it himself, via a bridge loan and rights offering structure totalling roughly $520 million, later repaid in shares. A shareholder lawsuit subsequently alleged that this structure extracted roughly $114 million in value to Duggan and co-CEO Maky Zanganeh at the expense of Summit's public shareholders, characterizing the loan terms in unusually blunt language.¹⁸

This is Summit's governance issue, not Akeso's, and the distinction matters — nothing here implicates Akeso's board or its disclosures. But it is directly relevant to an Akeso investor for a mechanical reason: essentially all of Akeso's ex-China economics for its most valuable asset flow through Summit's ability to execute trials, win approval, build a commercial organization, and stay well capitalized. A partner whose controlling shareholder is being sued over self-dealing in the financing of that very deal is a partner carrying elevated execution and financing risk. Akeso's milestone stream is only as good as Summit's balance sheet and focus.

Was $500 million the right price?

At the end of 2022, a $500 million upfront from a Western partner for a Chinese-originated molecule was an extraordinary number. Nothing at that scale had been normal. In hindsight, though, it was struck at the very beginning of a repricing that has since made it look modest.

The comparisons are instructive. GSK's 2024 deal with 恒瑞医药 Hengrui was worth up to roughly $12 billion with a $500 million upfront — matching Akeso's upfront for a portfolio deal rather than a single asset.¹⁹ AstraZeneca's arrangement with 石药集团 CSPC reached up to $18.5 billion with a $1.2 billion upfront. And most pointedly, BioNTech and Bristol Myers Squibb's collaboration on BNT327 — a rival PD-L1/VEGF bispecific attacking the same mechanism — carried a headline of approximately $11.1 billion, with the $1.5 billion upfront alone triple what Akeso received.²⁰

Zoom out and the category repricing is stark: total announced value of China biopharma out-licensing went from roughly $16.6 billion in 2023 to roughly $41.5 billion in 2024, and reportedly $137.7 billion in 2025.²¹

The fair conclusion — and it needs stating carefully, because deal terms are never strictly comparable molecule-for-molecule, and differ enormously on stage, indication breadth, royalty rates, and territory — is that Akeso priced ivonescimab richly for 2022 and cheaply relative to what a comparable asset would command today. That is not a management failure; nobody was getting $1.5 billion upfronts in 2022, and the deal supplied the capital that funded everything since. But it does mean the single most valuable transaction in Akeso's history captured a smaller share of the asset's eventual value than a later-timed deal would have. Investors sizing the milestone stream should size it against 2022 terms, not against today's headlines.

The trade, on both sides

Summit shares rose roughly 584% to 600% across full-year 2024, at one point carrying a market capitalization above $23 billion with zero approved products.¹² Then came the April 2025 survival readout and the 36% single-session drop.¹⁴

A company with no revenue reaching a $23 billion valuation, then losing a third of it in a day on an interim analysis of a secondary endpoint, is not a story about fundamentals. It is a story about how a single mechanism became a consensus trade on both sides of the Pacific, with position sizing that assumed a resolution the data had not yet delivered. Akeso's own shares have traced a similar, if less extreme, path.

Which raises the obvious question for anyone trying to value the company: underneath all that volatility, what does Akeso actually earn?

↑ Back to Top

VI. The Money: Revenue, the 2023 Profit Mirage, and Capital Allocation

Here is a sequence of numbers that, read without context, would make a value investor's eyes narrow.

FY2021: revenue of RMB 226 million, net loss of RMB 1,075 million. FY2022: revenue of RMB 838 million, net loss of RMB 1,168 million. FY2023: revenue of RMB 4,526 million, net profit of RMB 2,028 million. FY2024: revenue of RMB 2,124 million, net loss of RMB 515 million. FY2025: revenue of RMB 3,056 million, net loss of RMB 1,113 million.²¹

A company goes from persistent losses to a two-billion-renminbi profit, then straight back to losses, then to a larger loss on higher revenue. Something in that sequence is not what it appears.

The claim to falsify: "Akeso turned profitable in 2023"

It did. On paper. The statement is accurate and materially misleading at the same time, and untangling it is the single most important accounting exercise in this business.

FY2023's profit was almost entirely the accounting recognition of Summit's $500 million upfront — roughly RMB 2.92 billion of licensing income landing in a single reporting period.²¹ Strip that out and 2023 was a loss-making year with strongly growing drug sales, which is exactly what a commercial-stage biotech scaling a global trial program should look like.

The proof is what happened next. In FY2024, licensing income normalized to roughly RMB 122 million. Akeso returned to a net loss of RMB 515 million — despite product sales hitting an all-time high, up about 25%.²¹ The underlying business improved. The headline collapsed. Nothing about the operating trajectory changed; only the presence or absence of a one-time payment did.

This has a clean investment implication, and it is the discipline the whole financial section rests on. Judge Akeso on product-sales growth and on the discipline of its R&D reinvestment. Do not judge it on net income, which will swing violently with deal-signing cadence. A future licensing transaction could produce another "profitable year" that says nothing whatsoever about the health of the business — and the absence of one could produce a widening loss in a year where the drugs are selling better than ever. FY2025 was precisely that year: product sales grew 51.5% to a record, and the net loss widened to RMB 1,113 million because R&D spending scaled with a 15-trial Phase 3 program.³²¹

For the record, that spending is not trivial: R&D expense ran RMB 731.2 million in the first half of 2025 alone.¹¹ Running five global Phase 3 studies is not cheap, and the loss line is substantially a choice about pipeline investment rather than an operating failure. Whether it is a good choice depends entirely on whether the trials read out well — which is the same unresolved question from Section IV, now expressed in cash.

The balance sheet, and how it got that way

Akeso ended FY2025 with cash, equivalents, and deposits of approximately HKD 10.6 billion, against roughly HKD 5.75 billion of debt, leaving a positive net cash position near HKD 4.8 billion.²¹ By the standards of a clinical-stage biotech burning capital on a global Phase 3 program, that is a comfortable runway rather than a fragile one.

But how the company got to that cash position is the part that deserves the harder look, because the capital-allocation question for a pre-profit biotech is not "did they buy the right assets" — it is "how much of the equity did they sell, and what did they get for it?"

Cumulative capital raised since the IPO exceeds RMB 5 billion. The IPO itself contributed roughly RMB 2.31 billion. A January 2021 placement raised approximately RMB 1.07 billion. A July 2022 rights issue raised approximately RMB 520 million. Then came a placement of roughly $247 million in October 2024, and a further raise of approximately HK$3.49 billion in 2025.²¹

The effect on shareholders is measurable: shares outstanding rose from roughly 860 million to roughly 920 million over the past year, dilution of about 4% to 4.4%.²¹ There are no dividends and no buybacks. Every renminbi goes to R&D and commercial buildout.

The balanced read: for a company at this stage, paying dividends would be indefensible, and the capital has demonstrably gone somewhere productive — two approved drugs, NRDL access, a global trial program, and an early-stage next-generation pipeline. That is a better record than many peers who raised comparable sums and have less to show. But the honest forward statement is that dilution should be expected to continue rather than taper. A company running fifteen Phase 3 trials, several of them global, has a spending trajectory that current product sales do not yet cover, and the equity market has been its funding mechanism of choice through every previous shortfall. Anyone modelling per-share outcomes needs to model the share count going up.

The valuation the market is paying

Akeso's market capitalization stands at roughly HKD 81 billion, which puts it at something like 20 to 23 times sales.²¹ Hong Kong-listed biotech peers — Innovent, Junshi, BeiGene — trade in the neighbourhood of 10 to 14 times.²¹

That gap is the price of the option. Investors are not paying for RMB 3 billion of Chinese drug sales at a premium multiple; they are paying for the possibility that ivonescimab becomes a global standard of care in the largest oncology indication on earth. Forward earnings multiples run into the hundreds, which is another way of saying earnings are not the operative variable.

What that premium tells you analytically is simple and slightly uncomfortable: the market has already priced in a meaningful probability that the global bet works. The stock is not cheap on the assumption that HARMONi succeeds — it is roughly fairly priced on that assumption, and expensive on any other. That asymmetry is why the November PDUFA date and the maturing survival data matter more to the share price than any quarter of Chinese sales will.

Behind those allocation decisions sits one person who has made essentially all of them for fourteen years.

↑ Back to Top

VII. Michelle Xia and Current Management: Credibility Under the Microscope

There is a photograph that circulated after Forbes profiled her in July 2025 under the headline "How This New Biotech Billionaire Outmaneuvered Merck In China."⁵ Xia does not look like a person who set out to become a billionaire. She looks like what she is: a bench scientist who happens to run a company.

Dr. 夏瑜 Michelle Xia has been founder, Chair, CEO, and President of Akeso continuously since March 2012 — fourteen years in the seat, through the unsalaried years, the Hong Kong listing, two first-in-world approvals, and the most scrutinized oncology dataset to come out of China. She is formally designated a Controlling Shareholder under HKEX listing rules, with a direct holding of roughly 8% of the company plus associated holdings — a stake worth several billion Hong Kong dollars at current prices.²¹

The incentive structure is worth spelling out, because it is unusually clean for a company of this size. Her compensation runs approximately RMB 5.55 million, predominantly salary.²¹ Against an eight-figure-percentage equity position, that pay package is close to a rounding error. Her wealth is almost entirely a function of the share price, not of her employment contract.

The theoretical benefit of that structure is that it should discourage empire-building — growth pursued for the sake of scale rather than value. The theoretical cost, which is less often mentioned, is that a founder-chair-CEO with controlling-shareholder status and a board she has effectively assembled faces limited internal challenge. Concentrated alignment and concentrated power arrive together. So far the alignment has been the dominant effect; the capital-allocation record shows money going into R&D and commercial infrastructure rather than into diversifying acquisitions or vanity projects. But investors should be clear that the check on a bad decision here is the market, not the boardroom.

The finance seat is newer. Dr. Bing C. Wang joined as CFO in July 2024, arriving from Cellectis.²¹ Two years in the chair is not enough to assess a track record, and the timing places a relatively new CFO in charge of exactly the decisions — placement sizing, timing, dilution management — that will determine per-share outcomes through the heaviest spending period in the company's history. That is a flag to monitor, not a criticism.

The credibility test that actually matters

The most useful evidence about management quality is not what they say when things go well. It is what they say on the worst day.

Akeso's worst day, in narrative terms, was the April 2025 survival readout. The response was to argue in public that the market's reaction had "deviated from the core of the issue," and to lay out the specific technical position: HARMONi-2 had been designed and powered as a China registration study, not as the definitive global survival trial.¹³

Grade that against the standard failure modes. Blaming macro conditions or short sellers: absent. Refusing to engage with the number: absent. Vague reassurance that the data was "encouraging": absent. What Xia offered instead was a specific, technical, falsifiable claim — and falsifiable is the operative word, because it implied a test. If the argument was correct, mature survival data from properly powered trials would eventually show the benefit. If it was wrong, it would not.

Partial evidence has since come in on her side: a later interim cut reported overall survival reaching significance, and HARMONi-6 produced roughly a 34% death-risk reduction.³ That is meaningful support for her framing. It is not yet vindication, because HARMONi-3 also missed at interim, and because the argument's real test — HARMONi-7 and the mature global survival data — has not reported.

The calibrated conclusion: the record so far shows a management team that explains misses with specific technical arguments rather than evasion, which is a genuinely better pattern than the peer norm. It does not yet show a management team whose predictions have been proven right, because the prediction in question remains open. Those are different claims and the second one is not available on current evidence.

Narrative consistency, tested across reporting periods

A second, quieter credibility test is whether the story management tells changes when the results change. Compare the framing across the last two reporting cycles.

The 2025 interim results were presented under the banner of "historic results" and an "IO 2.0 + ADC 2.0" strategy said to establish "lasting competitive advantage."¹¹ The full-year 2025 release described the year as "a definitive strategic leap for Akeso across commercialization, global clinical expansion, and our broadening multi-platform innovation ecosystem."³

Two observations. First, the language is consistent between the two periods — the same strategic frame, the same pipeline architecture, no quiet repositioning after a difficult year. Consistency of narrative across filings is a low bar, but it is one that a meaningful number of companies fail, and Akeso clears it.

Second, and less flattering: "lasting competitive advantage" is management's characterization, and Section VIII will show that the structural evidence for durability is thin. A company whose advantage rests on data leadership and regulatory timing in a mechanism a better-capitalized rival is actively replicating has a lead, not a lasting advantage. Neither release quantifies what makes the advantage lasting, and neither engages with the fast-follower question. That is normal corporate communication, not misconduct — but it is exactly the kind of claim an independent reader should discount rather than adopt.

The pattern that emerges across both tests is a management team that is technically candid about specific setbacks and conventionally promotional about strategic framing. That is a better combination than the reverse, and it is roughly the industry norm rather than an outlier in either direction.

On governance more broadly: within the HKEX disclosures and financial reports reviewed for the periods covered here, no auditor qualifications, no confirmed short-seller report, and no material AGM shareholder dissent surfaced at Akeso itself.²¹ That is a bounded negative result over a defined set of records, not a clean bill of health — related-party transaction detail was not independently verified beyond summary disclosures, and the absence of discovered problems is not evidence of their absence. The governance issue that is documented in this story sits at Summit, not Akeso.

Management quality only matters, in the end, relative to the competitive field it has to operate in. And that field has changed dramatically since 2024.

↑ Back to Top

VIII. Industry Structure & Competitive Landscape

To understand why a Guangdong biotech's press release could move Merck's market capitalization, you have to understand the size of the pool being fought over.

The global PD-(L)1 inhibitor market was worth roughly $57.75 billion in 2025, with credible projections putting it somewhere between $150 billion and $210 billion by the early-to-mid 2030s.¹² Today it is effectively an oligopoly. Merck's Keytruda holds approximately 41% of the category and leads in non-small cell lung cancer, the single largest indication. Bristol Myers Squibb's Opdivo leads in melanoma and renal cell carcinoma. Roche's Tecentriq leads in small cell lung cancer.¹²

That structure has been stable for years. It is about to become extremely unstable, for a reason that has nothing to do with Akeso.

Why this moment, specifically

Keytruda faces a US patent cliff in 2028.

That single fact reorganizes the strategic incentives of every incumbent. When your $30-billion-a-year drug is going to face biosimilar competition in a defined number of quarters, defending monotherapy market share is not a strategy — it is a countdown. The rational move is to find the next-generation asset that makes the old molecule's genericization survivable, and the consensus answer across the industry has converged on PD-(L)1/VEGF bispecifics.

This is why ivonescimab landed the way it did. It was not attacking an entrenched incumbent from nowhere. It arrived at the precise moment when the incumbents were most motivated to either co-opt the mechanism or out-innovate it — and most sensitive to evidence that someone else had gotten there first.

The domestic field

Akeso's Chinese peers are, on this specific mechanism, mostly behind. Innovent has sintilimab and earlier-stage PD-1/VEGF bispecific work. Junshi has toripalimab and roughly RMB 2.5 billion in FY2025 revenue, still largely China-only.¹² Hengrui is a broad-portfolio player rather than a focused competitor here.

The peer worth studying closely is BeiGene, whose tislelizumab (Tevimbra) generates on the order of $171 million quarterly and is approved in 46 markets.¹² BeiGene has the strongest global commercial footprint of the Chinese cohort — and it got there by making the opposite strategic choice to Akeso's. Rather than out-licensing ex-China rights, BeiGene built its own international commercial organization, at enormous cost and over many years.

That contrast is the cleanest available natural experiment in Chinese biotech globalization, and it is genuinely unresolved. Akeso's approach is capital-efficient and fast: someone else pays for the trials, someone else builds the sales force, Akeso collects upfronts, milestones, and royalties. BeiGene's approach is slow and expensive but captures the full economics and, crucially, keeps control. If ivonescimab becomes a multi-billion-dollar global drug, Akeso will have licensed away most of its value; if it stalls, Akeso will have made the right call. Neither model has yet been proven superior, and an investor's view on which one wins should be held loosely.

The competitor that actually threatens the thesis

Everything above is context. This is the argument.

BNT327 is a PD-L1/VEGF bispecific antibody — the same mechanistic class as ivonescimab, attacking the same biology — now in global Phase 3 development across small cell lung cancer, non-small cell lung cancer, and triple-negative breast cancer. It is backed by BioNTech and Bristol Myers Squibb, under a collaboration worth approximately $11.1 billion in total with $1.5 billion paid upfront.²⁰

That is a scaled, deep-pocketed, fast-following rival explicitly constructed to contest the mechanism Akeso pioneered.

And it is the single clearest reason the bull case may not work, because it attacks the moat claim at its foundation. Akeso's advantage in PD-1/VEGF bispecifics is that it got there first and has the deepest maturing dataset. That advantage is real, and it is temporal. Being first in a mechanism is not a durable barrier when a competitor with a Merck-sized checkbook can fund a fast-follower into the same indications on a parallel timeline. There is no patent on the idea of combining PD-1/PD-L1 blockade with VEGF blockade in one molecule — only on specific molecules — and the clinical development capital required to run global Phase 3 programs is precisely the resource BioNTech and BMS have more of than Akeso does.

Reading the structure properly

Apply Porter's five forces and the picture is not flattering to the moat narrative.

Rivalry is intensifying rapidly, with multiple well-funded PD-(L)1/VEGF bispecifics racing toward the same indications on overlapping timelines. Buyer power is high and concentrated on both sides of the Pacific — NRDL price negotiation in China is a single-buyer monopsony, and US payers and pharmacy benefit managers exert their own considerable leverage. Supplier power is genuinely manageable, and this is where Akeso's original decision to build in-house manufacturing pays off: it is not queueing for CDMO capacity or negotiating with a contract manufacturer holding it hostage. Substitution threat is real and rising — antibody-drug conjugates, cell therapies, and other combination immuno-oncology regimens all compete for the same treatment lines. Barriers to entry are moderate: bispecific engineering and Phase 3 capital are real barriers, but BNT327 is standing proof they are not insurmountable.

Run Hamilton Helmer's 7 Powers and you get to the same place by a different route. Akeso's strongest claim is a hybrid of scale economies and cornered resource — first-to-approval status plus the richest maturing dataset in the class. But examine what that rests on and it is data leadership and regulatory timing, not switching costs, not network effects, not a counter-positioning that competitors are structurally unable to copy. Oncologists do not face switching costs when a better molecule arrives; they switch. There is no installed base to defend.

The conclusion, stated plainly: Akeso's advantage is temporary, not structural. That does not make it worthless — a multi-year head start in a market heading toward $150 billion-plus is enormously valuable, and first movers in oncology mechanisms frequently retain durable share through physician familiarity and trial-data depth. But it should not be described as a moat, and the article-level framing should not pretend otherwise.

Temporary advantages are vulnerable to specific, identifiable events. Which ones?

↑ Back to Top

IX. Risk Radar

Four risks matter here. Several more that a generic checklist would list do not, and are omitted.

Geopolitical and legislative. In December 2025, the US BIOSECURE Act became law, folded into the FY2026 National Defense Authorization Act. It restricts federal procurement involving designated Chinese "biotechnology companies of concern."¹² Akeso is not currently named, and the statute as written does not directly restrict a Summit-marketed, FDA-approved product.

But the mechanism to watch is broader than the specific text. The law establishes that the US Congress is willing to legislate against China-origin biotech as a category, and it creates precedent and machinery for expansion. The practical risk is not a ban; it is friction — in how a China-originated asset gets formulary placement, hospital adoption, and reimbursement in the United States even after FDA approval, and in how much political risk premium a Western partner attaches to Chinese-licensed assets in future deals. That is a latent, hard-to-quantify drag rather than an event risk, which is exactly why it tends to be underpriced.

Translation and regulatory execution. Covered substantively in Section IV, so only the live variable is restated here: whether the China efficacy signal reproduces in Western patients remains formally unresolved, with the Western-subgroup survival hazard ratio still maturing and the sintilimab precedent establishing that the FDA has rejected China-heavy oncology packages before. The concrete resolution event is the November 14, 2026 PDUFA date.¹⁶

Concentration. This is the risk most specific to Akeso's chosen structure. The entire ex-China value proposition for its most important molecule runs through one partner and one mechanism. Summit's balance sheet, Summit's governance — including the litigation over its financing structure — and Summit's trial execution are all now Akeso risks, sitting outside Akeso's control. If Summit stumbles operationally or financially, Akeso's remedies are contractual and slow. Diversification exists in the pipeline but not yet in the revenue.

Binary-event volatility. Since 2024, both Akeso and Summit shares have moved 30% to 50% on individual trial readouts, repeatedly, in both directions.¹²¹⁴ It is reasonable to expect near-term price action to continue tracking discrete data events — the maturing HARMONi-7 survival data, the PDUFA decision — more than any quarterly operating result. For an investor whose process assumes steady compounding, that is not a market quirk to be endured; it is a structural feature of the security, and one that should inform position sizing rather than being discovered after the fact.

Set those risks against the affirmative case, and the two sides of this argument are unusually well-matched.

↑ Back to Top

X. Bull vs. Bear: The Investment Case

The bull case

Start with what is not in dispute. Akeso is a founder-led originator with two approved drugs — both first-in-world in their mechanistic class — that are commercially ramping at roughly 50% annual growth with full national reimbursement in the world's second-largest pharmaceutical market.³ That is a real, growing, cash-generative base, and it exists independent of any global outcome.

On top of that base sits a validated global partnership with economics that have already partly paid: $500 million banked, up to $4.5 billion in milestones outstanding, plus royalties, with a Western partner that has completed a multiregional Phase 3 program and secured an FDA filing with a decision date.⁴¹⁶

The scientific claim is the strongest part. Akeso originated a mechanism that the largest pharmaceutical companies on earth are now racing to replicate — the clearest possible market validation, since BioNTech and BMS committed $11.1 billion to a molecule in the same class.²⁰ Being copied by people with more money than you is unpleasant, but it is also confirmation that you were right.

And on management: faced with a bad data print and a collapsing share price, Xia produced a specific technical argument rather than deflection, and subsequent data — the later significant survival cut, HARMONi-6's roughly 34% death-risk reduction — has partly supported it.³¹³

The bear case

Now the other side, and it is not merely the mirror image.

The single most important data point in the entire thesis — whether ivonescimab delivers superior overall survival to pembrolizumab in a global population — is unresolved, and has missed statistical significance at more than one interim look.¹³ Not once. That is the specific evidence that narrows the bull claim from "proven Keytruda-killer" to "strong progression data, survival benefit probable but not established."

The financial narrative has a similar problem. The company's one profitable year was a licensing-income artifact, not operating reality, and the stock carries a premium multiple to Hong Kong biotech peers that only makes sense if the global bet resolves favorably.²¹ An investor buying today is not being offered the option cheaply.

A mechanism-identical competitor with a bigger deal behind it is racing the same indications, and Akeso's advantage in that race is temporal rather than structural.²⁰ Dilution has been continuous and should be expected to continue as the trial program scales.²¹ And Summit's governance and capitalization add a layer of execution risk Akeso does not control.¹⁸

The activist's version

What would a skeptical long/short investor actually attack? Probably three things. First, the disclosure gap between headline revenue and product revenue — a company whose reported profitability swings on the timing of one-off payments invites the question of whether the presentation emphasizes the flattering line. Second, the founder-chair-CEO-controlling-shareholder concentration, with a two-year-old CFO, running the heaviest capital-raising period in the company's history. Third, and most pointedly: the valuation embeds an outcome that the company's own trials have twice failed to demonstrate at interim, and management's defense of those misses — however technically sound — is a defense of an unfalsified hypothesis, not of a proven result.

None of those are disqualifying. All three are fair.

The honest net framing

This is not a moat story. The evidence for a durable structural advantage is thin, and Section VIII laid out why: no switching costs, no network effects, no counter-positioning, a mechanism that a better-funded rival is actively replicating.

It is a data-and-timing story. The right question is not "does Akeso have a structural edge" — the answer to that is mostly no, and anyone underwriting the stock on moat logic is underwriting the wrong thing. The right question is "does the China readout confirm globally, and does Akeso convert its head start into entrenched share before the fast-followers arrive?"

That question is genuinely open. It has two near-term resolution points, and it deserves to be left open rather than resolved by assertion in either direction.

↑ Back to Top

XI. Playbook: Business & Investing Lessons

Platform-first beats asset-first, but only if you can survive the wait. Akeso's willingness to build in-house discovery, bispecific engineering, and manufacturing in its early years — when in-licensing would have been faster and cheaper — is what let it originate rather than follow. That is the transferable lesson for evaluating any biotech claiming a "me-too to first-in-class" transition: look at whether they built the capability or rented it, because rented capability cannot produce a novel molecular geometry on demand. The caveat is equally important: the platform strategy nearly starved the company in its first years, and for every Akeso there are companies that ran out of money before the platform produced anything.

Licensing is a legitimate globalization strategy, not a consolation prize. Rather than building US commercial infrastructure, Akeso sold geography for cash, milestones, and royalties — a template now standard among Chinese originators. The contrast with BeiGene's decision to build its own global organization is the sharpest available illustration of the trade-off between capital efficiency and value capture. Both are defensible. Neither is obviously right yet, and an investor who claims certainty about which model wins is claiming knowledge the evidence does not contain.

Interim data is not data. The repeated 30% to 50% single-session moves in both Akeso and Summit shares, in both directions, on interim analyses of secondary endpoints, are a case study in how markets extrapolate from immature information. The specific discipline: progression-free survival headlines move stocks; overall survival determines outcomes; and the gap between the two is where the money is lost. Waiting for mature survival data costs upside in the cases where the drug works. It also avoids the drawdowns in the cases where it does not.

How management explains a miss is usable evidence. Xia's April 2025 response is a template worth borrowing when assessing any management team: did they offer a specific, technical, falsifiable explanation with an implied test, or vague reassurance and blame-shifting? The first is not proof of being right — it is proof of being checkable, which is the more useful property. Then actually check it when the test arrives.

↑ Back to Top

XII. Epilogue: What to Watch

The forward calendar is unusually concrete for a company this speculative.

The FDA's PDUFA action date of November 14, 2026 for ivonescimab plus chemotherapy in second-line-and-beyond EGFR-mutated NSCLC is the nearest binary event.¹⁶ Beyond it: maturing HARMONi-7 survival data, which management itself has identified as the trial that settles the global question; continued cadonilimab indication expansion and reimbursement listings in China; and further BNT327 Phase 3 readouts, which serve as the direct competitive benchmark for the mechanism.³²⁰

Three metrics are worth tracking, and only three.

One: the ex-China overall survival hazard ratio, and its maturity. This is the number that resolves the bull/bear split. Not progression-free survival, not response rate, not the China data. The Western-population survival benefit, as it matures in HARMONi and HARMONi-7. Everything else in this story is downstream of it.

Two: China product-sales growth, cadonilimab and ivonescimab combined. This is the real underlying business, and it is the only line in the accounts that cannot be distorted by licensing noise. Watch it net of the headline revenue and net income figures, both of which will lie to you in any year with a large upfront payment.

Three: net cash burn and the cadence of equity placements. This is the direct read on how much further dilution the global program requires before it becomes self-funding. A company that raises less often than expected is telling you the trials are going well and the milestones are arriving; one that raises more often is telling you the opposite, generally before the data does.

There is one more thing on the board, and it deserves a mention proportional to its stage: the emerging "IO 2.0 + ADC 2.0" pipeline. AK150, a trispecific antibody now in the clinic. AK146D1, a Trop2/Nectin4 bispecific antibody-drug conjugate. AK138D1, a HER3 ADC. AK139, an IL-4R/ST2 bispecific moving toward autoimmune and respiratory disease — a genuine step outside oncology.³¹¹

This is real strategic optionality: diversification beyond a single mechanism and, in the immunology programs, beyond cancer entirely. It is also early-stage and pre-revenue. Akeso's own history is the relevant guide to how much weight it should carry — the company has converted exactly two programs from discovery to meaningful revenue in fourteen years, both of them taking roughly a decade. Certification, publication, and clinical entry are not commercialization. The next-generation pipeline earns a mention as a future call option on the platform. It does not earn weight in the current investment case.

↑ Back to Top

XIII. Outro

Akeso is a genuine scientific first and a genuine open question, and the discomfort of holding both ideas at once is the whole point.

The company has already done the hard part once: originate a novel molecule from a standing start in a provincial Chinese city, get it approved at home ahead of the entire world, and persuade a Western partner to pay real money upfront for the rights. Very few companies anywhere have done that. Nobody in Chinese biotech had done it in quite this way before.

Whether Akeso does the harder part — prove the survival benefit in Western patients, clear the FDA, and hold off well-funded fast-followers long enough to entrench — is still being written, in trial data that has not fully matured, in a regulatory decision ten weeks away, and in the choices of a founder who has bet essentially her entire net worth on the answer.

The press release that made Merck blink was real. What it proved is narrower than what it was reported to prove. Both of those statements will remain true until the survival curves say otherwise.

Akeso, Inc.

Table of Contents

Akeso: The Bispecific That Made Merck Blink

I. Introduction & Episode Roadmap

II. Origins: Building an Antibody Platform From Scratch (2012–2020)

III. Cadonilimab: The World's First Approved Immunotherapy Bispecific

What "first in the world" is actually worth

IV. Ivonescimab: The Keytruda Challenger, and What the Data Actually Shows

The molecule, in plain language

HARMONi-2: the readout that moved markets

The falsification test: what happened when they measured survival

HARMONi: does the China data replicate outside China?

The precedent nobody at Akeso wants cited

Where regulation actually stands, as of today

V. The Summit Therapeutics Deal: Betting $5B on a Bispecific

The governance problem inside the partner

Was $500 million the right price?

The trade, on both sides

VI. The Money: Revenue, the 2023 Profit Mirage, and Capital Allocation

The claim to falsify: "Akeso turned profitable in 2023"

The balance sheet, and how it got that way

The valuation the market is paying

VII. Michelle Xia and Current Management: Credibility Under the Microscope

The credibility test that actually matters

Narrative consistency, tested across reporting periods

VIII. Industry Structure & Competitive Landscape

Why this moment, specifically

The domestic field

The competitor that actually threatens the thesis

Reading the structure properly

IX. Risk Radar

X. Bull vs. Bear: The Investment Case

The bull case

The bear case

The activist's version

The honest net framing

XI. Playbook: Business & Investing Lessons

XII. Epilogue: What to Watch

XIII. Outro

References