The AI claim: data is the moat, safety is the wall
The AI story inside October's oncology news cycle is an interview, not a product launch. AstraZeneca's Puja Sapra describes a 'lab of the future' in which, in her words, "Where a self-driving car uses sensors and models to navigate its environment, this system uses AI to make predictions, robotic systems to execute experiments, and instruments to generate data."1
The competitive claim is about data, not model architecture. Sapra says: "Data is our differentiator. We've built an intentionally diverse portfolio across multiple disease areas and drug types. All of that data empowers us to fine-tune frontier AI models with richer, more representative training sets."1 The practical point for AI watchers is that the claimed advantage is proprietary, multimodal training data, not access to a better model.
The same interview is candid about the limits. Sapra says: "One of the hardest problems in de novo design is predicting whether a computationally generated molecule will be safe in the human body."1 Her proposed fix is virtual trials on advanced cell systems and micro-scale organ models. She says these "have the potential to generate enhanced biological signals without traditional testing bottlenecks, and they're a critical missing piece in closing the loop between AI-generated designs and clinical-ready candidates."1 That is a statement of potential, not a demonstrated result.
Agents, but with humans in the loop
The interview describes a shift toward agentic systems that generate candidates while predicting efficacy and safety. Sapra puts it this way: "The complexity of the biology goes hand-in-hand with the design of the molecule."1 She also expects automation to change the data supply: "This will generate AI-ready data at a scale that traditional workflows cannot match."1
On autonomy she is explicit that people stay involved: "Scientists will work hand-in-hand with these model systems. There will be a world where models will design molecules, then scientists will work with the systems to test those molecules and put all that data together."1 She adds that "human oversight remains at the heart of this approach—ensuring explainable and ethical AI for the benefit of patients."1
How much to trust the source
Via News measures how often claims extracted from each source hold up when checked. For the article carrying these quotes, only 6% of 16 checked claims held up.1 That is a small sample and a poor score. Read the quotes as what the speaker said, not as evidence that the approach works. The dossier contains no data on how many AI-designed molecules have reached patients, and we are not going to imply it does.
The money side: what a pre-revenue oncology bet costs
The AI discovery pitch comes from a company with approved drugs. The cleaner test of the cluster's economics is Summit Therapeutics, whose ivonescimab appeared in 10 presentations across many tumour types at ESMO 2026.2 Its filings show a large and growing cost of getting there.
Summit reported zero revenue in both FY2023 and FY2024.3 Its net loss was $614.9 million in FY2023 and $221.3 million in FY2024, then $1.08 billion in FY2025.3 The FY2025 loss is about 4.9 times the previous year's and about 2.6 times the $419.4 million of cash Summit held at Q2 2026.3 Loss per share went from -$0.99 in FY2023 to -$0.31 in FY2024 and -$1.44 in FY2025.3
Cash actually spent on operations was smaller than the headline loss. Operating cash flow was -$76.8 million in FY2023, -$142.1 million in FY2024 and -$322.9 million in FY2025.3 In FY2023 the operating loss was $609.7 million against a cash outflow of $76.8 million.3 The dossier does not say what explains that gap, so we do not speculate. The reader-relevant fact is that the accounting loss and the cash burn are different numbers.
Cash tells a stop-start financing story. Summit held $71.4 million at FY2023 year-end and $104.9 million at FY2024 year-end.3 In FY2025 it held $225.3 million, then $106.5 million at Q1 2026, then $419.4 million at Q2 2026.3 The sharp rises between quarters point to new money arriving, but the dossier does not identify the source, so we do not attribute it. One caution: the dossier's capex entries ($128,000, $139,000 and $657,000 for FY2023 to FY2025) look small beside these other figures, and we have not independently confirmed the units.3
The clinical catalyst
The clinical news behind Summit's drug comes from its partner. Akeso's Phase III HARMONi-GI1 reported positive overall survival for ivonescimab plus chemotherapy against durvalumab plus chemotherapy in first-line advanced biliary tract cancer. It was selected as a Late-Breaking Abstract for the ESMO Presidential Symposium.2 The dossier gives no survival figures, so we cannot say how large the benefit is. Durvalumab is an AstraZeneca-developed drug, so the comparison is a head-to-head against a large competitor's product.4
Regulators and launches
AstraZeneca announced US availability of ZEGFROVY (sunvozertinib) on October 22, 2026, for adults with previously treated NSCLC carrying EGFR exon 20 insertion mutations.5 The company's description is that it is the only approved oral, single-agent targeted therapy for these patients after platinum chemotherapy.5 That is a company claim in a press release. Gilead's Trodelvy plus Keytruda received a positive CHMP opinion in first-line PD-(L)1-positive triple-negative breast cancer, which the company says would help establish Trodelvy as a backbone therapy across PD-L1 status in the EU.6 We hold no source documents in this dossier for the other regulatory items in the cluster, so we do not detail them.
Press-release sources carry their own reliability caveat. Across the news wire documents used here, 53% of 6,520 checked claims held up.5 That is better than the interview source but far from a guarantee.
The uneven capital picture
Below the large caps, financing conditions diverge. Spago's board resolved on October 22, 2026 to apply for delisting from Nasdaq First North Growth Market, subject to an extraordinary general meeting, to improve conditions for financing the next phase of Tumorad development.2 That meeting is planned for November 19, 2026. The delisting application is intended no earlier than January 25, 2027, with last trading expected in February 2027.2 Akari, by contrast, is pitching a niche. Its statement is that ADC-refractory patients are a significant, growing unmet need and an exclusive opportunity for its novel payload, which current and most future ADCs will not be able to access.7 Akari plans Phase 1 clinical development of AKTX-101 in mid-2027.7 That is a company claim about its own market.
What to watch
- Whether AstraZeneca shows any measured result from AI-designed molecules, particularly on safety prediction, the problem its own spokesperson calls among the hardest.1
- Detailed intismeran results from Moderna and Merck at ESMO on October 24, and any regulatory filing discussions that follow.2
- Summit's next cash and loss figures against the $419.4 million it held at Q2 2026.3
- The Spago shareholder vote on November 19, 2026.2