---
title: "Anthropic at $83 and Its Push Into Pharma"
url: https://drugnews.com.tw/articles/2026-11-23-anthropic-ipo-ai-drug-development-en.html
date: 2026-11-23
lang: en
category: "商業分析系列"
tags: ["Anthropic","Claude","AI drug development","Mithril Medicine"]
summary: "Anthropic has measurable workflow results, while its own drug research is still preclinical."
json: https://drugnews.com.tw/articles/2026-11-23-anthropic-ipo-ai-drug-development-en.json
publisher: Drugnews｜藥時事 (https://drugnews.com.tw/)
---
## 三句看懂

- Anthropic has measurable workflow results, while its own drug research is still preclinical.
- Enveda measured human document review time.
- Insilico and XtalPi show two ways to structure the business.

## 關鍵數字（附來源）

| 項目 | 數值 | 日期 | 來源 |
|---|---|---|---|
| Anthropic 最終發行價 | 83 美元／股 |  |  |
| Anthropic 初步發行價區間 | 70–78 美元／股 |  |  |
| Anthropic Nasdaq 首日交易 ANT | 2026-11-19 美東日期 |  |  |
| 發行價估值 | 1.5 兆 美元 |  |  |
| 截至 2026/9/30 前十二個月營收 | 35.282 billion 美元 |  |  |
| Anthropic 已公開披露收購 Mithril | 2026-10-27 公司公布日期 |  |  |
| 收購前 Mithril 為一人公司 | 1 人 |  |  |
| 合作外部實驗室 | 11 家 |  |  |
| Mithril 罕見肌肉失養症計畫 | 2 種疾病 |  |  |
| 單一計畫自構想到小鼠數據 | 18 月，少於 |  |  |
| 最早人體試驗窗口 | 2028 年底 公司有條件目標 |  |  |
| 公開合作案例 | 6 家 |  |  |
| 單一 CAR-T 構件版本長度 | 70% 約縮短 |  |  |
| 模型提出的抗體設計 | 35,000 約 |  |  |
| 設計生成期間 | 1 月，約 |  |  |
| 實作及測試的變體 | 1,010 約 |  |  |
| 實作及測試所用時間 | 10 週，約 |  |  |
| 製造文件測試量 | 1,061 頁 |  |  |
| 人工製造文件審查工時 | 89% 減少 |  |  |
| Pleiades 早期流程測試範圍 | 45 組 |  |  |
| BMS 主管描述傳統初步分析用時 | 數天 定性敘述 |  |  |
| REC-1245 DAHLIA | 1/2 期；已有第一期早期資料 |  |  |
| 中國 GENESIS-IPF-3 首例給藥 | 2026-09-10 三期公司公告 |  |  |
| 隨機二期試驗人數 | 71 人 |  |  |
| 晶泰與 Pierre Fabre 公開合作 | 2026-11-09 港交所披露日 |  |  |

Anthropic priced its IPO at $83. Before listing, it had already bought a one-person drug developer.

## 01｜One founder, eleven labs

Claude's developer began trading on Nasdaq as ANT on November 19. [3] The final price exceeded the preliminary range of $70 to $78. [1][2]

Twelve-month revenue through September came to $35.3 billion, calculated from the prospectus. [1]

That figure covers the whole company. [1][2][3]

Nasdaq's Economic Institute put the valuation at about $1.5 trillion at the offer price. [14]

Anthropic disclosed its acquisition of Mithril Medicine on October 27, alongside six projects with drug developers and investors. [4]

The purchase price was undisclosed. The projects show where Claude can make itself useful even when a medicine is years away from a trial. [4]

![Figure 1｜Four points of entry for AI](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/figure-01.png)
*From candidate design to trial planning, each step produces a different piece of evidence.*

Mithril founder Cody Tranbarger used Claude to coordinate work with 11 outside labs. [4]

![11 · outside labs](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/inline-card-number-23515f2adfa5.png)

One muscular dystrophy program went from an idea to mouse data in under 18 months. [4] Mithril is working on two rare forms of muscular dystrophy. [4]

The team still has to choose a candidate and clear the work required for human testing.

Anthropic gave late 2028 as the earliest possible start for one trial, conditional on the work holding up. [4]

That date sets a clock for the project. [4]

In [our November 1 Chinese-language analysis](https://drugnews.com.tw/articles/2026-11-01-anthropic-ipo-ai-drug-lab.html), the business question was who would take the drugs forward.

The IPO brings a new audience to the same question. The six cases help put a price on specific work.

## 02｜What the partners could measure

Generate Biomedicines worked on a component of an experimental CAR-T approach for solid tumors.

One CAR-T component was about 70% shorter than the standard version. [4] The team also reported lab improvements.

![70% · shorter than the standard version](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/inline-card-number-ff359014f702.png)

Patient safety and clinical efficacy remain open. [4]

Divergence Labs used Claude to generate about 35,000 antibody designs in roughly a month for a food allergy project. [4]

Some scored as hits in early testing. The team still needs to bring its early findings together in a single candidate. [4]

![Figure 2｜Six partners and their assignments](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/figure-02.png)
*One case counts designs, another measures time, and another evaluates trial plans.*

BigHat Biosciences went further into the lab.

It built and tested about 1,010 variants over roughly 10 weeks, with improvements on selected properties. [4]

The case offers experimental evidence for lead optimization. It leaves the clinical candidate and timeline open. [4]

Enveda used Claude to review 1,061 pages of manufacturing records. [4] Human review time fell by 89%. [4]

The company also tried Claude on regulatory drafts. It still had people review the work, and the case did not report shorter agency review. [4]

The Column Group was setting up Pleiades to see how much drug research a small team could take on.

It evaluated an early workflow on 45 target-disease pairs. [4] An evaluation of that size tells us the scope of the test.

![45 · target-disease pairs](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/inline-card-number-cbf481254584.png)

The public account gives no measured improvement in the rate of successful programs. [4]

Bristol Myers Squibb used Claude to evaluate options for an early clinical program.

A company executive said that an initial analysis would otherwise have taken days.

The case gave no standardized time saving or evidence of a higher likelihood of trial success. [4][5]

Taken together, the cases show how quickly a team can get to the next decision.

Evidence for that decision can be an assay result, a set of manufacturing records or a clinical plan.

Each still needs someone who can judge its quality.

## 03｜Who owns the next step

The same questions work across the business models. What task does the company own? How far has that specific program advanced?

What could an outsider check next?

![Figure 3｜Three ways into AI drug development](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/figure-03.png)
*Some companies sell research capabilities. Others develop their own medicines or build tools around their own data.*

| Company and route | Role | Public progress | Next checkpoint |
| --- | --- | --- | --- |
| Anthropic, model supplier [4] | Rare disease research [4] | Mouse data [4] | Candidate selection [4] |
| Recursion, AI biotech [6] | Develops REC-1245 [6] | Early Phase 1 data [7] | Expansion data [7] |
| Lilly, in-house pharma [8] | Models using proprietary data [8] | TuneLab launched [8] | Evidence on a named asset [9] |
| Insilico, AI biotech [10] | Develops rentosertib [10] | First Phase 3 dose in China [10] | Phase 3 lung function data [10] |
| XtalPi, discovery partner [12] | Discovery. Pierre Fabre leads clinical development [12][13] | Agreement announced [12] | A named candidate [12] |

Insilico's randomized Phase 2a study enrolled 71 people. [11] Narrowing a list of candidates was one hurdle. Insilico faces a different one now.

![71 · people](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/inline-card-number-68a0535ed6e7.png)

It announced the first dose in a Chinese Phase 3 trial on September 10. [10]

A larger trial will show whether the earlier lung function signal holds up. [10][11] The XtalPi deal puts the commercial handoff earlier.

Under the agreement, XtalPi works on discovery and Pierre Fabre takes responsibility for clinical and commercial development.

A named candidate would give readers a stronger signal than another announcement about the platform. [12][13]

Lilly can approach the same work from inside pharma.

It already has the research data needed to build its own platform, and it gives other biotechs access through TuneLab.

Its joint lab with NVIDIA shows why general-purpose AI companies also face incumbent suppliers. [8][9]

## 04｜Two dates worth tracking

AI can make it cheaper to ask a question and quicker to close out an unpromising answer. The next expense is physical.

Drugmakers have to manufacture trial material, recruit participants and wait for the data.

Companies that control those decisions also carry the risk if the data disappoint.

![Figure 4｜The listing and Mithril's research timeline](https://drugnews.com.tw/assets/articles/anthropic-ipo-ai-drug-development-en/figure-04.png)
*The acquisition was already public before the IPO. The late-2028 trial window remains conditional.*

The partnership test can arrive sooner.

Watch whether Enveda, BMS or the other partners put Claude to work on a second identifiable asset, and whether they are willing to say where it made a difference.

For Mithril, first dosing will be a clearer milestone than another set of mouse data.

By the end of 2027, a partner will disclose using Claude for clinical planning or filing on a second named asset, while Mithril will still be awaiting its first human dose.

This article is for industry analysis and does not constitute investment or medical advice.

## References

1. Final Prospectus｜SEC EDGAR / Anthropic｜2026-11-18｜https://www.sec.gov/Archives/edgar/data/2130357/000162828026074944/anthropicpbc-424b4.htm
2. Roadshow launch announcement｜Anthropic｜2026-11-10｜https://www.anthropic.com/news/roadshow-launch
3. Anthropic Marks its IPO｜Nasdaq｜2026-11-19｜https://www.nasdaq.com/newsroom/anthropic-marks-ipo-at-nasdaq
4. Early Progress on Accelerating Drug Development with Claude｜Anthropic｜2026-10-27｜https://www.anthropic.com/research/early-progress-drug-development
5. Strategic Agreement with Anthropic｜Bristol Myers Squibb｜2026-05-20｜https://news.bms.com/news/corporate-financial/2026/Bristol-Myers-Squibb-Announces-Strategic-Agreement-with-Anthropic-to-Position-Claude-Enterprise-as-the-Shared-Intelligence-Platform-Across-Its-Global-Operations/default.aspx
6. First Quarter Update｜Recursion｜2026-05-06｜https://ir.recursion.com/news-releases/news-release-details/recursion-reports-first-quarter-financial-results-and-provides
7. Third Quarter Update｜Recursion｜2026-11-02｜https://ir.recursion.com/news-releases/news-release-details/recursion-reports-third-quarter-2026-financial-results-and
8. TuneLab Launch｜Eli Lilly｜2025-09-09｜https://investor.lilly.com/news-releases/news-release-details/lilly-launches-tunelab-platform-give-biotechnology-companies
9. Lilly and NVIDIA Co-Innovation Lab｜Eli Lilly｜2026-01-12｜https://lilly.gcs-web.com/node/53651/pdf
10. First Dose in GENESIS-IPF-3｜Insilico Medicine｜2026-09-10｜https://insilico.com/news/isn1009261-insilico-medicine-doses-first-patient-genesis-ipf-3
11. Randomized Phase 2a Trial｜Nature Medicine｜2025-06-03｜https://www.nature.com/articles/s41591-025-03743-2.pdf
12. XtalPi Exchange Filing｜XtalPi / HKEX｜2026-11-09｜https://www1.hkexnews.hk/listedco/listconews/sehk/2026/1109/2026110900037.pdf
13. Pierre Fabre and XtalPi Research Agreement｜Pierre Fabre｜2026-11-09｜https://www.pierre-fabre.com/en/news/ai-drug-discovery
14. IPO Valuation Context｜Nasdaq Economic Institute｜2026-11-20｜https://www.nasdaq.com/economic-institute/chartstopper-112026
15. Initial Public Offering Close｜Anthropic｜2026-11-20｜https://www.anthropic.com/news/ipo-close
