ByteDance Spins Off AI Pharmaceutical Company Anew Labs, Completes $290 Million
AI pharmaceutical company Anew Labs has completed its first round of independent financing, raising $290 million. Investors include HSG, IDG Capital, GL Ventures, 5Y Capital, Gaorong Capital, and several domestic pharmaceutical industry groups as strategic investors. In June, ByteDance spun off its AI pharmaceutical business line, establishing Anew Labs as an independent brand, while ByteDance retains majority control. The core team, algorithms, platform, and pipeline have been fully integrated into the new company, with computing power continuing to be supported by Volcano Engine.
The team was formed in 2021, led by Liu Kai, with about 36 to 50 core members based in Shanghai, Singapore, and San Jose. Scientific advisors include Liu Yongjun, former president of Innovent Biologics, Ji Ma, former chief scientist at Amgen, and Hua Zou, director of protein chemistry at Takeda California. The platform has released AnewSampling, AnewOmni, AnewFEP, AnewSynth, and scNext, covering dynamic structures of protein ligands, all-atom molecular generation, free energy calculations, synthetic feasibility, and virtual cells; structural prediction tools include Protenix, Seedfold, and the protein binding design tool PXDesign.
In April, the IL-17 small molecule project was publicly presented for the first time at the American Immunology Society annual meeting, claiming that a single small molecule can block three dimers (AA, AF, FF) of this family, targeting autoimmune pathways such as psoriasis and ankylosing spondylitis. Antibody drugs have validated the clinical value of simultaneously inhibiting A/F, while the small molecule tri-blocking is still in preclinical lead optimization; the IL-4 receptor project is at the lead stage, with other undisclosed targets still in the discovery phase.
The purpose of the spin-off is to enable independent decision-making, access industrial capital, recruit talent at the pace of a pharmaceutical company, while keeping the large company's computing power internally. The pipeline has not yet entered human trials; the $290 million purchase is for the platform and early assets, not for a listed drug. Strategic pharmaceutical companies investing are reserving interfaces for future joint ventures and equity.
In market mechanics, buyers are funds holding scientific assets in RMB and USD outside the large company's pipeline, while sellers are the controlling shareholders connecting generative models with the drug pipeline from the same source as the recommendation algorithms. This is industrial financing, not a merger exit. Funds are entering preclinical chemistry from profits of consumer internet and venture capital; beneficiaries are ByteDance, which remains consolidated, and industrial shareholders who gain independent director seats, while the pressured parties are independent AI pharmaceutical companies at the same stage without Volcano Engine's computing power, and traditional pharmaceutical companies that must use human data to validate generated molecules within their R&D budgets.
Source: Public Information
ABAB AI Insight
ByteDance has migrated the generative capabilities of its recommendation system to molecular design, and through the spin-off, it has separated the pharmaceutical risks from the narrative of the listed company. TikTok's algorithm predicts what will be watched next, while AnewOmni claims to have trained on five million biomolecular compounds to generate functional molecules across scales. The IL-17 tri-blocking is a milestone for investors, not an endpoint for regulators. The $290 million occurred three months after the spin-off, indicating that the June announcement was aimed at this round of pricing.
The capital path involves large company computing power combined with RMB industrial capital. Volcano Engine internalizes training costs, funds externalize preclinical risks, and pharmaceutical shareholders embed future licensing interfaces. Institutions like HSG are not buying short video cash flows but a pipeline entity that can sign contracts externally. Before human data is available, the valuation anchor relies on "world-first" conference abstracts and open-source model papers, similar to the narratives of Isomorphic and Insilico, but with an additional internet parent company among the shareholders.
Analogous to Alphabet spinning off Verily and Tencent investing in drug Ming, adding a layer of generative structural prediction. The industry is transitioning from issuing pre-training weights to issuing lead compounds. Those who can provide computing power, chemical advisors, and industrial shareholders simultaneously will be able to sell conference posters into Series A funding. The true pricing power remains in Phase I clinical trials: if it fails, the $290 million is merely an extended runway for the model company.
Structural judgments belong to the reconstruction of the industrial chain. The mechanism is that large companies nurture scientific models until they are demonstrable, then package the demonstrations into a financeable independent legal entity, allowing pharmaceutical capital to bid based on pipelines rather than advertising revenue. The spin-off does not change the majority control; what changes is who will pay the next clinical trial bill.
ABAB News · Cognitive Laws
Large companies spin off scientific businesses to allow industrial capital to bid based on pipelines.
Conference posters can complement financing but cannot replace human data.
Keeping computing power in the parent company allows risks to be sold to funds.