NextFin News — Shou Zi Chew stepped onto a Singapore stage on October 12 and talked about science. The TikTok chief executive, appearing at the DBS Asia Future Forward Conference, left short video, commerce and politics aside. He spoke instead about physics, biotechnology, chemistry and materials science.
Artificial intelligence, he said, is likely to produce genuine scientific breakthroughs in those fields within two to three years. The prediction was deliberate. It tracks investments ByteDance has already made far from its consumer apps.
In June the company separated its internal AI drug-discovery group into a standalone firm, Anew Labs. About fifty researchers, their software platforms and early pipeline assets moved into the new entity, with offices in Shanghai, Singapore and California. By September Anew Labs had raised $290 million at a $1.5 billion valuation. Leading Chinese venture firms backed the round; ByteDance kept a controlling stake of roughly 56 percent.
The split solves a tempo problem. Internet products reward rapid iteration. Drug programs run on multi-year evidence cycles and nine-figure budgets. Housing both inside one management system distorts incentives. An independent vehicle can raise specialist capital, hire pharmaceutical talent and absorb clinical risk without forcing the parent’s consumer businesses to share the same clock.
Modeling capacity sits underneath the pipeline. ByteDance’s Seed research group released Protenix-v2 in April, a system for predicting and designing biomolecular structures. Public benchmarks show clear gains on difficult antibody–antigen interfaces and large efficiency improvements in sampling. The project’s relatively open distribution aims to seed adoption among academic labs and smaller drug companies—an attempt to occupy the practical interface in a field still shaped by closed frontier systems.
Money is available at scale. In September ByteDance arranged a multi-year bank facility of nearly $30 billion for AI projects outside China, including data-center construction in Southeast Asia. Lenders extended the facility without collateral, a vote of confidence in the group’s cash generation. How much will ultimately support scientific work remains undisclosed. The pattern of spin-out financing, continued model investment and overseas infrastructure points to rising, not falling, priority for AI-for-science.
Discovery speed is only half the constraint. AI can compress target identification and candidate design from years into months. Phase III trials still require dozens of sites, hundreds of patients and long observation windows. Hypotheses multiply; wet-lab throughput and regulatory endpoints do not automatically keep pace. Researchers and executives across the field note the shortage of people fluent in both computational methods and experimental medicine, and they insist that an AI origin story is not clinical proof.
Other technology companies are mixing platform tools with proprietary pipelines, pairing with large pharma, or funding shared biological-simulation infrastructure. ByteDance has so far emphasized owned models, a controlled spin-out, external financing and selected therapeutic programs under its own roof.
Chew’s public comments stayed measured. He warned against systems that optimize a single objective past human intent, argued for human oversight in high-stakes settings, and acknowledged that fully generated video still draws limited engagement on TikTok even as recommendation algorithms dominate the product. The scientific outlook was optimistic; the governance language was careful.
The corporate record supplies the concrete claim he did not need to make. ByteDance has already committed people, models and outside capital to the idea that AI can change the economics of discovery. The next few years will show whether those commitments produce validated outcomes—or simply longer queues at the laboratory door.
更多精彩内容，关注钛媒体微信号（ID：taimeiti），或者下载钛媒体App