AI inference startup Qujing Technology (趋境科技) just closed a Series A round, pushing its total funding past 1 billion yuan (roughly $140 million) in under six months. Announcement came with some eye-popping numbers: the company says its “Token factory” is already churning out trillions of tokens daily, and parts of the business are profitable.
The round was led by Henan Investment Group’s Huirong Fund, with existing backers Zhenzhi Capital, Shangshi Capital, and Xinglian Capital all following on. The exact amount of this single tranche wasn’t disclosed.
Qujing doesn’t train its own large language models, nor does it try to serve every model under the sun. Instead, it focuses on optimizing inference for a handful of top-tier models — squeezing more tokens out of the same GPU, faster and more reliably.
A “Token factory,” as the company describes it, is essentially a coordinated system of chips, memory, networking, and inference software. The goal is to cut time-to-first-token, boost concurrency, and minimize the waste from idle hardware or failures — all at once.
Since Chinese New Year 2026, Qujing claims per-chip token throughput has more than tripled, and overall capacity has jumped over 30x. That’s the kind of efficiency that turns heads — and wallets.
The fresh capital will go toward expanding token production capacity, upgrading the company’s ATaaS (AI Token as a Service) platform, and pushing domestically produced chips into large-scale inference scenarios.