SiliconFlow adds a C round on the way to its Hong Kong listing

Share
SiliconFlow adds a C round on the way to its Hong Kong listing

SiliconFlow closed new money on Sunday, and the interesting part is who signed: the same state system that buys its tokens is now taking equity in the meter.

SiliconFlow has completed a second tranche of its B+ round and a C round, bringing its 2026 equity funding to nearly 2.9 billion yuan. The investor list reads like a state roll-call: China Internet Investment Fund, China Reform Fund, China Mobile's supply-chain fund, China Orient Asset International, Beijing Energy's fund, Shanghai Instrument Electric's smart-compute fund, Beijing High-Tech Venture Capital, plus provincial vehicles from Chengdu, Jiangxi and Shandong's rail investment arm. Existing backers Yaotu Capital, Shengyi Capital and Guotai Venture Group followed on. The company says the money goes to inference-engine work, heterogeneous compute scheduling, model-to-chip adaptation, its token supply platform and overseas expansion — the plumbing that turns rented GPUs into metered API calls, not the models themselves.

That positioning is the point. SiliconFlow does not train frontier models; it resells other people's open-weight ones, which makes it the layer where China's compute subsidy, model licensing and enterprise demand all meet. Frost & Sullivan ranks it as the country's largest independent-ecosystem token supplier by 2025 throughput and inside the domestic top five overall. Its own disclosures put registered users past 10 million and paying enterprise customers above 13,000, with the platform adapting models across Nvidia, AMD, Huawei's Ascend, Cambricon and Moore Threads silicon — the multi-vendor hedge that Chinese buyers care about more than benchmark scores.

The economics underneath are less flattering, and the company has already told the market so. SiliconFlow filed for a Hong Kong main-board listing on June 30, with Huatai International and Guotai Haitong as joint sponsors. The prospectus shows 2025 revenue of 55.3 million yuan, up 653 percent from 7.3 million, while the gross line flipped from a 2.9 million yuan profit to a 13.3 million yuan gross loss and the net loss widened 321.8 percent to 345 million yuan. Compute resource costs reached 59.6 million yuan, roughly 21 times the prior year, and sales and marketing hit 68.6 million yuan, up about 14 times. Public cloud — 52.9 percent of revenue — is sold below cost: the company spends about 1.19 yuan for every yuan it books there. Only on-premise deployments, 47.1 percent of revenue at an 82.5 percent gross margin, actually make money. This is the arithmetic we laid out last week — Deep Dive — Tokens halved in price. The bill to make them didn't — arriving on a prospectus page.

Ownership matters here too. Alibaba is the largest external shareholder at 7.42 percent. Founder Yuan Jinhui, whose earlier company OneFlow was absorbed into Meituan after being acquired by Lightyear, holds 14.35 percent directly and controls about 44.5 percent of votes with four employee incentive platforms. Founded in August 2023, the company has gone from nothing to a 2.9-billion-yuan funding year and a listing file in under three years.

The read: state money arriving pre-IPO in the token-metering layer is a hedge against the pricing war, not a bet on it. When Beijing's own funds are the customers, the shareholders and the subsidy channel at once, unit economics can stay negative longer than a private competitor could survive — which is exactly why the state is comfortable buying in, and why the private inference market should be nervous. This sits alongside the direct state stakes we covered this month — China's state AI fund just took a direct stake in Kling — and the consumption logs the Jiangsu fund published, where a Chinese state fund ran 6.07 billion tokens through its deal desk.

What to watch: whether the C round is explicitly framed as pre-IPO, and whether SiliconFlow's on-premise line keeps carrying the company as public-cloud prices keep falling.

If the state is simultaneously the buyer, the shareholder and the subsidy, is a token utility ever supposed to be profitable? Tell us in the comments.

Read more

Byteification turns Qwen3 and Llama 3 into byte-level models

Byteification turns Qwen3 and Llama 3 into byte-level models

The tokenizer is the last piece of the language-model pipeline that was never learned end to end — and a Nature paper out this week shows it can be swapped out for less than one percent of a pretraining budget. Researchers have retrofitted four open subword language models into true byte-level models using less than 1% of a typical pretraining budget, and the result — published 7 October in Nature — closes the performance gap that has kept byte-level models a niche curiosity for nearly a decad

OpenAI's dots go mobile and can start your Codex work

OpenAI's dots go mobile and can start your Codex work

One product update and two benchmark papers this hour, all circling the same question: how much of the work can you actually hand over, and what happens to the part you didn't? OpenAI's dots — the always-on agents that get their own cloud computer — can now be created and configured entirely from the ChatGPT app on iOS and Android, and a dot can start Codex work or pick up an existing Codex thread on your behalf. The October 9 release, billed as the first of a fresh run of updates for dots, mo

Coding agents as a hospital: Cockroach's five-month results

Coding agents as a hospital: Cockroach's five-month results

The agentic-coding debate keeps being framed as speed versus trust. Cockroach Labs just published the most detailed trust-first ledger yet — five months of numbers, including the parts that didn't work. Cockroach Labs ran its codebase like a teaching hospital for five months and says the pipeline merged well over a million lines of code with seven total reverts. In the setup the company calls MOLT Sinai, GitHub issues are patients, merging is discharge, and the humans in charge are Chiefs of M

Today in AI — October 10, 2026

Today in AI — October 10, 2026

A day where the boring machinery took center stage: proof checkers, order books, warehouses and the humble text message — plus fresh arguments about who gets to declare any of it safe. Models & Research * Mathematicians get a reliability primer for the Lean Theorem Prover. A guest post by Thomas Hales on Terence Tao's blog walks through what Lean actually guarantees — and devotes itself to the "Summer of Soundness Bugs," the string of kernel bugs found this summer that let false proofs thr