ByteDance's Volcano Engine Denies IPO Plans, Unveils Doubao 2.1 as Token Usage Hits 180 Trillion Daily

Tech News

Volcano Engine, the cloud services arm of ByteDance, has no plans to pursue a separate initial public offering, according to company president Tan Dai. Speaking at a media session during the 2026 FORCE conference on June 23, Tan addressed growing speculation that the unit might be spun off amid soaring valuations across the AI sector. “As far as I know, Volcano Engine currently has no plans for a separate spin-off IPO,” Tan said, adding that ByteDance remains firmly focused on advancing its core AI technologies — including the Doubao large language model, the Seedance video generation platform, and enterprise AI-native architecture deployment.

ByteDance has long been viewed as one of the most IPO-ready Chinese tech giants, but the company has consistently shrugged off market expectations. Over the past year, it has exited several non-core businesses — including gaming — to concentrate resources on AI and its flagship social media platforms. Despite recurring rumors of an imminent listing, ByteDance has repeatedly issued official denials, signaling no urgency to tap public markets.

Volcano Engine functions as ByteDance’s outward-facing technology platform, packaging the growth methodologies, infrastructure, and tooling developed internally and offering them to external enterprises. Its portfolio spans cloud infrastructure, video and content delivery, the VeDI data intelligence platform, AI services, and DevOps tooling — all aimed at helping businesses navigate digital transformation.

Doubao Reaches 180 Trillion Daily Tokens

During the same event, Tan revealed striking growth figures for Doubao, ByteDance’s flagship large language model. As of June 2026, the model processes an average of 180 trillion tokens per day — a staggering 1,500-fold increase since its initial launch, and more than a tenfold jump over the past twelve months alone. The numbers underscore the breakneck pace at which enterprises and developers across China are embedding LLM capabilities into their products and workflows.

Doubao 2.1 Series: Pro and Turbo

Fresh off those metrics, Volcano Engine officially unveiled the Doubao 2.1 model series, headlined by two variants designed for different deployment scenarios:

  • Doubao 2.1 Pro — the flagship model, priced at ¥6 per million input tokens and ¥30 per million output tokens. With cache hits, the input price drops to just ¥1.2 per million tokens, making it highly economical for applications with repetitive context patterns.

  • Doubao 2.1 Turbo — positioned as the cost-optimized alternative for large-scale inference workloads. It delivers comparable capability to the Pro model at roughly half the price, targeting high-volume production environments where marginal cost matters most.

The dual-tier strategy mirrors moves by other global AI labs — offering a full-fidelity model alongside a distilled or efficiency-tuned variant — and positions Volcano Engine to compete aggressively on both capability and price in China’s increasingly crowded LLM market.

With IPO speculation set aside for now, the message from ByteDance is unmistakable: the company is playing a long game in AI infrastructure, betting that developer adoption and real-world deployment — not financial engineering — will drive the next chapter of its growth.