GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

DeepSeek Revenue Nears $500M a Year as Founder Controls a 2027 Shanghai IPO

DeepSeek’s annualized revenue has reached $400-500 million, according to The Information, roughly doubling its 2025 run rate. The Chinese AI lab retains more than 50% gross profit on V4 API access — margins that rival the best software businesses and underpin plans for a STAR Market IPO as early as 2027.

The figures arrive after DeepSeek closed a $7.4 billion fundraise. Founder Liang Wenfeng contributed $3 billion of his own capital. Despite that scale of outside investment, most external investors hold no voting rights and face a five-year lockup. Wenfeng controls all material business decisions.

The IPO structure

A second fundraising round is in preparation, explicitly targeting dollar-denominated capital from outside the US — primarily Middle Eastern sovereign and institutional investors. By excluding US investors, DeepSeek avoids regulatory scrutiny that could complicate a STAR Market listing while preserving the founder-controlled structure.

Investment banks are already gearing up for the listing. At a $71 billion implied valuation based on current reporting, a 2027 IPO would rank among China’s largest tech debuts in years.

Why the margins hold

The reported 50%+ gross profit on V4 access is consistent with DeepSeek running inference on Huawei Ascend 910B hardware rather than Nvidia silicon. That hardware independence gives DeepSeek a structural cost advantage that US export controls have not closed — Nvidia chips are restricted, but the inference stack is already domestic.

DeepSeek V4 has become the benchmark others are priced against. On OpenRouter, V4 is the platform’s most-used model by token share. V4 Pro, released under MIT license, commands significant API volume even at $0.87/M output after the permanent June price cut. Volume at margin is a sustainable business; the IPO thesis is that it compounds.

What changes at $500M ARR

At this revenue level, DeepSeek is no longer a research lab that happens to charge for API access. It is a commercial AI business that has chosen to keep most of its architecture open. That combination — open weights, closed cap table, STAR Market exit — has no direct precedent among frontier AI labs.

OpenAI is targeting a 2025-2026 IPO on Nasdaq. Anthropic has filed a confidential S-1 at a $965 billion valuation. DeepSeek would list at a fraction of those valuations and far higher margins. Whether public markets will see that as a discount or a different category entirely depends on how they price sovereign AI risk.