DeepSeek is a Chinese artificial intelligence company based in Hangzhou that develops open-weight large language models, spun off in 2023 from the quantitative hedge fund High-Flyer. It became globally known in January 2025 when its R1 reasoning model matched top Western AI systems at a fraction of the reported training cost, triggering a massive selloff in US tech stocks. This page updates automatically every day: the timeline grows as news breaks, and the analysis refreshes with it.
AIAuraFarm
Start Aura Farming
Top AI money moves delivered every morning - free forever.
Auto-updated twice a week · Last updated 2026-08-12
📇 Company Profile
Founded2023
HeadquartersHangzhou, Zhejiang, China
LeaderLiang Wenfeng
Funding / valuationSecond funding round resumed after brief suspension, targeting roughly $8 billion at a $74 billion valuation, up from the previously reported $60 billion figure.
Flagship modelsD, e, e, p, S, e, e, k, -, V, 4, , F, l, a, s, h, , n, o, w, , h, a, s, , a, , p, u, b, l, i, c, , b, e, t, a, , A, P, I, , a, l, o, n, g, s, i, d, e, , V, 4, , P, r, o, ,, , w, i, t, h, , a, n, , a, g, g, r, e, s, s, i, v, e, , u, l, t, r, a, -, l, o, w, , p, r, i, c, i, n, g, , c, o, d, i, n, g, , m, o, d, e, l, , r, e, l, e, a, s, e, d, , A, u, g, u, s, t, , 1, .
🌾 Aura's Take (AI summary, refreshed daily): DeepSeek's story right now is a startup outgrowing its own playbook. The firm that built its brand on rock-bottom pricing is now walking that back, warning of a 'significant' price increase just as demand for its ultra-cheap coding model spikes. That is not weakness, it is leverage: usage is clearly outstripping capacity, and Hangzhou knows it can charge more without losing share from US labs still burning cash on compute. Meanwhile the reopened roughly $8 billion funding round, reportedly pushing valuation toward $74 billion with Monolith Management circling, suggests investors are betting DeepSeek can monetize its efficiency edge rather than just chase benchmarks. The V4 Flash beta and gigawatt-scale data center ambitions show a company pivoting from scrappy disruptor to infrastructure builder. The risk is execution: repeated V4 GA delays and founder-comment leaks that reportedly stalled fundraising once already hint at internal friction. Money follows momentum here, but DeepSeek needs to prove the price hikes stick without customers bolting to rivals.
This timeline grows automatically as news breaks. 26 events on record.
2026-08-06DeepSeek announced it will implement a 'significant' price increase across its AI API services without specifying the size or timing of the hike. [source]
2026-08-06DeepSeek reportedly resumed its second funding round, seeking to raise roughly $8 billion at a $74 billion valuation, according to Bloomberg. [source]
2026-08-01Fortune reported that DeepSeek, now valued around $60 billion with roughly 173 million downloads since its January 2025 launch, maintains a no-KPI, no-overtime culture according to founder Liang Wenfeng's internal staff discussion.
2026-08-01DeepSeek released a new ultra-cheap coding model that charges pennies for large volumes of code, intensifying the industry-wide race to commoditize AI intelligence. [source]
2026-07-31DeepSeek rolled out a public beta API for its flagship V4 Flash model, touting improved agentic capabilities. [source]
2026-07-30Bloomberg reported DeepSeek is developing a massive, gigawatt-scale AI data center in Inner Mongolia, a major infrastructure push despite lingering US chip export controls. [source]
2026-07-25The funding pause was reportedly driven partly by founder Liang Wenfeng's frustration over viral online reports about comments he made to investors, and management told backers it will prioritize AGI research over near-term commercialization. [source]
2026-07-25DeepSeek told prospective investors it is suspending its second fundraising round (targeting roughly $7.4 billion at a $74 billion valuation) for now, with the company saying it may resume the process later, Bloomberg reported. [source]
2026-07-20A quant fund managed by DeepSeek founder Liang Wenfeng's investment firm High-Flyer slumped roughly 15.7% in the week ended July 17 as a broader selloff in chip stocks hit Chinese quantitative hedge funds.
2026-07-16Reports indicated DeepSeek's new funding push has grown to a proposed $7.4 billion raise at a contemplated $74 billion valuation, with its planned IPO now specifically targeted at Shanghai's STAR Market as early as 2027.
2026-07-16Security researchers at Hunt.io revealed a Chinese state-linked cyber espionage campaign that used DeepSeek-V4-Pro as the reasoning engine alongside Claude Code to automate hacking, phishing-page cloning, and exploit development across multiple countries.
2026-07-14Bloomberg reported DeepSeek founder Liang Wenfeng's net worth more than doubled to $36 billion following the company's recent funding round, making him the world's wealthiest AI model founder, ahead of Anthropic's Dario Amodei and OpenAI's Greg Brockman.
2026-07-14DeepSeek has begun preparing for a mainland China IPO, targeting a filing as soon as this year with a debut possibly in 2027. [source]
2026-07-14DeepSeek is reportedly in talks to raise about $1.5 billion in new funding at a roughly $71 billion valuation, just weeks after closing its first $7 billion outside round. [source]
2026-07-07Reuters reported DeepSeek is developing its own inference-focused AI chip to reduce reliance on Nvidia and Huawei hardware, with the effort reportedly underway for about a year.
2026-06-30DeepSeek announced the official (non-preview) launch of V4 for mid-July 2026, adding a new peak/off-peak API pricing scheme charging double rates during Beijing business hours.
2026-06-16DeepSeek closes its first-ever external funding round, raising over $7.4 billion from Tencent, CATL and China's state AI fund at a $50B+ valuation.
2026-04-24DeepSeek previews its V4 Pro and V4 Flash models with 1-million-token context windows, built on Huawei Ascend chips, claiming near-parity with GPT-5.4 on some benchmarks.
2026-02-01Anthropic accuses DeepSeek of using thousands of fraudulent accounts to mine Claude conversations for training data.
2025-05-01DeepSeek releases the R1-0528 update, improving math and code reasoning.
2025-01-27DeepSeek's app surpasses ChatGPT to become the #1 free app on the US iOS App Store.
2025-01-20DeepSeek releases the R1 reasoning model and chatbot, matching OpenAI's o1 on benchmarks and sparking a $1 trillion US tech stock selloff, dubbed AI's 'Sputnik moment.'
2024-12-01DeepSeek releases V3, trained for a reported $5.6 million, drawing attention for its efficiency versus Western frontier models.
2024-06-01DeepSeek-Coder-V2 launches on the V2 mixture-of-experts backbone, supporting 338 programming languages.
2023-11-01DeepSeek ships its first public model, DeepSeek Coder, focused on code generation.
2023-07-17DeepSeek is spun off from hedge fund High-Flyer as an independent AI company founded by Liang Wenfeng.
Frequently Asked Questions
Who owns DeepSeek?
DeepSeek is majority owned and was solely funded by founder Liang Wenfeng, who also co-founded the hedge fund High-Flyer that backed it. After its June 2026 funding round, Liang still controls the company through a limited partnership, while Tencent, CATL and China's state AI fund hold minority stakes without voting rights except the state fund.
Is DeepSeek's AI free to use?
Yes, DeepSeek's chatbot is free to use on web and mobile, and its models (including V4 Pro and V4 Flash) are released as open weights on Hugging Face under an MIT license, letting developers download and modify them for local or commercial use.
How did DeepSeek disrupt the AI industry?
In January 2025, DeepSeek's R1 model delivered performance comparable to OpenAI's o1 while reportedly costing far less to train, which rattled markets and erased roughly $600 billion from Nvidia's market value in a single day. This challenged assumptions that frontier AI required massive US-style capital expenditure.
What chips does DeepSeek use for training?
DeepSeek historically trained on Nvidia H800 chips before US export restrictions tightened, and its 2026 V4 models were built with support from Huawei's Ascend 950 processors and Cambricon chips. This shift reflects Beijing's push to reduce reliance on Nvidia hardware.
How does DeepSeek V4 compare to GPT and Gemini?
DeepSeek claims V4 Pro rivals or beats GPT-5.2 and Gemini 3.0 Pro on some reasoning tasks and matches GPT-5.4 on coding benchmarks, though it lags top frontier models by an estimated 3 to 8 months on knowledge tests. Its main edge is price, with output tokens costing dramatically less than Western competitors.
