TechChora.com Breaking Tech News | Friday, April 24, 2026 | Reporter: TechChora Newsroom
DeepSeek, the Chinese artificial intelligence startup that shocked the global tech industry in January 2025 with an open-source model that matched US competitors at a fraction of the cost, is raising outside capital for the first time and has swiftly escalated its valuation target from $10 billion to more than $20 billion as interest from major investors intensifies. According to reporting by The Information, citing four sources with direct knowledge of the discussions, Chinese tech giants Tencent Holdings and Alibaba Group are both in talks to participate in the funding round, potentially creating a three-way power dynamic in China’s rapidly consolidating AI landscape.
The fundraising comes after DeepSeek spent its first years of existence rejecting outside capital entirely. The company is owned by Zhejiang High-Flyer Asset Management, the quantitative hedge fund founded by Liang Wenfeng, who also established DeepSeek in 2023. High-Flyer’s substantial capital and computing resources allowed DeepSeek to operate as a well-funded internal research project rather than a venture-backed startup chasing growth metrics. That model worked spectacularly from a technical standpoint. The R1 model DeepSeek released in January 2025 achieved performance comparable to OpenAI’s frontier systems despite China’s severely restricted access to cutting-edge Nvidia chips, triggering a brief market panic that wiped hundreds of billions off American AI and semiconductor stocks.
The decision to now open the door to external investors signals a shift in strategic necessity. One of the primary drivers is talent retention. Sources familiar with the discussions told The Information that DeepSeek is losing engineers to rival Chinese AI firms including ByteDance, Baidu, and Xiaomi, who are offering salaries two to three times higher. Without equity compensation tied to a formal valuation, DeepSeek cannot compete. Raising external funding at a defined valuation creates the option pricing framework needed to grant meaningful stock options to engineers being actively recruited by better-funded competitors.
Compute pressure is the second major factor. DeepSeek is in the middle of a profound technical transition, abandoning its previous reliance on Nvidia’s CUDA framework to run entirely on Huawei’s Ascend chips, specifically the Ascend 910C and its successors. This is not a simple hardware swap. It requires rewriting vast amounts of underlying code, from operator libraries to communication libraries, to be compatible with Huawei’s CANN framework. Sources told 36Kr that the transition caused training instability and subpar inter-chip communication speeds, delaying the release of DeepSeek’s planned V4 model from February 2026 until an unconfirmed future date.
The Huawei transition matters enormously beyond DeepSeek. Nvidia CEO Jensen Huang addressed the issue directly on the Dwarkesh Podcast this week, saying it would be ‘a horrible outcome’ for the United States if DeepSeek optimized its next-generation models for Huawei chips. Huang’s concern is that as AI diffuses globally, systems optimized for Chinese hardware standards could give China’s technology ecosystem a structural advantage, particularly in markets across Asia, Africa, and the Middle East that are choosing between American and Chinese AI infrastructure. Huawei’s Ascend 910C currently delivers about 60 percent of the inference performance of Nvidia’s H100, and American chips remain roughly five times more powerful than Chinese alternatives overall. But the gap is closing, and DeepSeek’s willingness to engineer around it rather than work through it is a direct challenge to American chip export control strategy.
The $20 billion valuation target places DeepSeek slightly above the $18 billion that Moonshot AI, the Chinese startup behind the Kimi large language model series, is seeking in its own current funding round. The comparison reflects a broader competition within China’s AI industry, where dozens of well-funded startups are racing to establish durable positions before the market consolidates. ByteDance, Baidu, Alibaba, Tencent, and Huawei all have their own frontier model programs, and the competitive dynamics are intensifying rapidly.
For Tencent and Alibaba, investing in DeepSeek is not purely a financial play. It is a strategic positioning decision. DeepSeek’s open-source technology model means its research outputs are theoretically available to everyone, but having an equity stake gives investors early access to research directions, potential preferential deployment agreements, and influence over how the company develops. Tencent’s Hunyuan model program and Alibaba’s Qwen series both compete in the same space where DeepSeek operates, creating an unusual situation where potential investors are also direct competitors.
DeepSeek’s focus on open-source development has created additional complexity around how it should be valued. Unlike closed-source AI companies that can point to subscription revenue, API pricing, and enterprise contracts as revenue metrics, DeepSeek has prioritized releasing its models freely to maximize adoption and influence over the global AI ecosystem. Investors in the current round must therefore bet on DeepSeek’s future ability to monetize its technology position, whether through enterprise services, infrastructure licensing, government contracts, or some other model that has not yet been publicly articulated.
The talks remain in flux. No deal has been confirmed, and the final valuation and investor composition could shift significantly before any announcement. But the mere fact that DeepSeek has opened its doors after years of self-sufficiency, combined with the rapid escalation of its valuation target as interest builds, suggests that the company sees an urgent window to secure the resources needed to maintain its competitive position in a race that is accelerating every quarter.
