Around May 6, 2026, a report from the UK's Financial Times and Reuters simultaneously drew the attention of the Chinese semiconductor and large model circles to the same name: the National Integrated Circuit Industry Investment Fund of China—this fund, long betting on companies like SMIC and Yangtze Memory Technologies, which has almost become synonymous with "foundries" and "equipment manufacturers," is negotiating with the cutting-edge AI laboratory/company DeepSeek, planning to act as a lead investor in its first round of financing. Several Chinese tech and crypto media quickly picked up this news, elevating DeepSeek, which had originally only circulated within technical circles, to the height of a potential target for "national capital votes."

According to public reports, the potential transaction has given DeepSeek an estimated valuation close to or around $45 billion, but negotiations are still ongoing, and the valuation and specific terms have not been finalized. If this first round of financing ultimately comes to fruition, its symbolic significance may even exceed the financing amount itself: a fund with a core label of "chip manufacturing and equipment," regarded as China's largest state-level semiconductor investment force, has for the first time been clearly placed at the top of the financing terms list for large model companies. Regardless of the outcome, this has already released a sufficiently clear signal—state capital is trying to cross the moat from single bets on computing power to the very center of the AI application layer.

The First Round Approaching $45 Billion

For any AI company, being discussed with an estimated valuation of around $45 billion in the first round of external financing is among the rarest cases. The UK's Financial Times directly titled "DeepSeek nears $45bn valuation as China’s ‘Big Fund’ leads investment talks," linking together "big fund leading, first round financing, $45 billion": on one side is a state-level fund that has long bet on chip manufacturing and equipment, and on the other side is a cutting-edge large model company that has just stepped onto the financing stage, with the price tag lifted to a height sufficient to rewrite the narrative.

Putting this figure back into the global AI startup landscape, it is not an easily brushed aside valuation range. $45 billion means that DeepSeek, while still in the first round of financing, has been pushed directly to the front row of the world AI company valuation sequence, which will naturally form a new anchor for Chinese AI entrepreneurs' psychological expectations: it turns out that domestically, as long as the technology and story are persuasive enough, the valuation imaginative space can also no longer passively benchmark overseas but can use a "first round exceeding $10 billion" price range to re-measure the capital market's scale for large model companies.

But it must also be noted that both the Financial Times and Reuters emphasized one premise: this transaction is not yet complete, and the valuation and investment terms are still under negotiation. The $45 billion is currently more like an "anchor price" on the negotiation table, and the final figure may still be adjusted based on factors such as investment scale and responsibilities arrangement. For DeepSeek and the big fund, this round of game is not just about whether they can secure $45 billion; it is also about finding a balance point that both sides can withstand and can explain externally among price, control rights, and long-term expectations.

The Big Fund Steps Out of its Comfort Zone

If we were to trace the investment direction of the big fund over the past few years along a timeline, it would resemble a profile of industry extending downstream along "silicon wafers": from SMIC to Yangtze Memory Technologies, then to the landscape built around manufacturing, design, packaging testing, and equipment materials; this largest, state-supported semiconductor investment institution in China has always focused its money on the hardest core end. The names on the public record of investments rarely stray beyond the official track of "integrated circuit industry," more often addressing shortfalls and filling gaps, assembling wafer production capacity, storage capabilities, and key equipment piece by piece, forming a typical "hardware-oriented" path dependency, which is also implicitly recognized by the market as its "chip comfort zone."

Because of this, when the reported negotiation target changed to DeepSeek—a company publicly described as a cutting-edge AI laboratory/company focused on AI software and application layers such as large models—this move seemed particularly abrupt. For the big fund, investing in SMIC and Yangtze Memory Technologies primarily considers production capacity, yield, and equipment depreciation; but when facing DeepSeek, the evaluative criteria change to algorithms, model iterations, and application scenarios, transforming asset forms from factories and machines to parameters and code. This shift can easily be interpreted as a deliberate extension of the chain: no longer standing only at the “power creation” end, but attempting to bind to the downstream “power utilization,” integrating large models, such as power black holes, into the ecosystem it supports while also responding to more macro AI development strategic demands, making chip investments no longer isolated supply projects.

Whether this means that the investment boundary of the big fund is expanding from “hardware first” to “synergy of hardware and software” is still far from a conclusion, but at least in the potential deal with DeepSeek, it has stepped out of the secure zone of pure hardware logic. Once the transaction is finalized, a state-level semiconductor fund appearing on the shareholder list of large model companies will force the market to redefine its role: it is no longer merely a silent investor behind wafer factories and storage manufacturers, but begins to actively intervene in how the demand for computing power is created and amplified. This role displacement itself constitutes a substantive rewrite of the existing investment narrative.

From Computing Power to Models: China's AI Acceleration Bet

If we break down the Chinese AI industry into a vertical chain, the upstream consists of manufacturing links like SMIC and Yangtze Memory Technologies, responsible for producing wafers and storage; the midstream involves various computing infrastructure, stacking chips into racks and data centers to form training clusters; only then does it progress to large model R&D and application services that compress computing power, algorithms, and data into specific products. DeepSeek is publicly described as a cutting-edge AI laboratory/company focused on directions such as large models, naturally positioned at the "model layer" of this chain, determining how computing power is consumed, how data is organized, and how algorithms are iterated, while the big fund has traditionally only appeared at the higher manufacturing end, rarely directly touching this layer.

Once the national-level semiconductor investment platform extends funding to the equity of model companies, it would effectively reconnect the previously relatively fragmented three ends with capital: the expansion of upstream chips and computing clusters will no longer just satisfy abstract "AI demands," but needs to connect with specific model routes and training plans; model companies will also have a major shareholder with hardware discourse power, potentially having greater coordination space regarding computing power pricing, supply stability, and technology route selection. For the domestic market, this binding will change the previous pattern of “computing power, algorithms, and data each fighting their own battles,” pushing large models from being demand holders to becoming central players in power planning and standard negotiations.

Compared to overseas, large model companies like OpenAI have long been integrated into a unified capital structure from chips, cloud computing to application distribution under the joint support of large tech firms and financial capital, where the model layer and computing power, funding have almost grown synchronously. In China, previously more national capital was bet on chips, and market capital was invested in applications, while the medium-large models were dispersed and taken on by internet companies and social capital. Now, the big fund attempts to cross traditional boundaries, directly connecting with model entities like DeepSeek, using state-level funding to accelerate replenishment in the "model engine" of the intermediate segment. The pace of catching up may not necessarily replicate that of overseas, but by inserting a national capital "nail" at the model layer, China's large model track is forming a more integrated industrial trajectory with more local path dependency from manufacturing to algorithms.

Behind State Capital Support: Opportunities and Constraints Coexist

If the big fund ultimately invests in DeepSeek, the first impact will certainly be the amplification effect in terms of funds and resources. As China’s largest state-supported semiconductor investment institution, the big fund has accumulated a complete manufacturing and equipment network through projects like SMIC and Yangtze Memory Technologies over the past few years. Once this “hardware base” is connected with large model companies, it means that DeepSeek will have opportunities to directly link to national-level industrial chains and local policy resources in procurement of computing power, chip selection, and industrial cooperation. More importantly, the policy signals represented by the big fund itself would create a sort of "implicit guarantee" regarding regulatory, industrial planning, and the attitude of upstream and downstream partners, helping DeepSeek occupy a more central position in the domestic AI ecosystem, rather than merely being a passive adapter of hardware as an algorithm supplier.

However, the entry of state capital has never been a one-dimensional "blood transfusion." Historical experience shows that as a state-level capital platform, the big fund is often accompanied by clear policy orientations and compliance requirements, which will reflect higher prudence in corporate governance and business decision-making. For DeepSeek, introducing the big fund means an increased weight of national will at the board level; some high-risk, highly controversial business attempts need not only to find technological viability but also to account for policy boundaries and social impact; external cooperation, particularly with cross-border capital and platforms, may shift from simple commercial matching to simultaneously satisfying multiple constraints of industrial synergy and policy safety. In the context of large model companies pursuing extreme product iteration speed, this "multi-step thinking" decision-making approach serves as both a protective umbrella and a potential brake.

More realistically, as of May 6, 2026, all of this remains uncertain on the negotiation table. The Financial Times and Reuters have only confirmed that the big fund is in negotiations with DeepSeek regarding a potential investment, intending to act as a lead investor in its first round of financing, with an estimated valuation close to or around $45 billion, but they emphasized that the transaction terms have not been finalized, and both sides have not disclosed any official announcements or final structures. Various statements regarding DeepSeek's equity structure—including founder Liang Wenfeng's personal shareholding ratio and whether he will co-invest—come from C-side tweets, marked in briefs as unverified information; speculation about whether Tencent or Alibaba will participate in this round of financing or whether the starting valuation is lower than the current reported figure is similarly categorized as unverified market rumors. Until these key variables are settled, the only thing we can confirm is: the big fund is trying to extend its policy and capital leverage into the large model layer, and the final direction of this negotiation will largely determine whether DeepSeek moves rapidly on the "national capital + industrial chain" track or continues to explore on a looser yet more uncertain market capital path.

A Transaction to Leverage a New Competitive Point for Chinese AI

If one considers that the big fund's bets on SMIC and Yangtze Memory Technologies over the past few years have been paving the foundation for China's computing power and manufacturing base, then the negotiation with DeepSeek represents the first time it has pushed into the “superstructure above this foundation” of large models. Public reports have clearly stated that if this potential transaction proceeds, it will mark DeepSeek's first round of financing led by a national-level fund, which signifies a clear deviation from the big fund's consistent “hardware-oriented” trajectory: shifting from merely being an "infrastructure investor" focused only on foundries, storage, and equipment, to attempting to grasp both the computing base and model layer applications simultaneously in a "co-integration of hardware and software" role. Multiple Chinese and English media have interpreted this role switch as a signal of state capital entering the large model track more directly, rather than just participating indirectly in "laying the groundwork" for computing power and chips.

Once this first round of financing is truly established, it will release to the entire market not just the signal that “a certain company has secured funding,” but also a template for financing paradigms: national-level funds sitting at the forefront of the table signify that subsequent local large model companies, when designing their equity structure and planning their financing pace, can hardly ignore the path of “national capital + industrial chain synergy.” For local industrial funds, industry leaders, and other social capital, a large model investment led by the big fund may also be seen as the starting point of a rhythm—whoever follows first and to what extent will implicitly rearrange the flow map of AI capital in China for the coming years. However, as of May 6, 2026, negotiations have yet to be announced as complete, and uncertainties remain regarding the details of the terms and whether more state-owned or private capital will be introduced, which means that the final announced transaction result will determine whether this investment is merely a rare case or marks the starting gun for China's entry into a new round of state capital layout cycles.

Join our community, let's discuss and become stronger together!
Official Telegram Community: https://t.me/aicoincn
AiCoin Chinese Twitter: https://x.com/AiCoinzh
OKX Welfare Group: https://aicoin.com/link/chat?cid=l61eM4owQ
Binance Welfare Group: https://aicoin.com/link/chat?cid=ynr7d1P6Z

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。