DeepSeek is once again in the process of raising funds, with plans to go public.

According to reports from foreign media, DeepSeek is contemplating a new round of financing and preparing for an initial public offering (IPO). It intends to submit its listing application as early as the end of this year, with the goal of listing on the Shanghai Stock Exchange next year.

In the previous financing round, which raised 50 billion RMB, founder Liang Wenfeng, Tencent, and CATL collectively contributed 35 billion RMB, leaving limited space for external investors.

Now, with the listing on the horizon, the new 10 billion RMB financing round offers investors who missed out previously a chance to invest on the eve of the IPO, while also providing a valuation benchmark for employees holding stock options.

Going public also implies that Liang Wenfeng and the core team may need to make subtle adjustments and compromises, incorporating more pragmatic considerations regarding business and commercial models.

Whether it's monetizing top-tier coding capabilities through token sales or actively venturing into high-frequency agent markets like office assistants, DeepSeek may need to strike a more balanced path between its ideals and commercial realities.

Why Raise Another 10 Billion?

Many people's initial reaction to the news of DeepSeek's new financing is likely one of surprise.

After all, it's been less than two months since the last round of financing, which raised 50 billion RMB. From the perspective of computational costs and daily operations alone, the timing and scale of this new financing seem somewhat unusual.

Let's first consider the macro context: the new pathway recently opened by the Shanghai Stock Exchange for hard-tech companies. In the past, China's domestic capital markets heavily emphasized profitability, causing many excellent tech companies to turn to U.S. or Hong Kong stock markets during their loss-making phases.

Now, the rules have changed. The Shanghai Stock Exchange has introduced new regulations stating that as long as your technology is sufficiently advanced, you can still list domestically even if you're currently unprofitable.

The rationale behind this new policy is clear: on one hand, it provides a stable financing channel for domestic tech leaders; on the other hand, it gives domestic investors a chance to more directly share in the dividends of top-tier technological development.

In recent years, as Chinese concept stocks have gradually listed in Hong Kong, access has become relatively convenient through programs like Stock Connect. However, investing in purely U.S.-listed companies like Pinduoduo remains relatively difficult for many investors.

Of course, the large model industry is extremely capital-intensive and currently shows little profitability.

DeepSeek has recently expanded its workforce significantly, at least doubling its staff. Competitive salaries are now required to attract top algorithm and engineering talent from the market.

Moreover, DeepSeek is reportedly considering developing its own chips and building large-scale data centers, all of which require substantial hardware investments. Therefore, from a long-term perspective, they indeed need a continuous supply of funds.

However, these factors may not be directly correlated with this round's 10 billion RMB financing.

In the previous round, they raised 50 billion RMB just under two months ago. While 10 billion RMB sounds like a lot, it's just a fraction compared to the last round.

If their spending rate is manageable, the 50 billion RMB should sustain DeepSeek for some time, eliminating the need for such a hasty additional financing round in such a short period. If they were truly burning through cash, they wouldn't settle for just 10 billion RMB.

The key lies in the financing structure of the previous 50 billion RMB round.

In that round, founder Liang Wenfeng himself contributed 20 billion RMB, while Tencent and CATL took 15 billion RMB. The remaining amount was divided among a few large institutions, leaving little room for others.

This actually highlights a point: while DeepSeek needs funds, they are highly selective about external capital and maintain tight control over their equity structure.

Therefore, the need for cash should be a secondary factor in this new 10 billion RMB financing. The other two purposes are still market valuation anchoring and boosting employee confidence.

DeepSeek now has a large number of technical experts. For these top talents, the value of their stock options cannot rely solely on the boss's promises; they need external market pricing in real money.

This external investment at a valuation of 71 billion USD provides clear value endorsement for employees' stock options. Look, despite only two months passing, the valuation has surged again.

Another purpose, when viewed in conjunction with their preparations for a domestic listing, is that this 10 billion RMB additional financing resembles a pre-IPO opportunity to 'buy in.' In the previous round, many resourceful top-tier institutions couldn't secure a spot due to limited capacity. Now, with DeepSeek's listing imminent, releasing some capacity offers these missed institutions a chance to get on board.

This 10 billion RMB influx essentially forms the future roster of cornerstone investors for the listing.

As everyone knows, bringing these stakeholders on board before the listing plays a crucial role in the subsequent IPO process and post-listing stock price stability.

Reconciling Idealism with the Secondary Market

To this day, DeepSeek remains a unique player in the large model race.

DeepSeek always reminds people of early-stage OpenAI—a group of pure tech geeks, obsessed with the ultimate ideal of achieving Artificial General Intelligence (AGI) while seemingly indifferent to immediate commercial interests.

While other major companies have spent billions on marketing and aggressive user acquisition across various channels to gain consumer users, DeepSeek hasn't deliberately pursued this. Instead, it has naturally attracted users through its hardcore technical reputation.

For a long time, DeepSeek's product form has been text-based interaction. Even adding multimodal functions like image understanding was a late move compared to others.

The product lacks overwhelming promotional efforts. When the industry is focusing on agent applications and office automation, DeepSeek remains unswayed, still dedicating all its computational power and energy to improving foundational model capabilities and adapting to domestic computational resources.

However, this geek's paradise-like state will inevitably face pressure as the listing approaches.

Financing essentially means surrendering some freedom in exchange for staying in the game, while going public takes this a step further.

Private financing in the primary market can still rely on the founder's charisma and technological vision. Once truly listed, the company will face all-encompassing and extremely rigorous business scrutiny.

The secondary market prioritizes realistic financial returns, with performance data ultimately reflected coldly in stock price fluctuations and transmitted layer by layer to management.

Few companies can excel at resisting external pressure from the capital markets; Pinduoduo serves as a benchmark in this regard.

During Pinduoduo's earnings calls, management often starts by tempering expectations from institutional analysts and shareholders. Even when presenting impressive profit growth and stellar data, they repeatedly emphasize that such high growth is unsustainable, actively cooling down the market.

They seem to possess strong strategic resolve, refusing to be swayed by short-term market sentiment.

However, Pinduoduo's approach is more of a communication strategy and attitude toward the capital markets.

They fully understand what shareholders want to hear and know which statements can boost the stock price but choose to deliberately remain restrained.

Looking back at Pinduoduo's every major strategic move, whether it's investing heavily in Temu's global expansion, launching hundred-billion-yuan subsidies and trillion-yuan support programs, or stockpiling hundreds of billions of USD in cash reserves, these decisions are all commercially logical and self-contained: Either they're trading current investments for larger future scale benefits, or they're stockpiling resources to prepare for macroeconomic cycles or new projects, still within the framework of traditional business common sense.

However, DeepSeek's behavioral trajectory thus far defies this mature commercial model. To this day, it's still difficult to discern any clear, systematic commercial monetization considerations from them.

This idealism is precious and admirable but objectively lacks commercial sustainability.

Now that they've chosen to sprint toward listing, this likely means that Liang Wenfeng and the core team have subtly compromised and shifted their stance. From here on, DeepSeek will inevitably become more proactive in commercialization.

Currently, two directions lie ahead of them.

One is to follow Anthropic's route and directly convert top-tier coding capabilities into scalable B-end revenue. Coding remains the only segment in the large model field where a commercial closed loop has been successfully achieved.

DeepSeek possesses first-rate technical capabilities, which are its core moat. However, compared to vendors like Zhipu, which have meticulously cultivated this area, DeepSeek has previously put insufficient effort—or interest—into 'packaging and selling coding capabilities.'

The other direction is to directly enter agent scenes like office assistants. With improvements in foundational model capabilities, agent implementation will be the biggest industry disruption in the coming year.

For example, Tencent's heavily promoted Workbuddy now has over 13 million daily active users, performing impressively in actual office scenarios.

Alibaba recently overhauled its operations by merging three core products—Wukong, MuleRun, and QoderWork—under DingTalk's new leader, Chen Yusen, clearly aiming to concentrate forces and break through in this sector.

If DeepSeek can maintain its model leadership while riding the trend and entering such highly commercial application scenarios, its post-listing valuation narrative will become even more compelling.