On the evening of May 8th, "White Whale Lab" released an exclusive report stating that negotiations over financing terms between DeepSeek and Alibaba had collapsed, with both sides failing to reach an agreement on specific financing terms. It was claimed that this was due to a conflict between Alibaba's strong desire for a closed-loop ecosystem in its AI strategy and DeepSeek's positioning as an independent model company.

This news has not yet received a response from Alibaba, and given that investment and financing negotiations involve a process of bargaining, everything remains uncertain until a formal financing agreement is signed.

On the afternoon of May 9th, some market insiders also claimed that Alibaba had likely not engaged in negotiations.

However, the terms "market insiders" and "likely" are both very vague, lacking definitive details.

Alibaba and Tencent were among the first tech companies rumored to be participating in DeepSeek's financing. At the time, the valuation was an incredible $20 billion, but it has now risen to $40-50 billion.

The threshold for basic research and development of large models is extremely high, and DeepSeek is clearly an exceptionally rare and high-quality target. Simply participating in this deal is already a testament to one's status and capabilities.

But for Alibaba, whether to invest or not ultimately comes down to its own business logic: What would Alibaba gain by investing? What would it lose by not investing?

What Alibaba Wants

I've seen many comments that are pleased with this news. The general sentiment is that Alibaba's track record in external investments and acquisitions has not been particularly good in recent years. If DeepSeek were to take Alibaba's money, it might be led astray by Alibaba's organizational, ecological, and commercialization logic.

This concern is understandable.

DeepSeek currently has a very unique public image, or more precisely, it carries many expectations that transcend commercial considerations.

It's not the kind of company that wears a suit and talks about closed-loop business models; rather, it's a company with a strong technical idealism.

Judging from past acquisition cases, it's easy to think that keeping DeepSeek away from Alibaba might be a good thing.

But this judgment is a bit lazy. There's a missing link in the logic chain.

Alibaba's external investment capabilities are often compared to those of Tencent.

Relatively speaking, Tencent has indeed been more successful, with examples like Meituan vs. Ele.me, Didi vs. Kuaidi, and Kugou/Kuwo vs. Xiami.

The reason is not complicated. Tencent is the starting point for traffic and social chains. Everyone uses WeChat and QQ, making it naturally easier for Tencent to provide entry points, scenarios, and users for its invested companies.

Alibaba is different. Alibaba is closer to transactions, which are often the endpoint of traffic monetization and the most efficient traffic killers.

So, in the past, many internet products felt different when they took money from Tencent versus Alibaba.

Meituan is the most typical case.

When Wang Xing had to choose between Alibaba and Tencent, Alibaba offered conditions that included providing over a billion yuan but stipulated that Meituan could no longer accept money from Tencent.

Wang Xing didn't agree, and Alibaba ultimately chose to support Ele.me. Meituan was free to be featured in WeChat's grid, while Ele.me struggled to even launch a WeChat mini-program.

But large models are not like past internet products.

Large models are primarily about technological competition. As long as the model's capabilities are sufficiently leading, everything else can be reconfigured.

Traffic can be diverted, channels can be changed, and commercialization paths can be supplemented. None of Alibaba's past unsuccessful investment cases failed because the technology wasn't leading.

So, we can't simply use Alibaba's past investment track record to conclude that "Alibaba investing in DeepSeek would be a bad thing." Just because someone isn't good at singing and rapping doesn't mean they can't play basketball.

Whether Alibaba invests or not ultimately comes down to a business logic.

It needs to see if the investment will yield a return, whether that's financial or ecological, such as in cloud services, API distribution, enterprise clients, or model applications. Ultimately, it all comes back to profit.

DeepSeek, to this day, is not a company overly concerned with commercialization. It's more like a company sitting on a gold mine but temporarily not interested in digging it up.

This doesn't mean it lacks commercial value. On the contrary, DeepSeek's value is too obvious.

Its model capabilities speak for themselves. More importantly, it has a very scarce consumer-facing entry point. Major companies have been spending crazy ly (a typo in the original, meaning " crazy 地" or "frantically") on promoting their AI products over the past year. Doubao, Yuanbao, and Qianwen have all been aggressive. Yet, DeepSeek has achieved significant user mindshare without any promotion.

A company with a model, open-source reputation, and user entry point is truly a rare find.

But for Alibaba, the calculation can't just be about sentiment.

Alibaba has invested in multiple large model companies, including Zhipu AI, Baichuan Intelligence, 01.AI, Kimi and MiniMax.

For the financings of Kimi and MiniMax, there have been clear reports confirming that they included arrangements for discounted computing power; the large model of 01.AI was directly built on Alibaba Cloud's platform; Wang Xiaochuan of Baichuan Intelligence publicly stated deep cooperation with Alibaba Cloud in model pre-training and deployment.

This makes sense. Alibaba doesn't want to be just a financial investor; the mid- to short-term financial performance of large models is highly uncertain. When Alibaba invests in large model companies, it naturally hopes these companies will integrate into the Alibaba Cloud ecosystem, bringing computing power consumption, customer cases, and an ecological position to Alibaba Cloud.

Even speaking at a Cloud Town event would be a plus.

Microsoft's investment in OpenAI follows a similar logic.

Microsoft didn't just give OpenAI money; it also provided Azure. OpenAI didn't just give Microsoft financial returns; it also brought cloud business growth, an AI entry point, and a strategic position in the enterprise market.

Until recently, when their cooperative relationship was adjusted, OpenAI and Microsoft had a strong binding relationship, with Azure being the exclusive cloud provider for OpenAI's models and products.

The "White Whale Lab" report mentioned Alibaba's desire for a "closed-loop ecosystem," which is likely referring to these aspects.

ByteDance Doesn't Invest Either

Should Alibaba actively abandon its conditions to secure a shareholder seat?

I don't think so.

The commercial returns Alibaba can gain from investing in DeepSeek essentially come down to two paths.

The first path is direct profit sharing. This path is currently unclear, as Liang Wenfeng is entirely focused on achieving AGI and basically doesn't consider commercial aspects.

The second path is indirect conversion into Alibaba's revenue through synergies in cloud services and the AI ecosystem. This path, however, is precisely the core conflict in the current negotiations.

It's important to note that Alibaba's attitude toward open-source has recently undergone a subtle shift, from "actively embracing open-source" to "conditionally embracing open-source."

Alibaba is more inclined to monetize its strongest model capabilities through cloud APIs and commercialization systems rather than opening them up unconditionally. DeepSeek, on the other hand, shows no signs of tightening its open-source strategy, representing a conflict in values.

A deeper reason is that Alibaba still has cards to play, as Qianwen remains in the first tier of model research and development.

The opportunity with large models is significant enough that every major company has internally debated the same question: Should we invest in a top-tier model company or develop our own top-tier model? As long as there's a choice, everyone will choose the latter.

ByteDance was the most decisive. As reported by LatePost, ByteDance internally discussed investing in large model companies but ultimately chose not to, stopping all external investments in model companies by mid-2023. Zhang Yiming's question at the time was: Why not do it ourselves?

ByteDance's Volcano Engine MaaS call volume now ranks first domestically, and Doubao's daily token consumption is far ahead, to the point where subscription packages up to 5,000 yuan per year are being discussed.

Alibaba's approach has been to heavily invest in self-research while also scanning the market for opportunities.

It invests frequently externally, but part of the motivation is to build an ecosystem, turning invested companies into major clients of Alibaba Cloud while hedging against the risk of missing out on leading models. When ecological synergy cannot be achieved, Alibaba's conservatism is a rational choice.

In contrast, Tencent faces a slightly different situation.

The performance of its Hunyuan series models has been relatively pressured among major companies.

According to a report by Dianchang, Martin Lau once reflected internally on why Yuanbao's retention rate and user growth fell short of expectations, with one core reason pointing to the inadequacy of the Hunyuan large model: It was difficult to use for custom training in specific scenarios, leading to obvious fragmentation in the search chain.

According to Bloomberg's report last month, Tencent has a strong appetite for acquiring shares in DeepSeek. Sources said Tencent proposed acquiring up to 20% of DeepSeek in this financing round, but DeepSeek was reluctant to cede such a large proportion of control.

Tencent generally does not interfere in the operations of its invested companies, which aligns with Liang Wenfeng's aspirations. Tencent should have no problem participating in this round; the question is how much of a stake it can secure.

Of course, even if Alibaba ultimately does not become a shareholder of DeepSeek, it will not completely lose access to DeepSeek's technological achievements.

As long as DeepSeek continues to adhere to open-source principles, anyone can use it.

Alibaba can use it, Tencent can use it, ByteDance can use it, and we all can. The difference is that having equity might mean deeper cooperation space, earlier access to technology, and stronger bargaining chips for exclusivity; without equity, one can only use the existing offerings under the open-source license.

"Investment talks collapsing" does not equal "cooperation breaking down." Alibaba can still participate in DeepSeek's ecosystem as a cloud service provider or ecological partner without necessarily securing a shareholder seat. Equity investment is just one way to bind interests; if this way doesn't work out, continuing to cooperate in another way may not be a worse outcome for Alibaba.

Of course, all of this The premise (a typo in the original, meaning " provided that " or " premise is") that DeepSeek continues to be open-source.