Name: 阿里巴巴如何悄悄成為AI領域的領導者！ (How Alibaba Quietly Became a Leader in AI)
Uploaded: 2025-11-22T18:02:02.000Z
Duration: 8 min 50 s

The global AI race is accelerating and Alibaba wants a piece of the action.

全球 AI 競賽正加速進行，Alibaba 也想分一杯羹。

In September 2025, the Chinese tech giant announced a fresh wave of investment in AI and infrastructure.

在 2025 年 9 月，這間中國科技巨頭宣佈了對 AI 和基礎設施的新一波投資。

That was on top of the $53 billion it pledged earlier in the year.

這是在今年稍早承諾的 530 億美元之外。

It was by far one of its boldest strategic bets to date.

這是迄今為止最具野心的策略性投資之一。

There has been a huge shift in what this company is about.

這間公司的定位發生了巨大的轉變。

it comes at a very particular moment where clearly the world is looking at AI as the next big thing.

而且這發生在一個非常關鍵的時刻，顯然全世界都將 AI 視為下一個重大發展。

And while the company's technology has positioned it as a serious contender in the quest for AI leadership, its rise has received less fanfare than rivals like OpenAI or Deepsea.

雖然這間公司的技術使其成為 AI 領導地位的有力競爭者，但它的崛起卻不像 OpenAI 或 Deepsea 那樣受到矚目。

Deepsea changed a bit in how we see Chinese team.

Deepsea 稍微改變了我們對中國團隊的看法。

We stop seeing them as in worse version of the Western mobile.

我們不再將他們視為西方手機的較差版本。

Alibaba is participating and being an important player.

Alibaba 積極參與，並扮演著重要角色。

This is a huge tech giant with more than 900 million users reinventing itself.

這是一家擁有超過九億用戶、正在重塑自我的巨大科技巨頭。

So how did Alibaba quietly become a leader in the AI race?

那麼，Alibaba 是如何悄悄成為 AI 競賽的領導者呢？

I'm Arjun Karpel and I've been covering tech for more than a decade.

我是 Arjun Karpel，我已經關注科技領域十多年了。

But three of those years I've been in China, delving into the country's most fascinating giants.

其中三年，我待在中國，深入瞭解這個國家最引人入勝的科技巨頭。

And in this episode I'll unpack how it became a global AI powerhouse.

在這集節目中，我將拆解它如何成為全球 AI 強權。

Alibaba's AI journey really started in 2016.

Alibaba 的 AI 之旅其實始於 2016 年。

This is when they built their first AI lab and did some experiments and it really led them then to go more fundamental in 2017, 2018, when they also built the Damo Academy, which was an important research institute.

那時他們建立了第一個 AI 實驗室並進行了一些實驗，這也真正促使他們在 2017、2018 年走向更基礎的研究，當時他們也建立了達摩院，這是一個重要的研究機構。

Acceleration happened really during the COVID years 2019, 2021, when they really started to build their own foundational models and their own chips.

加速發展發生在 2019 年到 2021 年的 COVID 期間，那時他們開始建立自己的基礎模型和晶片。

But what really spurred Alibaba on was the rise of ChatGPT.

但真正促使 Alibaba 前進的是 ChatGPT 的崛起。

In late 2022, the app hit 1 million users in under a week, kicking off the AI race.

在 2022 年底，這款 app 在一週內就達到了 100 萬用戶，這也正式啟動了 AI 競賽。

In response, Alibaba launched its own world class AI research lab and began backing startups, hoping to become China's answer to open AI.

為此，Alibaba 推出了自己的世界級 AI 研究實驗室，並開始扶植新創公司，希望能成為中國對抗 open AI 的答案。

Eddie wu's appointment as CEO less than a year later cemented Alibaba's commitment to its reinvention as an AI company.

不到一年後，Eddie Wu 被任命為執行長，這鞏固了 Alibaba 轉型為 AI 公司的決心。

In his first letter to employees, Wu called for Alibaba to return to the startup mindset and set two strategic priorities, user first and AI driven.

在他給員工的第一封信中，Wu 呼籲 Alibaba 回歸新創公司的思維，並設定了兩個策略重點：用戶至上和 AI 驅動。

And in the 15 months following his appointment, Alibaba's quarterly capital expenditure surged more than doubling compared to previous years as the company ramped up investment in AI infrastructure and cloud services.

在他上任後的 15 個月裡，Alibaba 的季度資本支出激增，與前幾年相比翻倍，因為公司加大了對 AI 基礎設施和雲端服務的投資。

At the same time, it began selling off businesses it didn't see as critical in order to free up resources for its new initiatives.

同時，他們開始出售那些他們認為不重要的業務，以便為新計畫騰出資源。

All of these efforts have culminated into a suite of successful AI products.

所有這些努力都匯聚成了一系列成功的 AI 產品。

First, there's Qwen, Alibaba's family of large language models that can understand text, answer questions, translate language and work with images.

首先是 Qwen，這是 Alibaba 的大型語言模型系列，可以理解文字、回答問題、翻譯語言並處理圖像。

It's akin to the models that OpenAI has developed that underpin ChatGPT.

這就像 OpenAI 開發的、作為 ChatGPT 基礎的模型。

Unlike most of OpenAI's models though, Alibaba's offering is open source, meaning developers can modify, redistribute and even use the models for commercial purposes.

不過，與 OpenAI 的大多數模型不同的是，Alibaba 的產品是開源的，這意味著開發者可以修改、重新發布，甚至將這些模型用於商業用途。

The default way to release models in China is to open source them, so to share them freely.

在中國，發布模型的預設方式是開源，也就是免費分享。

And on the other hand, in the US the default way is mostly to release closed models so that you can't really download them.

而在美國，預設方式大多是發布閉源模型，你其實無法下載它們。

And there is a big difference because if you release a closed model, you can make a specific type of business model which is selling access to the model.

這之間有很大的差異，因為如果你發布閉源模型，你可以建立一種特定的商業模式，也就是出售模型的存取權。

But on the other hand, when you release an open source model, you get a lot more adoption.

但另一方面，當你發布開源模型時，你會獲得更多的採用率。

There is a full ecosystem that grow around this type of models.

圍繞著這類模型會成長出一個完整的生態系統。

China's latest AI breakthrough has leapfrogged the world.

中國最新的 AI 突破已經讓世界望塵莫及。

I think we should take the development out of China very, very seriously.

我認為我們應該非常、非常嚴肅地看待中國的發展。

At the start of 2025, Chinese AI startup Deepseek shocked the western world with the release of its R1 model.

在 2025 年初，中國 AI 新創公司 Deepseek 發表了 R1 模型，震驚了西方世界。

They strategically about seven, eight days before the Chinese New Year, which is China's largest holiday, it made big news.

他們策略性地選在農曆年前，也就是中國最大的節日，大約七、八天前發布，這引起了很大的新聞。

Alibaba team that was in charge of Q1 canceled the holiday, stayed in the office for 8 days and by 28th of January, Alibaba came up with the Q1 2.5 which was on par with deep seq model.

負責 Q1 的 Alibaba 團隊取消了假期，在辦公室待了 8 天，並在 1 月 28 日前推出了 Q1 2.5，其表現與 Deepseek 模型不相上下。

And in April 2025, Alibaba followed with QN3, its most advanced model to date.

而在 2025 年 4 月，Alibaba 接著推出了 QN3，這是他們迄今為止最先進的模型。

QN also powers Alibaba's AI assistant app, Quark, which has almost 150 million monthly active users.

QN 也為 Alibaba 的 AI 助理 app Quark 提供支援，該 app 擁有近 1.5 億的月活躍用戶。

So Quarq allows you to enter this AI toolbox for business use.

所以 Quark 讓你能夠進入這個用於商業用途的 AI 工具箱。

So it is an operating system for life that is run on a smart AI algorithm.

這是一個運行在智慧 AI 演算法上的生活作業系統。

You are ultimately speaking with an intelligent genie who helps you get stuff done.

你最終是在與一位能幫助你完成事情的智慧精靈對話。

ModelScope is like Alibaba's version of an AI app store for developers and researchers.

ModelScope 就像是 Alibaba 專為開發者和研究人員打造的 AI app 市集。

It offers hundreds of pre trained models that users can customize and build on, making it easier to plug AI into their businesses without starting from scratch.

它提供了數百種預先訓練好的模型，使用者可以客製化並在其基礎上開發，讓企業能輕鬆導入 AI，而無需從零開始。

They have 16 million developers on that platform and they build thousands and thousands of models on there.

他們的平臺上有 1600 萬名開發者，並在此建立了成千上萬的模型。

So they really believe in the open source approach because it allows them to expand fast throughout an ecosystem.

所以他們非常相信開源方法，因為這能讓他們快速擴展整個生態系。

Behind the scenes, Alibaba has turned its cloud platform Aliyun into an AI distribution hub.

在幕後，Alibaba 已將其雲端平臺 Aliyun 轉變為 AI 分發中心。

Businesses can run QN and other models on Alibaba cloud to power tools like Chatbots.

企業可以在 Alibaba Cloud 上運行 QN 和其他模型，為 Chatbots 等工具提供支援。

Alibaba is positioning Aliyun as a serious rival to Amazon's AWS and Microsoft's Azure.

Alibaba 正將 Aliyun 定位為 Amazon AWS 和 Microsoft Azure 的有力競爭者。

In terms of model performance, Alibaba is a little bit closer to an Amazon than compared to an OpenAI or a Microsoft.

在模型效能方面，Alibaba 比起 OpenAI 或 Microsoft，與 Amazon 更為接近。

But we should not forget model performance is not what makes a company successful in AI.

但我們不應忘記，模型效能並非讓公司在 AI 領域成功的關鍵。

It is about scale, it's about adoption and it's about solving real problems.

關鍵在於規模、採用率以及解決實際問題的能力。

And this is where Chinese companies in general, but Alibaba in particular is really, really good at.

這正是中國公司普遍，尤其是 Alibaba 特別擅長的地方。

A Morgan Stanley report released in May 2025 projected that AI investments in China could deliver of 52% by 2030.

一份於 2025 年 5 月發布的 Morgan Stanley 報告預測，中國在 AI 的投資到 2030 年可能帶來 52% 的回報。

And Alibaba is already seeing some of that.

而 Alibaba 已經看到了一些成果。

In August 2025, its Cloud Intelligence Group posted a 26% year on year revenue increase while AI related projects achieved triple digit growth for eight consecutive quarters.

在 2025 年 8 月，其雲端智慧事業群的營收年增 26%，而與 AI 相關的專案已連續八個季度實現三位數的成長。

These big investment comes at a pivotal strategic moment where Baba already has a very strong foundation, a very popular open source model, a very successful application on the ground across the country, when Baba already is building together with local governments, universities, et cetera, multiple AI labs.

這些重大的投資發生在一個關鍵的策略時刻，當時 Alibaba 已經擁有非常堅實的基礎、一個非常受歡迎的開源模型、以及在全國各地非常成功的落地應用，同時 Alibaba 也正與地方政府、大學等合作，建立多個 AI 實驗室。

The challenge for cloud outside of China is that there's a lot of competition and I think Alibaba probably has to accelerate that part of the business, which exactly is why they're investing so much money in AI Industry partnership announcements have also bolstered its credibility, including one with BMW to integrate Qwen LLMs into BMW's intelligent vehicles launching in 2026.

中國以外的雲端服務面臨的挑戰是競爭非常激烈，我認為 Alibaba 可能必須加速這塊業務的發展，這也正是他們投入大量資金於 AI 的原因。產業合作夥伴關係的宣佈也增強了其信譽，包括與 BMW 合作，將 Qwen LLMs 整合到將於 2026 年推出的 BMW 智慧汽車中。

Another deal with Apple could bring Alibaba's AI to iPhones in China.

另一項與 Apple 的協議可能讓 Alibaba 的 AI 進入中國的 iPhone。

The partnerships are real test cases for them to build deep expertise and and models that are completely tuned to very complicated industries like automotive.

這些合作夥伴關係是他們建立深厚專業知識和針對汽車等複雜產業進行完全調校模型的真實案例。

Partnerships with an Apple or a BMW help them also to make sure that this business does not go to competitors and that they have at least an arm and a leg in those global organizations.

與 Apple 或 BMW 的合作夥伴關係也有助於確保這塊業務不會落入競爭對手手中，並讓他們在全球組織中至少佔有一席之地。

Wherever you look, whatever you touch, where China is moving closer towards that vision of dominating AI race by 2030, Alibaba is participating and being an important player.

無論你往哪裡看，觸摸什麼，中國正朝著 2030 年主導 AI 競賽的願景邁進，Alibaba 正在參與其中，並扮演著重要的角色。