---
title: "AI devouring software? Bean buns are devouring everything"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/280857758.md"
description: "In 2011, Silicon Valley investor Marc Andreessen proclaimed that \"software is eating the world,\" marking the beginning of a golden decade for mobile internet. By 2026, Naval's statement that \"software is being eaten by AI\" sparked heated discussions, signaling the end of the software era and the arrival of the AI era. The global SaaS index has fallen nearly 30% since its peak in 2025, intensifying industry panic. The true endgame is not countless AI tools replacing applications, but a super AI gateway that consumes all applications. In China, this gateway has already emerged—Doubao"
datetime: "2026-03-28T01:14:04.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/280857758.md)
  - [en](https://longbridge.com/en/news/280857758.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/280857758.md)
---

> Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/280857758.md) | [繁體中文](https://longbridge.com/zh-HK/news/280857758.md)


# AI devouring software? Bean buns are devouring everything

In 2011, legendary Silicon Valley investor Marc Andreessen proclaimed, "Software is eating the world," ushering in the golden decade of mobile internet, with countless apps dominating mobile phone desktops. In 2026, Naval's statement, "Software was eaten by AI," went viral, garnering over 100 million views and definitively announcing the end of the software era and the arrival of the AI ​​era. Industry panic has spread: the global SaaS index has plummeted nearly 30% from its 2025 peak, and the "software doomsday" narrative is rampant. But the real endgame isn't countless AI tools replacing apps; it's a super AI gateway that devours all apps. In China, this gateway has already been born—Doubao. 315 million monthly active users, a peak of 145 million daily active users, a 30-day retention rate of 44.5%, and an average of 4.2 app launches per day per user… These aren't just data from ordinary AI tools; they're the declaration of the birth of China's second "national-level super app," after WeChat. When you can shop, hail a ride, order takeout, make videos, and book flights simply by speaking a single sentence to Doubao (a mobile app), Is there still any need for Taobao, Meituan, Didi, and Ctrip on your phone? From "Software Eating the World" to "AI Eating Software" The golden decade of the mobile internet was essentially a "war for app dominance." We break down all our daily needs into individual scenarios and pack them into individual apps. The UI is the only translator between user intent and machine capabilities. To book a flight, you open Ctrip, enter your departure point, destination, and time, filter, compare, and place an order; to order takeout, you open Meituan, select a merchant, choose dishes, fill in your address, and pay. For the past two decades, the battles among all internet giants have essentially been about vying for a place on users' phone home screens, for clicks and user time. But the emergence of AI Agents has completely bridged this gap in human-computer interaction. Users no longer need to translate their intentions, no longer need to learn the operating logic of different apps; they only need to use natural language to say what they want. Booking a flight—no need to open Ctrip—just say, "Book me the cheapest direct flight from Beijing to Shanghai tomorrow afternoon, with a window seat," and AI can complete the entire process in the background. For expense reimbursement, there's no need to open the OA system. A simple command like, "Export my taxi receipts from this week, fill out the reimbursement form, and send it to finance," and AI can complete the process automatically. For shopping, there's no need to open Taobao. A simple command like, "Please choose an amino acid facial cleanser suitable for sensitive skin, within a budget of 200 yuan, and that can be shipped today," and AI can directly recommend products and complete the order. When all operations can be completed by AI in the background, the value of traditional apps is reduced to their backend service capabilities. In the future, only two types of apps will survive: those that encapsulate their service capabilities into APIs for AI to call; and those that are completely forgotten by users, relegated to a corner of their phone's home screen. The global software (SaaS) stock index has fallen nearly 30% from its peak in September 2025, and the market is already experiencing panic about a "SaaSpocalypse." Goldman Sachs, HSBC, and other institutions have released reports predicting that by 2027, the AI market will cross the traditional SaaS market, achieving a complete dominance over the software industry. Doubao is already the second super entry point after WeChat. There are three core criteria for judging whether a product can become a super entry point: a national-level user base, high-frequency and high-stickiness user behavior, and the ability to cover all scenarios of needs. Doubao has already crossed all of these thresholds. With a user base of national scale, it has already left all competitors far behind. We have compiled core user data for mainstream AI products in China in February 2026, and the gap is obvious: Data source: QuestMobile February 2026 China Mobile Internet Industry Report, official data disclosed by various companies. What does 315 million monthly active users mean? This number has exceeded the monthly active user scale of Meituan and Didi, and is very close to the monthly active user level of Taobao. More importantly, Doubao's user growth relied almost entirely on large-scale advertising campaigns DataEye monitoring data shows that in November 2025, Doubao's advertising creative expenditure accounted for only 11% of its total spending, while Tencent's Yuanbao and Alibaba's Tongyi Qianwen accounted for 46% and 34% respectively. ByteDance's investment in Doubao's promotion was only one-third of its competitors'. Low customer acquisition cost, high user retention, and high usage frequency are precisely the core characteristics of a super app. ByteDance's "APP factory" paved the way for Doubao. The core barrier to Doubao's success is not its large-scale technology, but the full-scenario ecosystem empire that ByteDance has built over ten years. ByteDance boasts the most comprehensive national-level APP matrix in China's internet industry, with products boasting over 100 million monthly active users (MAU), including: Douyin (936 million MAU, number one in China) Toutiao (300 million+ MAU) Tomato Novels and Hongguo Short Drama (over 100 million daily active users, nearly 300 million monthly active users) Qishui Music (140 million MAU) Covering content, news, entertainment, e-commerce, In local life, office, and learning scenarios, TikTok boasts over 2 billion monthly active users globally, spanning more than 150 countries and regions. Doubao is the sole gateway to this ecosystem empire. Doubao has already successfully established a closed-loop business model for "one-sentence transactions." The core of a super gateway is never about how much information it can provide, but rather whether it can complete the entire closed loop "from demand to transaction." Doubao has already achieved this. In content creation scenarios, Doubao has achieved a closed-loop end-to-end process: "Generate short video scripts with one sentence → Create videos with one click in CapCut → Publish directly on Douyin." This is a seamless workflow that no other AI tool has achieved. In e-commerce scenarios, Doubao's "voice shopping" function, currently in beta testing, allows users to simply state their shopping needs, and AI will directly recommend products from the Douyin Mall. Orders and payments can be completed within the dialogue interface without redirecting to Douyin, directly connecting to Douyin's trillion-dollar e-commerce supply chain In local lifestyle scenarios, Doubao has partnered with ByteDance's "DouShengSheng" group-buying platform to provide a seamless service: "Find stores, claim coupons, and redeem coupons with one sentence." Integration with travel, cultural tourism, and office scenarios is also in beta testing. From dialogue to transactions to service delivery, Doubao has completed the core closed loop of a super gateway. WeChat uses social interaction to pack all services into one app; Doubao uses AI to pack all services into a single sentence. The collective predicament of AI in major tech companies: not that the technology is inadequate, but that they dare not revolutionize themselves. In stark contrast to Doubao's meteoric rise is the collective collapse of AI in the BAT (Baidu, Alibaba, Tencent) giants. We never doubt the technological capabilities of Alibaba, Tencent, and Baidu, but their AI products, from the very first day of their existence, are shackled by the genes of the old era. Their core revenue comes from the foundation of mobile internet-era apps: Alibaba's e-commerce, Tencent's social media, and Baidu's search—all trillion-dollar cash cows. Their AI can only add value to these cash cows, not disrupt them. This is the classic "innovator's dilemma." Alibaba's Tongyi Qianwen: A False Prosperity Built on Subsidies

Tongyi Qianwen's 203 million monthly active users seem impressive, but in reality, it's a traffic frenzy fueled by a 3 billion RMB Spring Festival red envelope subsidy.

Behind the 552.83% monthly growth rate lies a strong reliance on Alibaba's e-commerce ecosystem—it's essentially a shopping guide tool for Taobao, and all its functions ultimately aim to drive traffic to Taobao.

Once the subsidies stop, users immediately leave. After the event ended, Tongyi Qianwen's user retention and usage time both declined significantly.

Even more fatally, this subsidy frenzy not only failed to establish a distinct user mindset for Tongyi Qianwen, but also caused a server crash. In March 2026, Lin Junyang, the core person in charge of Tongyi Qianwen, suddenly resigned, the team was split and reorganized, and the product's future was fraught with uncertainty Tencent Yuanbao: An appendage in the cracks of WeChat. Tencent Yuanbao, with WeChat's over 1 billion users, only achieved 109 million monthly active users, less than a third of Doubao's. The core reason is simple: from its inception, Yuanbao was confined to the role of a "WeChat office plugin," daring not to touch WeChat's social foundation, let alone disrupt the WeChat mini-program ecosystem. WeChat's core logic is to integrate all services into its ecosystem, while Yuanbao's AI capabilities essentially undermine the entry value of mini-programs. Tencent cannot allow its AI product to revolutionize WeChat. Worse still, Yuanbao also experienced a serious incident where AI verbally abused users, exposing significant flaws in the model's instruction execution and ethical alignment, severely damaging user trust. Baidu's "Wenxin Yiyan": Early to Start, Late to Finish (I'm almost embarrassed to say it) Baidu was the first major company in China to release a large-scale model, holding a first-mover advantage, yet it squandered its advantages Wenxin Yiyan's monthly active users have dwindled to only 24.5 million, declining for two consecutive months and far behind Doubao. The core problem is that Baidu has tied Wenxin Yiyan to its search engine. It remains a secondary function of Baidu Search, not an independent AI entry point. Users with questions open Baidu first, not Wenxin Yiyan. The early paid model caused Wenxin Yiyan to miss the golden window for C-end user growth, falling into a vicious cycle of "user lack - data hunger - slow iteration." While BAT (Baidu, Alibaba, Tencent) were still finding scenarios for their AI and trying to empower their old businesses, Doubao had already used AI to create a completely new internet. This is the fundamental difference. All the efforts of big companies in the AI field are essentially painting an old house, while ByteDance directly built a new city with Doubao. The Fate of Vertical AI: No Matter How Powerful the Technology, It Can't Win the War Without an Ecosystem. Besides major companies, vertical AI players like Kimi and MiniMax have also surprised the market Kimi's long text processing capabilities remain top-tier in the industry, with an average usage time of 25.6 minutes per user, firmly holding the top spot. MiniMax's multimodal capabilities and global reach have also made it a dark horse in the industry. However, we must acknowledge a harsh reality: no matter how powerful their technology, they cannot win the war without an ecosystem. AI without an ecosystem is ultimately just a tool, never a true gateway. The essence of a super gateway is to cover the needs of users throughout their entire lifecycle, creating sustained user engagement. Users use Kimi to read papers and edit documents, then close it afterward. It cannot handle users' daily needs like shopping, ordering takeout, or booking flights, and cannot form a complete closed loop of "demand-instruction-transaction-service-feedback." It can only remain a "productivity tool," never becoming a super gateway. The technological gap will eventually be closed, but the ecosystem barriers will only widen. Even more critically, the technological gap in large-scale models will eventually be closed Parameters, inference capabilities, and context windows—these technical metrics can be caught up with or even surpassed by major companies in just a few months. Baidu and Alibaba quickly launched long-text processing functions similar to Kimi in their respective platforms, directly negating its innovative advantage. As the old saying goes in the industry: when AI functions are embedded in super portals like WeChat, Taobao, and Douyin, any single, independent application will find it difficult to acquire long-term users. MiniMax is the same; over 70% of its revenue comes from overseas, its domestic C-end user base is extremely small, and it lacks a localized ecosystem, forcing it to rely solely on API services, never reaching the threshold of a super gateway. The ultimate outcome of the AI era is never a technology war, but an ecosystem war. ByteDance has spent ten years building an ecosystem empire covering all user scenarios, and Doubao is the only key to unlocking this empire. This is the ultimate barrier that no other player possesses Technology can allow you to create a good product, but only an ecosystem can allow you to create an era. One day, Doubao will devour Taobao, Meituan, and Didi, becoming the endpoint of all needs. We've been saying that AI is devouring software, but what exactly is it devouring? It's not the software's service capabilities, but rather its entry-level value. Taobao's core competitive advantage is its supply chain; Meituan's core competitive advantage is its local business resources; Didi's core competitive advantage is its transportation capacity. But these barriers are merely backend service capabilities in the face of the AI super gateway. Users don't need to care whether the goods come from Taobao or JD.com, whether the food delivery comes from Meituan or Ele.me, or whether the ride-hailing service comes from Didi or Gaode. Users only care whether their needs can be met as quickly and efficiently as possible. When Doubao can access Taobao's supply chain, Meituan's merchant resources, and Didi's transportation capacity, the moats of these giants' gateways are completely eroded by AI They will ultimately be marginalized, reduced to backend supply chain service providers for the AI super gateway. Doubao is replicating WeChat's success, even more thoroughly than WeChat. WeChat uses mini-programs, turning countless apps into its own service providers; Doubao, using AI, eliminates the need for mini-programs altogether. WeChat requires users to click, operate, and navigate within mini-programs, while Doubao hides all these operations in the background. Users simply state their needs and pay for the final result. This represents a complete reconstruction of human-computer interaction logic and a complete subversion of the mobile internet landscape. The ultimate future of the internet will undoubtedly be a "one superpower and many strong players": The "one superpower" is a super AI portal like Doubao, controlling all user needs and traffic; the "many strong players" are countless service providers in vertical fields, providing backend service capabilities for the super portal. Doubao, leveraging ByteDance's full-scenario ecosystem, has already gained a head start, becoming the most likely player to capture this super portal The ultimate outcome of the mobile internet was traffic-driven; the ultimate outcome of the AI era is intent-driven. Whoever controls user intent controls the entire internet. When Doubao can meet all your needs, how many apps will be left on your phone? Looking back at the past thirty years of the Chinese internet, every technological revolution has given birth to new super portals and also caused the giants of the old era to fall from their pedestals. In the PC era, Baidu became the gateway to the internet through search; in the mobile internet era, WeChat became a national super app through social networking; and in the AI era, Doubao is rewriting the final chapter of the internet through natural language interaction. "AI devouring software" is not just an industry slogan fueled by anxiety; it's a reality unfolding before our eyes. Doubao isn't just devouring individual apps on our phones; it's devouring the entire old order of the mobile internet era. Finally, we want to leave you with a thought-provoking question: When you can satisfy all your daily needs with a single sentence from Doubao, how many apps will be left on your phone?

### Related Stocks

- [Tianhong CSI Computer Theme ETF (159998.CN)](https://longbridge.com/en/quote/159998.CN.md)
- [Guotai CES Semiconductor Chip Industry ETF (512760.CN)](https://longbridge.com/en/quote/512760.CN.md)
- [State StreetSPDRS&PSftwr&SvcsETF (XSW.US)](https://longbridge.com/en/quote/XSW.US.md)
- [VanEck Semiconductor ETF (SMH.US)](https://longbridge.com/en/quote/SMH.US.md)
- [State Street® SPDR® S&P® Smcndctr ETF (XSD.US)](https://longbridge.com/en/quote/XSD.US.md)
- [Invesco Semiconductors ETF (PSI.US)](https://longbridge.com/en/quote/PSI.US.md)
- [ChinaAMC SSE STAR Semiconductor Material & Equipment Thematic ETF (588170.CN)](https://longbridge.com/en/quote/588170.CN.md)
- [China Southern CSI Semiconductor Industry Custom ETF (159325.CN)](https://longbridge.com/en/quote/159325.CN.md)
- [Guotai CSI Computer Theme ETF (512720.CN)](https://longbridge.com/en/quote/512720.CN.md)
- [WisdomTree Cloud Computing ETF (WCLD.US)](https://longbridge.com/en/quote/WCLD.US.md)
- [iShares Global Tech ETF (IXN.US)](https://longbridge.com/en/quote/IXN.US.md)
- [iShares Expanded Tech-Software Sect ETF (IGV.US)](https://longbridge.com/en/quote/IGV.US.md)
- [iShares Semiconductor ETF (SOXX.US)](https://longbridge.com/en/quote/SOXX.US.md)
- [Global X Cloud Computing ETF (CLOU.US)](https://longbridge.com/en/quote/CLOU.US.md)
- [ChinaAMC Guozheng Semiconductor Chip ETF (159995.CN)](https://longbridge.com/en/quote/159995.CN.md)
- [ProShares Big Data Refiners ETF (DAT.US)](https://longbridge.com/en/quote/DAT.US.md)
- [The Technology Select Sector SPDR® ETF (XLK.US)](https://longbridge.com/en/quote/XLK.US.md)
- [GTJA Allianz SSE STAR Chip Design Thematic ETF (588780.CN)](https://longbridge.com/en/quote/588780.CN.md)
- [First Trust Cloud Computing ETF (SKYY.US)](https://longbridge.com/en/quote/SKYY.US.md)
- [Direxion Daily Semicondct Bull 3X ETF (SOXL.US)](https://longbridge.com/en/quote/SOXL.US.md)
- [GTJA Allianz CSI All Share Semiconductors & Semiconductor Equipment ETF (512480.CN)](https://longbridge.com/en/quote/512480.CN.md)

## Related News & Research

- [South Korean AI chip startup Rebellions eyes new shores for rack-scale invasion](https://longbridge.com/en/news/281155540.md)
- [Codenotary Launches the First Enterprise Agentic Network Monitoring for Security, Performance and Cost Visibility](https://longbridge.com/en/news/281211815.md)
- [02:21 ETNICE otorga a Bell el premio al Socio Implementador de Integración del Año](https://longbridge.com/en/news/281475103.md)
- [Bureau Veritas To Offer AI Assessment To European Enterprises In Partnership With AWS](https://longbridge.com/en/news/281397262.md)
- [Micron Readies Client Storage for AI](https://longbridge.com/en/news/281383944.md)