The fierce competition in China's internet industry, characterized by a long-standing war of cash burning, has re-emerged in the context of artificial intelligence and large language modelsThis intense rivalry is reminiscent of previous battles in the sector, where industries such as group buying, meal delivery, e-commerce, travel, ride-hailing, and the sharing economy faced aggressive financial skirmishesThe arena was marked by technology giants alongside venture capitalists relentlessly pouring funds into their initiatives.

In the wake of a significant downturn catalyzed by macroeconomic factors, these tech powerhouses seemed to have paused their extravagant spending, focusing instead on refining their business modelsEmerging phenomena like the metaverse and Web3 failed to gain significant traction, leaving the market relatively quietHowever, the dawn of AI applications has sparked a revival of aggressive competition akin to previous years, with major players known as the "Big Five AI Models" — including Zhiyou AI, the Dark Side of the Moon, Baichuan Smart, MiniMax, and Lingyi Wanwu — along with tech behemoths like ByteDance and Tencent, now participating in an intense user acquisition contest across various platforms since early 2023.

One entrepreneur who jumped into the AI model race following the launch of ChatGPT observed a striking shift in industry focus

“Last year, everyone was still grappling with product development and trying to catch up to international standardsThis year, we’re all worrying about user growth,” he remarkedAs of June 2023, AI-generated content (AIGC) applications in China saw a staggering rise, reaching 61.7 million monthly active users — an increase of 653% year-on-yearNevertheless, the influx of users substantially raised customer acquisition costs.

Leading applications such as Kimi from the Dark Side of the Moon and Doubao from ByteDance turned this technological revolution into a battle of financial muscleReports from early 2023 revealed that Kimi’s average customer acquisition cost lingered between 12-13 yuan per user, with daily costs soaring to at least 200,000 yuanBy this year, media coverage indicated that Kimi was spending around 30 yuan for each registered user through advertising efforts on platforms such as Bilibili.

This disruption marks the return of a cash-burning spree in the industry, prompting new entrants like the Dark Side of the Moon to strategize on how to compete against established giants such as ByteDance

• Apple's Electric Vehicle Project Comes to a Halt
• Fed Likely to Hike Rates by 75 bps in September
• NASDAQ Suffers Five Consecutive Losses
• Profit Strategies in Individual Stock Options
• Fed to Halt Rate Increases at 4%

The conflict over user acquisition is not just a typical competition for market share; it embodies a deeper strategic battle reminiscent of previous rounds of internet warfare.

Any new business model necessitates initial investments, often yielding little in return, to cultivate the marketFor internet companies, the overwhelming capital backing enables extraordinary expenditures, with the allure of achieving a “winner takes all” scenarioTake, for example, Meituan, which emerged victorious from the “Thousand Group War,” monopolizing merchant resources and securing its position as the leader in meal deliverySimilarly, Didi capitalized on its financial prowess to defeat rivals like Kuaidi and Uber, establishing dominance in the ride-hailing marketHowever, once the battlefield cooled, consumers no longer reaped the benefits of past subsidies.

Investments in traditional internet models prioritize access to users and competitive positioning, often underpinning subsequent funding rounds

What drives AI model applications to invest so heavily in advertising is more than merely user growth; it encompasses a multifaceted approach to acquiring invaluable data that fuels algorithm training.

Kimi's surging user engagement not only strengthens the Dark Side’s position among its investors but also generates vital new data points essential for refining its underlying modelAs algorithms get optimized through internal updates and capital infusions bolster computational power, one crucial element remains: high-quality data, which the models must source externallyAn Oxford University study published in the prestigious journal Nature raised concerns that excessive reliance on artificially generated data in training could lead to a self-degrading cycle, where AI degrades the original content into nonsensical output in just a few generations.

Perhaps to secure precise data, Kimi focuses on engaging students and young professionals on platforms like Bilibili, where over 80 million users view AI-related content monthly

This reflects a staggering annual increase of over 80% in daily views for AI content and over 60% growth in active user participation among AI content creatorsKimi's overwhelming presence on Bilibili has outpaced its competitors, including Zhiyou, MiniMax, and Kunlun Wanwei, which have found it challenging to match Kimi's cost-per-user under the current fast pace.

Funding and high-cost advertising have generated a positive feedback loop for the Dark Side of the MoonJuly statistics from a third-party AI product ranking platform indicated that Kimi's monthly visit count reached 24.56 million, dramatically outpacing its closest competitor, Baidu’s Wenxin YiyanFollowing a substantial funding round in August, the Dark Side of the Moon was valued at approximately $3.3 billion (21 billion yuan), reinforcing its leadership position in the domestic startup landscape.

Yet, despite the influx of users, the high spending does not translate into sustainable revenue models

According to data from Zpeidia, Kimi received 24.05 million visits in July, with only about 3.99 million being unique visits, implying that users opened the app a mere six times on averageEven minimal interactions by new users lead to considerable computational costs for the backend, given the free usage model.

Unlike Meituan and Didi — which could later withdraw subsidies after establishing user habits — the Dark Side’s current customer acquisition strategy does not secure enough engagement to retain a significant user baseThis unsustainable approach raises questions about its viability going forward.

In the realm of business models for AI applications, the path toward lucrative revenue generation via consumer subscriptions remains treacherousFor instance, OpenAI, the parent company of the globally recognized ChatGPT, boasted approximately 7.7 million subscribers and revenues of roughly $1.9 billion

However, these figures are dwarfed when contrasted with its operational costs, which reportedly reach $8.5 billion annually.

The Dark Side of the Moon previously experimented with a feature allowing users to gift Kimi, offering purchase options from as low as 5.2 to 399 yuan for premium servicesThis initiative seemed more exploratory rather than a committed revenue generation strategyIndustry opinions on whether ToB (business-to-business) efforts might offset ToC (consumer) deficits vary widelyFor instance, the founder of Lingyi Wanwu has adamantly claimed, “We focus solely on ToC, avoiding unprofitable ToB ventures,” while the CEO of Zhiyou AI argues that B2B markets exhibit stronger willingness to pay compared to individual consumers.

There are, however, concerns surrounding this approachWhile the ToB model exhibits clearer revenue trajectories, the sector has found itself embroiled in price wars initiated by larger corporations

Reports suggest that domestic companies like Alibaba Cloud and Baidu Smart Cloud had profit margins exceeding 60% prior to May, aligning with international counterparts, but recent price cuts have pushed those margins into the negative territory.

For major players like Alibaba, Baidu, ByteDance, and Tencent, an unrestrained engagement in these pricing skirmishes is mitigated by the profits from their respective cloud servicesFor example, while OpenAI has generated significant pledges of financial support, its demand for cloud services has translated into significant profits for its high-profile backer, MicrosoftAllegations have surfaced indicating that an $800 million installment from a recent $1 billion investment in the Dark Side included substantial allocations aimed at utilizing Alibaba’s cloud offerings.

Yet, despite the perceived short-term gains, venture capital-backed startups are left with no choice but to explore ToB opportunities

Zhiyou AI embodies this mindset, championing ToB initiatives while still investigating potential ToC pathwaysIn contrast, MiniMax approaches both fronts, launching enterprise-level APIs while leveraging Kimi's rising profile in the consumer space.

This parallel also resonates in sectors like autonomous driving, where the need for extensive data collection to enhance driving software is as pressing as purchasing power for companiesTraditional automotive manufacturers utilize hardware sales to subsidize software investments while firms relying solely on software face challenges partnering with vehicle manufacturers to maintain financial viability.

Even as various industries confront insurmountable challenges from established giants, these hurdles permeate all sectorsThe age-old dilemma remains pertinent: What are the implications of industry titans, such as the BAT (Baidu, Alibaba, Tencent), imitating or overshadowing the innovations of a burgeoning company?

During the cash-burning phases that marked earlier eras, entrepreneurs often encountered the question, “What will you do if BAT copies you?” This query looms large over modern AI applications as well

Recently, as the Dark Side of the Moon celebrated Kimi’s significant expansion in capabilities from processing 200,000 words to an astounding 2 million words, it sought to distinguish itself in the crowded marketplaceAccording to data analytics firm Similarweb, this enhancement drove a peak in daily active users to 346,000, with weekly engagement surging by 45%.

However, once such advancements are made public, competitors inevitably adaptWithin a month of Kimi's breakthrough, the AI products backed by Alibaba and 360AI announced their own impressive long-text processing capabilities, subsequently pressuring the Dark Side of the Moon’s positioning in the marketAs these larger firms adopted advanced capabilities, Kimi found its rival, Doubao, emerging rapidly, intensified by ByteDance’s supportive marketing tactics on platforms such as TikTok, effectively pushing Kimi to defend its territory in Bilibili.

As the pulses of competition intensified, the battle has shifted towards more diverse channels, with larger companies leveraging existing ecosystems to their advantage

The involvement of established giants has established a backdrop where emerging models must compete against deeply entrenched content ecosystemsFor instance, where several models source their data, larger platforms leverage extensive content networks — like Baidu utilizing its own Baijiahao or Tencent capitalizing on WeChat’s ecosystemIn contrast, startups like Kimi often rely on public information and third-party sources.

Your success in the ever-evolving AI landscape hinges significantly on how well you navigate these challengesAs capital abundance facilitates technological advances, the onus remains on firms to strategically allocate resources toward initiatives promising long-term returnsThe failure to effectively tackle the complexities of competition and investment may eventually transition startups from trying to fend off competition to needing to prepare for acquisition attempts by the very companies they once aimed to surpass.