Last month, the Baidu Cloud Intelligence Conference, themed “Intelligent Leap,” captured significant attentionThis gathering brought together industry leaders and innovative enterprises, with Baidu, as the host, heavily engaged in activitiesNot only did it strike a strategic partnership with Samsung, but it also showcased its prowess during the main forum by inviting key partners to jointly display exemplary case studiesAdditionally, it created an immersive technology exhibition area of over 10,000 square meters to demonstrate its smart cloud products extensively.

The slogan of the event, “Open, Cooperation, and Win-Win,” reflects Baidu Cloud's efforts in forging alliances, promoting products, and developing its ecosystemHowever, in reality, the landscape of China’s public cloud market suggests a different narrativeCurrently dominated by major players like Alibaba Cloud, Huawei Cloud, Tianyi Cloud, and Tencent Cloud, Baidu Cloud finds itself in the role of a persistent challenger.

With smaller volumes but grand aspirations, Baidu Cloud is grappling with the complexities of the cloud computing arena, often described as one filled with richness yet mediocrity.

The Story of AI Clouds: Are They Lucrative?

The cloud market may appear bustling, but those within understand the actual state of affairs

Recent data from the international market research firm IDC shows a decline in the growth rate of China's public cloud marketIn 2023, the total scale of the public cloud market reached $39.49 billion, with a growth rate of 12.7% for cloud infrastructure and platform services (IaaS+PaaS), marking a consistent decline over three consecutive years.

Despite this downward trend, Baidu Cloud has historically stood on the sidelinesSince its inception in 2012, it lagged behind Alibaba and Tencent by two to three yearsAs of the second quarter of 2024, Alibaba Cloud commanded over 36% market share while Tencent held more than 16%. In contrast, Baidu Cloud was relegated to the “others” category, virtually invisible on the leaderboard.

Yet, as AI began to ascend in popularity, Baidu saw a ray of potentialAn IDC report from August 26 revealed that the AI public cloud market in China reached a scale of 12.61 billion yuan in 2023, a staggering year-on-year growth of 58.2%, far exceeding the growth of public cloud services

This moment presented Baidu with a long-awaited opportunity—after establishing its “intelligent cloud” strategy in 2019, AI emerged as Baidu's core focusSince then, Baidu's visibility in the AI public cloud sector has grown significantly, with IDC reporting a market share of 26.6% for Baidu Smart Cloud in this area, maintaining its top position for five consecutive years.

However, the size of the AI cloud market remains uncertainAccording to IDC's latest report, while Baidu Smart Cloud leads the Chinese large model platform market with a share of 19.9% in 2023, it secured only 10 out of nearly 260 project bids during the first five months of 2024. Given that large model projects are intricately connected to cloud services, this performance raises serious questions about Baidu Cloud's ability to meet market expectations.

Moreover, it is essential to note that the AI market is evolving from a heavy focus on internet users to engaging traditional industries (government, finance, telecommunications, manufacturing, energy) with comprehensive solutions

As the customer base and demands shift, the potential for the AI cloud market requires re-evaluation.

A senior executive from a competitor commented that Baidu Cloud's market position hinges on its differentiated AI product offerings—AI computing power at the IaaS layer, AI platforms at the PaaS layer, AI models at the MaaS layer, AI applications at the SaaS layer, alongside its Wenxin large model and various computing platformsThese elements collectively support its overall product framework.

"Fortunately, Baidu Cloud still has AI as a core component; if it focused solely on basic cloud products, it would find itself in a fiercely competitive red ocean."

Expanding Horizons: The Cost of Aggressive Growth

Baidu Cloud’s ambitions are widely recognizedAccording to internal disclosure from CEO Robin Li, Baidu Cloud has already penetrated over 30 industries, including education, real estate, machinery, and business services

alefox

Such broad outreach aligns with Alibaba's Zhang Yong’s point that “every industry deserves a reimagining through AI.”

Additionally, Baidu Cloud's multi-industry approach has yielded positive resultsIn the past three quarters (2023 Q4 to 2024 Q2), revenue growth rates were 11%, 12%, and 14%. In the earnings call for the second quarter of 2024, it was revealed that revenue from large models had increased from 4.8% of total income in Q4 2023 to 9% in Q2 2024.

While this growth is commendable, some insider perspectives raise concernsIndustry insiders suggest that Baidu Cloud might exaggerate user experiences to attract developers, which could lead to scrutiny“Marketing and technology are fundamentally different logics,” they argue.

Nonetheless, Baidu Cloud seems desperate for market validationRecords indicate that Baidu secured 143 projects in the first half of 2024, with 30 significantly tied to large model or AI-related products, bringing in much higher profit margins than labor-intensive packages

Better yet, its large model platform, Qianfan, recorded a staggering 700 million daily calls by September—over ten times the figure from 2023, illustrating strong market acceptance.

Despite this success, a paradox looms"The competition in large models essentially boils down to computing power, an area where Baidu is still not on par with leading competitorsWith a price war underway, Baidu cannot afford to back down." An industry insider voiced his observations, cautioning that this could potentially squeeze profitability.

"Previously, there was room to make profits, but drastic price cuts have transformed margins from over 60% to zeroEach invocation incurs a loss; the competition turns into who has deep pockets rather than who can expand the market." A player within the industry expressed concerns over the brutal price wars that have erupted.

The dilemma lies in whether to target existing customers or pursue new clientele

For instance, startup Minimax organized a bidding event to procure computing power, where Alibaba Cloud, Baidu Cloud, and ByteDance's Volcano Engine all participatedYet the only concern from Minimax—the client—was price.

No one wants to lose valuable existing clients like MinimaxTypical price points from various cloud platforms range from 60% to 50% of regular pricingHowever, in this bidding war, rates dropped to a mere 20%, which is extraordinarily severe.

Baidu Cloud's rationale reflects a broader sentiment among leading players: the large model market remains in its infancy, and client numbers are paramount to market growth.

However, the trend is more revealingIndustry insights predict a potential exponential increase in large model calls over the next 1 to 2 yearsUnder this marginal effect, cloud providers’ computing costs may flatten with growing demand, ultimately leading to positive profit scenarios

Consequently, as elimination rounds commence, only 3 to 5 foundational model enterprises will likely survive, relying heavily on absolute market share dominance.

When faced with such an impasse, Baidu Cloud bets on its technical capabilities through its Wenxin large model, Qianfan large model platform, and Baijiao heterogeneous computing platform, all directed toward optimizing computing tasks and performance to resolve issues of computing shortages and high costsThese endeavors are essential for garnering market approval.

While this strategy appears promising, the challenge remains that competitors share similar ambitions, presenting an unavoidable scenario of product homogeneity.

The Reality of Funding AI Clouds: Can They Survive?

In truth, the widespread usage of prominent large model platforms incurs massive costs, raising questions about sustainability as Baidu Cloud encounters rumors about its commitment to these initiatives.

Last month, whispers circulated about Baidu potentially abandoning its universal foundational large model due to escalating competition and disproportionate input-output ratios

As the gossip gained traction, Baidu had to respondOn September 9, Zhang Quanfeng from Baidu’s Wenxin Yi Yan marketing department declared on social media, "The rumor about abandoning the universal large model development is pure misinformation! Wenxin Yi Yan has just undergone a comprehensive functional upgradeWe will continue to ramp up investment in universal large model research and development."

The large model remains crucial to Baidu Cloud, and a withdrawal at this point would be illogicalYet, the rumor serves as a stark reminder of the harsh realities: how long can players sustain continuous cash burn?

A Gartner report released in August projects that generative AI is still 2-5 years away from reaching production maturityHowever, within this timeline, the financial demands on Baidu Cloud for operational efficiency are dauntingFor instance, Anthropic’s CEO revealed that their current AI model training costs reach a staggering $1 billion, with projections indicating that costs could hit $10 billion in just three years.

If costs are immense for foreign companies, the situation is not much better for Baidu domestically

A year ago, Guosheng Securities estimated that a single training session for GPT-3 costs around $1.4 millionFor larger language models, expenses could range from $2 million to $12 millionGiven Baidu's plans to benchmark against ChatGPT, the expenses for its Wenxin large model are inevitably astronomicalMoreover, the cost of cloud infrastructure intensifies the financial burdenIn this arms race, the question of how much Baidu will invest becomes crucial.

Has Baidu adequately prepared for this challenge? Concerns arise when considering Baidu's Q2 2024 financial report, which indicates total revenue of 33.9 billion yuan, with core revenue of 26.7 billion yuan and a core operating profit of 5.6 billion yuan—up 23% year-on-yearDespite a notable figure, its primary revenue stream—online advertising—declined by 2% to 19.2 billion yuanIn contrast, Baidu Cloud's revenue that quarter was 5.1 billion yuan, of which only 9% stemmed from external demand for large models and generative AI services

Importantly, the lack of clear data on profit composition from these two major revenue sources warrants concern.

“Baidu’s capacity to fund its cloud and large model initiatives remains uncertain,” an industry analyst remarked on Baidu Cloud's future investmentsDespite Baidu holding cash, equivalents, and investments totaling approximately 162 billion yuan as of June 30, 2024, the reality is that only around 6.2 billion yuan is flexible cashWith the persistent price wars in the cloud platform space, Baidu could face complex challenges ahead.

In fact, Baidu Smart Cloud has been continuously reducing product prices this year: in May, it announced free access to its ERNIE-Speed, ERNIE-Lite, and ERNIE-Tiny series model services; in July, it slashed prices for flagship models ERNIE 4.0 and ERNIE 3.5, with rates plunging to 0.03 yuan (for input) and 0.06 yuan (for output) per 1,000 tokens.

This may explain the hidden profits in its financial report: spending without profit is unappealing.

The situation extends beyond just Baidu Cloud; it reflects its parent company, struggling with insufficient revenue generation to weather this intense period of competition