15 亿美元 AI 独角兽崩塌,全是印度程序员冒充!微软亚马逊惨遭忽悠

深潮Published on 2025-05-25Last updated on 2025-05-25

Builder.ai 自称用 AI 简化软件开发,吸引微软、软银等巨额投资,估值一度超 15 亿美元。但实际靠人工冒充 AI,创始人虚报三倍营收。丑闻曝光后,投资被冻结,公司原地破产。

印度老哥是真的有点猛啊!

今天要说的这位,是AI编程公司Builder.ai的创始人兼前CEO——Sachin Dev Duggal。

  • 他不仅造了个「全是人工,没有智能」的假AI公司

  • 从软银、微软等巨头手里骗到了数亿美元融资,估值干到15亿

  • 而且还敢对投资人虚报300%的营收

是的,这家公司的后台里并没有AI,有的只是一群印度老哥假装AI写代码。

更劲爆的是,这一骗愣是坚持了8年。

但这周他算是彻底玩完了。

创始人兼前首席执行官Sachin Dev Duggal

随着最近一次「欺诈」的曝光,上一轮的投资人吓得赶紧冻结了投资账户里的剩下的3700万美元(共投资5000万美元),只给公司账户留下500万美元,而这500万美元还受限于政府的资金出境规定,也没法用来发工资。

没办法,Builder.ai只能申请破产,此时的CEO早已换成来「擦屁股」的Manpreet Ratia——创始人Sachin Dev Duggal于2月份辞去CEO职务,由Ratia接替。

这场闹剧直接导致了自2022年ChatGPT发布以来,AI初创公司中规模最大的一次倒闭事件——这家公司在上一轮融资中估值已经超过15亿美元。

Builder.ai的破产清算通告

Builder.ai的官网已经无法访问,只剩下两个联系邮箱

而这场风波中的「冤大头」除了上面提到的提供了5000万美元的Viola Credit外,还有两年前牵头2.5亿美元融资的全球最大主权财富基金之一——卡塔尔投资局(QIA)。

以及同年也进行了投资,并成为战略伙伴的微软。甚至,他们还把Builder.ai集成到了云服务中。

黄金时代

Builder.ai诞生于伦敦,源于其创始人Sachin Dev Duggal对传统软件开发的不满。

在AI驱动的叙事黄金时代,Builder.ai有一个好到不容忽视的宣传口号:让软件开发「像点披萨一样简单」。

这家成立于2016年的初创公司声称,它可以通过一个据称由AI驱动的平台,让非工程师也能够构建复杂的应用程序,从而普及软件开发。

AI的宣传口号对投资者而言效果奇佳。

Builder.ai的前身叫做Engineer.ai,总部位于伦敦和洛杉矶的公司2018年从包括Deepcore Inc.在内的投资者处筹集了 2950万美元资金,Deepcore Inc.是软银(SoftBank)的全资子公司。

其他投资方还包括总部位于苏黎世的风险投资公司 Lakestar(Facebook Inc.和 Airbnb Inc.的早期投资者)以及总部位于新加坡的Jungle Ventures。

创始人Sachin Dev Duggal在早年的一次科技会议上

到2022年,Builder.ai已筹集了1.95亿美元,并于2023年5月,在由卡塔尔投资局 (QIA) 领投的一轮融资中又增加了2.5亿美元。

同年,微软作为战略投资者和合作伙伴加入,将其Builder.ai平台集成到其云服务产品中。

这带来了巨大的认可,随之而来的期望也同样巨大。

在接下来的8年里,它筹集了超过4.45亿美元的资金,其投资者包括微软和卡塔尔投资局,公司估值也跨过了 13 亿美元的大关。

Builder.ai提供的解决方案是:将模块化代码组件与人类开发者相结合,并由AI进行协调。

其名为「Builder Studio」的平台,配备了一个名为「Natasha」(娜塔莎)的数字助理,承诺提供由AI驱动的无缝用户体验。

Builder.ai酷炫的官网,如今已经全部无法打开

但这个愿景的背后实际上是:大部分工作是由印度的开发人员完成的,而非AI。

2019年,「华尔街日报」揭露了一个令人尴尬的真相:Builder.ai的AI更多的是营销噱头,而非工程突破。

多位现任和前任员工表示,一些定价和时间表的计算是由传统软件来做的,剩下的大部分工作也都是由员工手动完成。

如果你告诉客户你在使用AI,他们很可能不会想到上世纪50年代的技术。决策树是一项非常老旧且简单的技术。

这些人表示,公司缺乏自然语言处理技术,并且公司内部使用的决策树不应被视为AI。

正如报道所言,Builder.ai这家AI公司「全是人工,没有智能」。

这种叙事与现实之间的鸿沟将决定该公司的发展轨迹。

只有人工,没有智能

Builder.ai欺骗的迹象不仅仅出现在2019年华尔街日报的报道。

根据Reddit上多位前员工和知情人士的爆料,Builder.ai公司可能一开始就只有人工,没有智能。

多位前员工表示,管理层不可能不知道正在进行的欺诈,只是视而不见。这公司工作两年,几乎没有看到有项目交付。

而且有前员工透露Builder.ai极限压低员工工资,甚至称「给的薪资太垃圾」,并且不是AI导向,而是营销导向的公司。

一名用户在一年前就发现自己在使用Builder.ai的服务中发现很多「无法理解」之处。

包括:开发体验极差、缺少模块、代码无法使用、无法访问IDE甚至有些代码完全无法修改。

还有知情人士直接透露Builder.ai其实就是用「ai域名」来欺诈的公司。公司里雇佣了大量的低成本开发人员来「假装AI」。

清算时刻

随着时间的推移,Builder.ai内部的裂痕也一直在扩大。

据内部人士透露,该公司长期以来一直依赖夸大的营收预测和AI方面的宣传来获得融资。

庞大的全球员工队伍和耗资巨大的扩张计划,包括在东南亚和中东开拓新市场,使资金消耗率不断上升。

与此同时,前CEO的法律问题也层出不穷。

据「金融时报」报道,Duggal卷入了印度一桩洗钱刑事案件的调查。对此,Builder.ai的总法律顾问曾在一篇现已删除的博客中回应称,Duggal只是该案的一名证人。

不过,Duggal还是在2月辞去了CEO职务,但仍留在董事会并保留了他的「wizard」头衔。

接替他的,是亚马逊和Flipkart的前高管Manpreet Ratia,后者此前曾担任Builder.ai投资方Jungle Ventures的管理合伙人。

紧接着,清算的时刻就来临了。

2025年5月,Builder.ai的高级投资方之一Viola Credit从该公司账户中扣押了3700万美元,并触发违约。

仅在两个月前接手收拾残局的首席执行官Manpreet Ratia手中仅剩下500万美元现金。

几天后,他申请了破产。

事实证明,Builder.ai向贷款方提供了夸大的财务预测,谎报了其营收健康状况。

这一违反契约条款的行为让Viola Credit得以采取断然措施。

但这次结构性崩溃背后更大的原因是,他们的商业模式从未与他们的品牌宣传相匹配。

Ratia在一次全公司范围的电话会议中,承认了败局已定。大部分全球员工遭到解雇,曾被定位为AI创新旗舰的产品也被搁置。

5月20日,它正式宣告破产。

在失败之前的一个月,该公司进行了最后时刻的重组,裁掉了770名员工中的220人。

Builder.ai本周表示,由于「无法从历史挑战和过去的决策中恢复过来,这些因素给公司的财务状况带来了巨大压力」,尽管管理层「不懈努力」,但公司将任命一名行政官来监督破产程序。

据「金融时报」报道,Builder.ai总共欠了亚马逊8500万美元,欠微软3000万美元。

创业明星

为何Duggal一开始能获得投资人的青睐?不论是卡塔尔资金、软银还是微软,都不是轻易能够欺骗的。

这就不得不提Duggal「光鲜」的履历了。

Sachin Dev Duggal在14岁时开始通过组装PC电脑开启职业生涯,到17岁时,他为德意志银行创建了世界上首批自动化货币套利交易系统之一。

他在21岁仍在帝国理工学院就读期间,启动了他的下一个创业项目——云计算公司Nivio。

在离开估值为1亿美元的Nivio之后,Duggal开始专注于打造一个名为Shoto的照片分享应用。

然而,他很难找到符合自己需求的前端开发人员。Duggal不禁思考:如果连他自己都难以找到可靠的帮手,那么没有工程背景的人又该如何开始构建一款应用呢?

于是,他创立了Builder.ai,旨在让软件构建「变得像点披萨一样简单」。

后面的故事,大家也都知道了。

AI洗白

在行业上,Builder.ai这种将传统技术服务包装成AI来骗取资金的模式,被称为「AI洗白」(AI washing)。

而它的失败,也重新引发了关于在AI交易中进行技术尽职调查必要性的讨论。

对于客户而言,其中许多是初创公司和中小企业,这次突然的停运让他们手忙脚乱地重建或迁移他们的应用程序。这凸显了依赖新兴参与者提供关键任务软件基础设施的风险。

尽管遭遇了这次打击,但更广泛的低代码/无代码市场依然保持韧性。

Gartner预测,到2028年,60%的新企业应用程序将使用此类平台开发。预计到今年年底,全球市场规模将达到 260亿美元。

从Gartner的赞誉到Fast Company的排名,从明星投资者到其网站上展示的顶级公司标识,Builder.ai似乎是AI时代的伟大成功故事之一。

但像许多建立在炒作之上的公司一样,它混淆了规模与可持续性,以及知名度与生存能力。

最终,Builder.ai的故事与其说是一项失败的技术,不如说是假装它曾经奏效所带来的后果。

在 ChatGPT 带动的投资热浪里,规模、估值与曝光度并不等于护城河。

Builder.ai的故事像极了昔日Theranos——当技术承诺与实际能力出现1毫米的裂缝,资本市场就会在下一秒撕开1千米的深渊。

参考资料:

https://www.ft.com/content/926f4969-fda7-4e78-b106-4888c8704bda

https://www.financialexpress.com/business/start-ups/why-did-microsoft-backed-1-3bn-builderai-collapse-accused-of-using-indian-codersforaiwork/3854944/

Trending Cryptos

Related Reads

AI is Sweeping the Globe, So Why is Crypto + AI in a Slump?

AI Booms, But Crypto + AI Remains Sluggish: A Demand-Side Analysis Despite the AI industry's explosive growth and massive investment, the convergence of blockchain and AI (Crypto + AI) has seen limited traction. The core issue is a severe supply-demand mismatch, not a flawed premise. Analyzing four key sub-sectors reveals specific gaps: 1. **Decentralized Compute/Storage:** Offer logical benefits like data sovereignty and cost savings but lack a decisive technical advantage over entrenched cloud giants (AWS, GCP). Enterprises prioritize performance and stability and are unwilling to bear the switching risk and uncertainty of decentralized networks. 2. **Model Verification/Privacy (e.g., ZKML):** Address important long-term issues like auditability and data privacy, but these are not urgent operational pain points for most businesses today. Widespread demand will likely follow regulatory mandates (like the EU AI Act), not precede them. 3. **AI Agent Infrastructure:** Projects are building infrastructure for a future of autonomous, interacting agents. However, the current market focus is on internal process automation within corporate firewalls. The technology is ahead of market readiness. 4. **AI Agent Payments:** This is the only sub-sector where blockchain is on a level playing field with traditional finance. Both are trying to solve the unsolved problem of real-time, micro-transactions for machines, making it the most immediately competitive area. The overarching problem is that the AI industry invests heavily in solutions that solve immediate bottlenecks (e.g., faster memory, more power). Most Crypto + AI solutions target secondary, longer-term concerns (decentralization, transparency) and often come with performance trade-offs. The lack of a flagship, large-scale commercial success case further hinders mainstream capital inflow. The path forward requires either aligning more closely with the current industry's performance demands or patiently building the foundational infrastructure for the next phase of AI.

Foresight News5m ago

AI is Sweeping the Globe, So Why is Crypto + AI in a Slump?

Foresight News5m ago

Continuous Net Outflows from ETFs, Are Institutions Exiting?

US spot Bitcoin ETFs have experienced approximately $6 billion in net outflows over the past six weeks, marking the longest consecutive weekly withdrawal streak since their launch in 2024. The iShares Bitcoin Trust (IBIT) from BlackRock has been particularly affected, accounting for over 70% of recent outflows. On-chain analysis indicates that long-term Bitcoin holders (holding for over 155 days), who control about 83% of the circulating supply, remain steadfast. The selling pressure is primarily coming from allocators who entered through ETF brokerage accounts. This represents the first major collective capitulation since Bitcoin gained mainstream Wall Street recognition, driven more by risk-off portfolio adjustments than a fundamental rejection of the asset. Factors such as rising inflation, a hawkish shift in Federal Reserve policy, massive capital inflows into AI infrastructure, and attractive IPO opportunities have redirected speculative funds. Bitcoin, treated as a high-beta risk asset, was among the first to be sold. While the pace of outflows has slowed significantly—from $1.72 billion in early June to $226.8 million mid-month—the structural issue remains. IBIT's large size means its outflows alone exert substantial market pressure. With spot market volume thin, new capital inflows absent, and ETF buying muted, the market lacks sufficient buying support to absorb this selling. The coming sessions are critical. If IBIT outflows decelerate and Bitcoin reclaims $60,000, this phase could be seen as a healthy reset. However, if heavy IBIT redemptions resume and the price falls below $58,000, it would signal a more sustained institutional exit, requiring non-ETF buyers to shoulder the entire selling pressure alone. The ETF, while lowering entry barriers, has not removed Bitcoin's inherent volatility.

marsbit46m ago

Continuous Net Outflows from ETFs, Are Institutions Exiting?

marsbit46m ago

Introduction to the Concept of World Models: A Story from Psychology to the Main Battlefield of AI

**World Models: From Psychology to AI's Core Concept** "World model" is a trending but often confusing term in AI, describing a system that allows machines to internally simulate, predict, and rehearse potential outcomes before taking real-world action—like a mental "sandbox." While definitions vary—Yann LeCun emphasizes physical understanding, OpenAI's Sora is a video-based "world simulator," Google DeepMind's Genie 3 creates interactive 3D environments, and companies like Alibaba and Tesla focus on practical applications—the core goal is consistent: reduce reliance on vast real-world data by creating an internal, predictive model for safer and more efficient AI. The concept has deep roots, tracing back to psychologist Kenneth Craik (1943). In AI, it was revitalized by researchers like David Ha and Jürgen Schmidhuber (2018). Major technical approaches include: 1) generative video models (e.g., Sora) for visual realism; 2) abstract predictive models (e.g., LeCun's JEPA) for efficiency and physical reasoning; and 3) explicit 3D simulators (e.g., NVIDIA Omniverse) for precision. Fei-Fei Li proposes a classification based on the AI action loop: renderers (output observations), simulators (output world states), and planners (output actions). The emerging "World Action Model" (WAM) paradigm aims to unify future prediction and action generation. An industry framework is forming: upstream (data, compute, sensors), midstream (general and vertical platforms), and downstream applications (autonomous driving, robotics, gaming, etc.). Autonomous driving is currently the most mature use case. The current lack of a unified definition reflects the field's early, dynamic stage, similar to past tech revolutions. Different approaches—focusing on pixels, physics, or behavior—represent parallel explorations of how best to compress and understand the world. This diversity, while seemingly chaotic, signals that world models have moved from an academic idea to a critical industrial battleground, ultimately aiming to give machines the ability to understand, imagine, and reason about the world.

marsbit1h ago

Introduction to the Concept of World Models: A Story from Psychology to the Main Battlefield of AI

marsbit1h ago

Building the Bright Path While Secretly Crossing Chencang: Is Walsh Paving the Way for a September "Rate Cut"?

The title "Building the Plank Road Openly While Secretly Crossing at Chencang: Is Walsh Paving the Way for a September 'Rate Cut'?" suggests Federal Reserve Chair Kevin Walsh's hawkish stance may be a deliberate smokescreen. Academy Securities analyst Peter Tchir argues in a report that markets, currently pricing a 75% chance of a September hike, are missing a potential path to a September rate cut that Walsh himself might be quietly preparing. Tchir posits that Walsh's hawkish rhetoric aims to suppress long-term yield risks (with the 10-year Treasury yield falling recently) while creating room for a narrative shift based on upcoming data. The potential political endgame, according to this view, could be rate cuts in September and October, ahead of the midterm elections. This hinges on a political logic where the Trump administration's preference for lower rates remains unchanged. A core part of Tchir's argument involves redefining inflation metrics. He contends the Fed under Walsh may deprioritize the PCE index, criticizing its lagging components like Owners' Equivalent Rent (OER). Instead, he points to alternative, more real-time indicators like the New Tenant Repeat Rent Index (NTRR) and the Truflation daily index, which shows core inflation around 1.45%. He suggests the Fed could shift its data narrative to justify policy easing. Furthermore, Tchir downplays AI-driven inflation fears. He argues that consumer price sensitivity, evidenced by negative market reactions to price hikes (e.g., Apple), contradicts persistent inflation narratives. He also separates AI/data center spending—which he sees as relatively rate-insensitive—from broader consumer affordability issues, implying rate hikes are misdirected. Based on this analysis, Tchir sees a re-pricing of rate cut expectations as likely, creating opportunities in short-duration Treasuries. He maintains a neutral-to-slightly-bullish view on the long end of the yield curve. For equities, he recommends a significant overweight in energy (especially global nuclear assets) and, within defense/security themes, an overweight in biotech/pharma versus an underweight in semiconductors, expressing caution on AI/data center valuations.

marsbit1h ago

Building the Bright Path While Secretly Crossing Chencang: Is Walsh Paving the Way for a September "Rate Cut"?

marsbit1h ago

Trading

Spot

Hot Articles

Discussions

Welcome to the HTX Community. Here, you can stay informed about the latest platform developments and gain access to professional market insights. Users' opinions on the price of AI (AI) are presented below.

活动图片