字号·· | 护眼
thenextweb

CarbonAI融资1200万美元,用AI代理实现商业房地产承销自动化CarbonAI raises $12M to automate commercial real estate underwriting with AI agents

点「原文对照」整页切到原文,或双击某段只看那段的原文。

图片来源:Igor Omilaev,UnsplashCredit: Igor Omilaev on Unsplash

CarbonAI 获得 1200 万美元融资,由 Cowboy Ventures 领投,用于开发处理商业房地产贷款的承销、尽职调查和贷后监控的 AI 代理。欧盟《人工智能法案》将消费者信贷决策视为高风险,但不包括企业贷款;当赞助人提供个人担保时,这一界限变得模糊。

CarbonAI has raised $12M led by Cowboy Ventures for AI agents that handle underwriting, due diligence and post-close monitoring on commercial real estate loans. The EU AI Act treats consumer credit decisions as high-risk but not corporate lending, a line that blurs when a sponsor gives a personal guarantee.

CarbonAI 获得 1200 万美元融资,用于开发处理商业房地产贷款分析工作的 AI 代理。该公司总部位于纽约,仅在美国销售。在欧洲,同样的软件将不受《人工智能法案》信贷规则的约束,直至赞助人签署个人担保。

CarbonAI has raised $12M for AI agents that do the analytical work behind commercial property loans. The New York company sells only in the US. In Europe the same software would sit outside the AI Act’s credit rules, until a sponsor signs a personal guarantee.

根据公告,Cowboy Ventures 领投此轮融资,参投方包括 Timber Grove Ventures、Bling Capital 和 Max Ventures。未披露融资阶段或估值。

Cowboy Ventures led the round, according to the announcement, with Timber Grove Ventures, Bling Capital and Max Ventures. Neither a stage nor a valuation was disclosed.

这些代理从租金清单、租赁合同、财务报表和法律文件中提取数据,构建承销模型,起草备忘录并标记风险。每个数据均可追溯至源文件。需要人工判断的决策将转交给贷款团队。

The agents pull figures from rent rolls, leases, financial statements and legal documents, build the underwriting model, draft the memo and flag risks. Each number links back to its source document. Decisions needing human judgment go to the lending team.

联合创始人 Jeff Saul 和 Josh Kaufman 曾在 Starwood Capital、摩根士丹利和古根海姆从事商业房地产贷款承销工作,随后打造了交易平台 Nativ 并将其出售给 Built。Greystone 是目前唯一被命名的客户。

Co-founders Jeff Saul and Josh Kaufman underwrote commercial property loans at Starwood Capital, Morgan Stanley and Guggenheim, then built the deal platform Nativ and sold it to Built. Greystone is the only customer named.

在欧洲,房贷承销代理被视为高风险。Vesta 上周获得 3000 万美元融资,用于开发做出房贷决策的 AI 代理,而欧盟对此类系统的监管方式截然不同。

Mortgage underwriting agents are high-risk in Europe. Vesta raised $30M last week for agents that make mortgage decisions, work the EU treats very differently.

《人工智能法案》附件 III 覆盖用于评估“自然 persons 的信用状况”或设定其信用评分的系统,其中仅有欺诈检测一个例外。公司不属于自然 persons。

Annex III of the AI Act covers systems used to evaluate “the creditworthiness of natural persons” or set their credit score, with one carve-out for fraud detection. Companies are not natural persons.

柏林的 Taktile 在 6 月获得 1.1 亿美元融资,用于开发承销企业贷款的 AI 代理,因同样原因而不受该条款约束。

Berlin’s Taktile raised $110M in June for agents that underwrite business loans, and sits outside the same provision for the same reason.

CarbonAI 表示,其客户包括债务基金、抵押房地产投资信托和机构多户住宅贷款人。其借款人均为企业。

CarbonAI says its customers include debt funds, mortgage real estate investment trusts and agency multifamily lenders. Its borrowers are companies.

这条线路并不干净。商业地产贷款通常会附带发起人的个人担保,一旦对个人的信用状况进行评估,该条款就会立即生效。这些义务被推迟到了2027年12月。

The line is not clean. Commercial property loans routinely carry personal guarantees from their sponsors, and the provision bites the moment an individual’s creditworthiness is assessed. Those duties were pushed to December 2027.

该公告引用了抵押银行协会的数据,称美国商业地产市场规模为20万亿美元,而今年到期的抵押贷款规模为8750亿美元,相比之下,每年用于承销这些贷款的支出仅约为500亿美元。

The announcement cites a $20 trillion US commercial property market and $875B of mortgages maturing this year, on Mortgage Bankers Association figures, against roughly $50B a year spent underwriting them.

欧洲的市场并不处于这种状况。瑞银的房地产贷款负责人本月离职,距其受雇来建立该业务尚不足两年。

Europe’s market is not in that condition. Revolut’s property lending chief left this month, less than two years after being hired to build the business.

由于利率仍然居高不下,当地的贷款机构正在争夺更少的交易,彭博社报道称,这正在压缩他们的利润空间。

Lenders there are competing for fewer deals as rates stay high, Bloomberg reported, which is squeezing their margins.

CarbonAI的内部数据是值得关注的。该公司称,其在六个月内实现了七位数的合同收入,并且在选定的承销任务上节省了高达90%的时间,但未给出具体方法论。一家名为Build的英国公司,专注于为数据中心提供类似服务,声称其效率提升超过95%。

CarbonAI’s own figures are the ones to watch. It reports seven-figure contracted revenue within six months, and up to 90% of time saved on selected underwriting tasks, with no methodology given. Build, a British-founded firm doing the same for data centres, claims over 95%.

阅读媒体原文 ↗