Gravity 融资3050万美元以建立面向 AI 代理的广告交易所

Why it matters

Gravity is betting that AI assistants become both advertising inventory and buyers. Owning demand, supply and the exchange could give Gravity leverage if platforms accept ads built from conversation context.

Gravity, founded by Zach Oldham (@zachtheoldham) and Leo Martinez (@leojamartinez), has raised a $30.5 million Series A to build advertising infrastructure for AI chatbots and software agents. Business Insider reported on August 6th that Lightspeed Venture Partners and Committed Capital co-led the round, bringing Gravity's total funding to $38.5 million. (businessinsider.com)

该公司由 Zach Oldham (@zachtheoldham) 和 Leo Martinez (@leojamartinez) 创立，已完成3050万美元的A轮融资，旨在为AI聊天机器人和软件代理构建广告基础设施。Business Insider 报道，Lightspeed Venture Partners 和 Committed Capital 共同领投本轮融资，使 Gravity 的总融资额达到3850万美元。(businessinsider.com)

The financing backs an attempt to establish Gravity as the intermediary between brands seeking space inside AI products and developers looking for an alternative to subscriptions. Gravity combines a demand-side platform for advertisers, a supply-side platform for AI developers and the exchange that runs between them. The San Francisco company also sells campaign creation, measurement and conversion-tracking tools. (businessinsider.com)

这笔融资支持了 Gravity 成为品牌（寻求在 AI 产品中获得展示位）与开发者（寻找订阅以外替代方案）之间中介的尝试。Gravity 将广告主的需求方平台、AI 开发者的供给方平台以及二者之间运行的交易所结合在一起。这家旧金山公司还出售活动创建、衡量和转化跟踪工具。(businessinsider.com)

That full-stack structure is central to Oldham's pitch. A company controlling advertiser demand and developer inventory can gather performance data from both ends of each transaction, improve its matching system and make itself harder to replace as the network grows. Gravity still has to compete for both sides against established advertising platforms and new products from companies already embedded in marketers' budgets.

这种全栈结构是 Oldham 推介的核心。控制广告主需求和开发者库存的公司能够从每次交易的双方收集性能数据，改进其匹配系统，并随着网络增长使自身更难被替代。Gravity 仍需在两端与已建立的广告平台以及那些已被纳入营销预算的公司推出的新产品竞争。

According to Business Insider, Gravity has placed text ads in AI assistants and chatbots including ChatGPT, Codebuff, EcoGPT, Daimon, Magneta, Runable and CTO.new. Advertisers using Gravity have included Best Buy, Target, Vercel and MongoDB. Those names establish early access to recognizable brands, though Gravity has not published revenue, campaign spending or conversion figures that would show how large those relationships are. (businessinsider.com)

据 Business Insider 报道，Gravity 已在包括 ChatGPT、Codebuff、EcoGPT、Daimon、Magneta、Runable 和 CTO.new 等 AI 助手和聊天机器人中投放文本广告。使用 Gravity 的广告主包括 Best Buy、Target、Vercel 和 MongoDB。这些名字表明其获得了可识别品牌的早期接入，尽管 Gravity 尚未公布可显示这些关系规模的收入、活动支出或转化数据。(businessinsider.com)

From ad measurement to AI conversations

Oldham previously founded Flax Labs, an ecommerce growth agency built around software that attributed profit to individual ads. Flax Labs says he founded the business in 2020. On his personal site, Oldham writes that he bootstrapped Flax Labs into a 45-person services company and brand incubator before starting Gravity. (zachtheoldham.com)

从广告测量到 AI 对话

Oldham 此前创立了 Flax Labs，一家围绕将利润归因于单个广告的软件构建的电子商务增长机构。Flax Labs 表示 他于2020年创立了该业务。在他的个人网站上，Oldham 写道他在创立 Gravity 之前，自主将 Flax Labs 做成了一个拥有45人规模的服务公司和品牌孵化器。(zachtheoldham.com)

That background explains Gravity's focus on the machinery behind advertising rather than a single consumer-facing ad format. The company's developer documentation offers SDK and API integrations for AI chats, coding assistants, search products and other applications. Gravity says an ad request runs alongside the application's model call, with its system matching an advertiser to the conversation, running an auction and returning creative for the developer to display. (trygravity.ai)

这一背景解释了 Gravity 更关注广告背后的机制而非单一面向消费者的广告形式。该公司的开发者文档为 AI 聊天、代码助手、搜索产品及其他应用提供 SDK 和 API 集成。Gravity 表示，广告请求与应用的模型调用并行运行，其系统将广告主与对话进行匹配、执行竞价并返回供开发者展示的创意素材。(trygravity.ai)

Gravity publicly bills advertisers on a cost-per-thousand-impressions model. Its technical documentation says the matching engine embeds recent messages, retrieves relevant campaigns, holds an auction and generates creative tailored to the conversation. The same documentation says richer contextual signals improve matching and attribution, making the handling of conversation data a core product issue as Gravity tries to persuade larger platforms to adopt its system. Gravity says its tracking pixels are first-party and do not share data across unrelated sites. (docs.trygravity.ai)

Gravity 对广告主公开以按千次展示成本计费。其技术文档称，匹配引擎会将最近消息进行嵌入，检索相关活动，进行竞价并生成针对对话定制的创意。同一文档还指出，更丰富的上下文信号可以改善匹配和归因，这使得对话数据的处理成为 Gravity 在尝试说服更大平台采用其系统时的核心产品问题。Gravity 表示其跟踪像素为第一方，不会在无关网站之间共享数据。(docs.trygravity.ai)

Martinez, Gravity's CTO, is responsible for the technical side of making those auctions fast and inexpensive enough to run inside a live conversation. In a customer case study, Inference.net said Gravity replaced a general-purpose 70-billion-parameter model with a specialized 1-billion-parameter model. Inference.net claimed the change cut inference costs by roughly 90% and reduced 99th-percentile latency from nearly two seconds to 351 milliseconds while serving millions of requests per day. Those performance figures come from Gravity's infrastructure provider rather than an independent benchmark. (inference.net)

Gravity 的首席技术官 Martinez 负责使这些竞价在实时对话中运行得既快速又低成本的技术方面。在一份客户案例研究中，Inference.net 表示 Gravity 将一个通用的700亿参数模型替换为一个专用的10亿参数模型。Inference.net 声称，此更改将推理成本大约降低了90%，并将第99百分位延迟从接近两秒减少到351毫秒，同时每天处理数百万次请求。这些性能数据来自 Gravity 的基础设施提供方，而非独立基准测试。(inference.net)

Ads aimed at agents

Gravity is also testing what Oldham calls "agent-to-agent" advertising. Instead of displaying a sponsored message to a person, Gravity supplies a purchasing agent with product catalogs, features and current promotions when the agent is evaluating what to recommend. A request for shoes, for example, could prompt Gravity to provide the agent with relevant inventory from participating advertisers. (businessinsider.com)

面向代理的广告

Gravity 还在测试 Oldham 所称的 “agent-to-agent” 广告。Gravity 并非向人展示赞助信息，而是在代理评估推荐内容时，为采购代理提供产品目录、功能和当前促销。例如，对鞋子的请求可能会促使 Gravity 向代理提供来自参与广告主的相关库存。(businessinsider.com)

The planned next step is transaction execution. Gravity wants to add payment integrations that let an agent complete a purchase after receiving user approval. That would move Gravity deeper into the transaction, connecting product discovery, advertising attribution and checkout rather than stopping after an impression or referral. It would also make disclosure and ranking rules consequential: an agent's recommendation may be shaped by paid inventory even when the user never encounters a conventional ad unit.

下一步计划是执行交易。Gravity 希望添加支付集成，使代理在获得用户批准后能够完成购买。这将使 Gravity 更深入介入交易，将产品发现、广告归因与结账连接起来，而不仅止于展示或引荐。它还将使披露和排序规则变得有实际影响：即便用户从未遇到传统广告单元，代理的推荐也可能受付费库存的影响。

Oldham argues advertising can subsidize AI services and reduce the need for paid subscriptions. He has described Gravity's goal as "making intelligence free," applying the commercial model that funded search and social media to AI products. Gravity's current engineering job listing says the company serves millions of ads per day and identifies Caffeinated Capital and Basis Set Ventures as existing backers. Both operating scale and investor information on that page are company-supplied. (zachtheoldham.com)

Oldham 认为广告可以补贴 AI 服务，减少对付费订阅的依赖。他将 Gravity 的目标描述为“让智能变得免费”，将支撑搜索和社交媒体的商业模型应用于 AI 产品。Gravity 的当前工程岗位招聘信息称公司每天投放数百万条广告，并列出了 Caffeinated Capital 和 Basis Set Ventures 作为现有投资方。该页面上的运营规模和投资者信息均由公司提供。(zachtheoldham.com)

The market Gravity is pursuing is expected to attract much larger competitors. WPP Media's 2026 midyear forecast projects generative search advertising revenue will exceed $100 billion globally by 2030. Gravity's wager is that independent infrastructure can capture part of that spending before the largest AI platforms consolidate buying, measurement and inventory within their own systems. (cdn.wpp.com)

Gravity 所追逐的市场预计将吸引更大规模的竞争者。WPP Media 的2026年年中预测预计，到2030年，生成式搜索广告收入将在全球超过1000亿美元。Gravity 的赌注是，独立基础设施可以在最大 AI 平台在自身系统内整合采购、衡量和库存之前，占据其中一部分开支。(cdn.wpp.com)