• 亿邦公众号
    • APP下载

AI记忆初创公司Engram完成9800万美元融资 主打削减token成本

作者:亿邦AI 文章来源:亿邦动力

2026-06-24 15:25

作者:亿邦AI 文章来源:亿邦动力

2026-06-24 15:25

邦小白快读

EN
全文速览

本文核心是介绍成立仅8个月的AI记忆初创公司Engram完成9800万美元融资的相关信息,核心干货整理如下:

1. 产品核心能力:Engram将自身定位为AI的习得性记忆,模型可记忆企业专属工作流与上下文,预判用户需求输出精准响应,官方数据显示其效果可匹配甚至超越前沿实验室产品,最高可将AI运行所需的计价单位token数量减少100倍,能大幅降低AI使用成本。

2. 团队与运营现状:这家公司仅13名员工,CEO有哥伦比亚大学计算神经科学博士学位,长期研究记忆领域,成立不足一年就拿下了微软、Notion等知名客户,本次融资资金将用于算力支持与人才招募。

本文对AI相关品牌的定位和发展有不少参考干货,整理如下:

1. 当前行业需求趋势:新一代复杂度更高的AI模型成本不降反升,打破了规模扩大会带动成本下降的行业共识,当前数据量和AI成本都处于爆发阶段,企业用户对AI降本有极强的需求,这是明确的市场机会。

2. 可参考的品牌定位策略:Engram走差异化垂直专精路线,没有定位成全能AI厂商,而是明确承认放弃部分通用能力,聚焦AI记忆和降本这个细分领域,已经获得顶级资本和头部客户认可,这种差异化定位对新品牌打造很有借鉴价值。

3. 产品研发方向:AI记忆层构建、降本是当前明确的研发方向,可重点布局。

本文透露了AI服务领域的最新机会和可借鉴的经验,干货整理如下:

1. 明确的市场机会:当前AI行业成本痛点突出,新一代大模型成本高于前代,数据量和AI成本同步爆发,所有企业应用AI都面临成本压力,主打AI降本的垂直服务有广阔的增长空间。

2. 可学习的商业模式:小团队不需要和OpenAI等大厂正面竞争,走差异化垂直路线,聚焦细分痛点做深做透,8个月就能拿到顶级资本融资,拿下头部客户,这种轻资产快速验证的模式非常适合中小创业者。

3. 风险提示:走垂直路线需要主动放弃部分通用能力,要明确自身优势边界,不要盲目扩张赛道,聚焦自身核心优势才能快速发展。

本文对工厂推进AI数字化转型有不少启示,干货整理如下:

1. 工厂引入AI的核心注意点:当前新一代AI模型复杂度提升后,成本并没有随着规模扩大下降,反而比前代更高,工厂推进数字化转型引入AI的时候,不要盲目追求大参数通用模型,要重点关注落地成本问题。

2. 可挖掘的商业机会:AI在垂直领域的成本优化空间很大,面向制造企业的AI服务,围绕生产流程的记忆梳理、上下文优化降本有很大需求,就算是工厂自身做数字化升级,也可以关注这个方向的方案,提升落地性价比。

3. 转型启示:工厂数字化不需要追求大而全,针对自身生产、设计、管理的具体痛点,引入专业化的垂直AI方案,就能实现不错的降本增效效果,不用跟风追逐通用大模型。

本文给AI相关服务商梳理了行业趋势和可参考的方案,干货整理如下:

1. 行业发展趋势:当前AI行业数据量和AI运行成本都进入爆发阶段,新一代复杂大模型成本高于前代,打破了规模降本的行业传统认知,企业侧对AI降本的需求十分迫切,这是服务商接下来的核心增长机会。

2. 客户核心痛点:现有主流大模型存在记忆局限,过多上下文输入会提升模型负载,直接推高运行成本,企业需要能适配自身专属工作流的AI记忆方案来控制成本。

3. 可参考的解决方案:Engram的习得性记忆方案,通过梳理企业整体流程,最高可将token需求降低100倍,效果能匹配前沿产品,走垂直专精的差异化路线,不跟大厂拼通用能力,这个模式非常适合中小AI服务商参考。

本文给AI平台的生态布局和运营带来不少参考,干货整理如下:

1. 当前市场对平台的核心需求:AI开发者和企业用户都普遍面临AI运行成本过高的痛点,对能降低token成本、优化AI记忆能力的垂直服务商需求强烈,平台需要围绕这个需求调整生态布局。

2. 平台可调整的运营做法:可以针对性招商引入类似Engram这类主打AI降本的垂直初创企业,丰富平台AI服务品类,满足企业用户降本的核心需求,形成和大厂通用大模型平台差异化的竞争优势。

3. 风向规避提示:当前AI行业并不是参数越大、越通用就越好,垂直细分的成本优化方案已经得到资本和大客户的认可,平台布局要避免盲目追逐大而全的通用项目,重视细分赛道的优质创业项目,规避资源错配的风险。

本文展现了当前AI产业的新动向和新范式,对产业研究有较高价值,干货整理如下:

1. 产业新动向:AI大模型产业的细分赛道开始崛起,聚焦AI记忆、降本的垂直细分方向已经获得顶级资本认可,成立仅8个月的13人小团队就能拿到9800万美元融资,还拿下微软等头部客户,说明垂直细分赛道已经成为AI产业的新增长极。

2. 产业新问题:新一代复杂度更高的大模型实际成本高于前代,打破了行业一直以来“规模扩大带动成本下降”的传统认知,数据量和AI成本同步爆发已经成为产业发展的新痛点,这是值得深入研究的新课题。

3. 新商业模式:差异化垂直专精的创业范式已经得到验证,创业者不需要追求全面超越大厂产品,主动放弃部分通用能力,聚焦细分领域做深做强,就能实现快速成长,这种新的创业模式值得深入研究。

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article covers the $98 million funding round recently closed by Engram, an AI memory startup founded just 8 months ago. Key takeaways are as follows:

1. Core product capability: Engram positions itself as "learnable memory" for AI systems. Its model can remember an enterprise's exclusive workflows and context, anticipate user needs, and deliver highly accurate responses. Official data shows its performance matches or even outperforms products built by cutting-edge AI labs; it can reduce token consumption — the unit that AI services are priced on — by up to 100x, significantly cutting AI adoption costs for businesses.

2. Team and current operations: The company currently has only 13 employees. Its CEO holds a PhD in computational neuroscience from Columbia University and has long focused on memory research. Less than one year after its founding, Engram has already secured well-known clients including Microsoft and Notion. The new funding will be used to expand computing capacity and recruit new talent.

This article shares multiple actionable insights for AI brand positioning and growth, summarized below:

1. Current industry demand trend: The cost of more complex new-generation AI models has actually increased rather than decreased, breaking the long-held industry consensus that scaling would bring cost reductions. Today, both data volume and AI costs are growing exponentially, and enterprise clients have extremely strong demand for AI cost reduction — this is a clear market opportunity.

2. A referenceable brand positioning strategy: Engram adopts a differentiated, niche-focused strategy. Instead of positioning itself as a full-stack general AI provider, it openly chooses to forfeit some general capabilities to focus exclusively on the AI memory and cost reduction niche. This positioning has already earned recognition from top-tier investors and leading clients, making it a highly valuable model for new AI brands to emulate.

3. Product R&D direction: Building AI memory layers and delivering cost reduction are clear high-potential R&D directions worth prioritizing for investment.

This article reveals the latest opportunities and replicable lessons in the AI services space, summarized below:

1. Clear market opportunity: Cost pressure is a prominent pain point in today's AI industry. New-generation large models are more costly than their predecessors, and data volume and AI costs are growing in lockstep. All enterprises adopting AI face cost pressure, so vertical services focused on AI cost reduction have enormous room for growth.

2. A learnable business model: Small teams do not need to compete head-on with industry giants like OpenAI. By pursuing a differentiated vertical strategy and deepening focus on a narrow pain point, startups can secure top-tier venture funding and win leading clients in as little as 8 months. This asset-light, rapid validation model is extremely suitable for small and medium-sized entrepreneurs.

3. Risk reminder: Pursuing a vertical strategy requires proactively giving up some general capabilities. Founders must clarify the boundaries of their core strengths, avoid blindly expanding into unrelated tracks, and stay focused to achieve rapid growth.

This article offers key insights for factories advancing AI-driven digital transformation, summarized below:

1. Core consideration when adopting AI: After the complexity increase of new-generation AI models, costs have not fallen with scale — they are actually higher than previous generations. When introducing AI for digital transformation, factories should not blindly pursue large-parameter general-purpose models. Instead, they should prioritize implementation cost in decision-making.

2. Underexplored business opportunities: There is huge room for AI cost optimization in vertical sectors. For AI services targeting manufacturing, there is massive unmet demand for production workflow memory organization and context-based cost optimization. Even factories conducting in-house digital upgrades can explore solutions in this area to improve cost-effectiveness.

3. Transformation insight: Factory digital transformation does not require a one-size-fits-all, full-scope overhaul. By introducing specialized vertical AI solutions targeted at specific pain points in production, design and management, factories can achieve meaningful cost reduction and efficiency improvement without following the hype around general large models.

This article outlines industry trends and referenceable solutions for AI service providers, summarized below:

1. Industry development trend: Today, both data volume and AI operating costs are growing exponentially. New-generation complex large models cost more than their predecessors, breaking the traditional industry assumption that scaling reduces costs. Enterprises have an extremely urgent demand for AI cost reduction, which is the core growth opportunity for service providers going forward.

2. Core client pain points: Existing mainstream large models have inherent memory limitations. Excess context input increases model load and directly drives up operating costs. Enterprises need AI memory solutions adapted to their exclusive workflows to control costs.

3. A referenceable solution: Engram's learnable memory solution organizes an enterprise's end-to-end workflows to cut token demand by up to 100x, with performance matching cutting-edge industry products. Its strategy of vertical specialization and differentiation, competing with giants on general capabilities, is extremely suitable for small and medium-sized AI service providers to emulate.

This article offers useful reference for ecosystem layout and operations of AI platforms, summarized below:

1. Core market demand for platforms: AI developers and enterprise users widely face the pain point of excessively high AI operating costs. There is strong demand for vertical service providers that reduce token costs and improve AI memory capabilities, and platforms should adjust their ecosystem layout around this demand.

2. Adjustable operational strategies for platforms: Platforms can proactively recruit vertical startups like Engram that focus on AI cost reduction, enrich their AI service offerings to meet core enterprise demand for cost cutting, and build a differentiated competitive advantage against giant general large model platforms.

3. Risk guidance: Bigger and more general is not necessarily better in the current AI industry. Vertical, niche cost optimization solutions have already won recognition from both capital and major clients. Platforms should avoid blindly pursuing large-scale general projects, prioritize high-quality startups in niche tracks, and mitigate the risk of resource misallocation.

This article presents new trends and new paradigms in the current AI industry, with high value for industrial research, summarized below:

1. New industrial trends: Niche tracks in the AI large model industry are starting to emerge. Vertical directions focused on AI memory and cost reduction have already gained recognition from top-tier capital. A 13-person startup founded just 8 months ago can raise $98 million and win leading clients like Microsoft, proving that vertical niche tracks have become new growth poles for the AI industry.

2. New industrial questions: The actual cost of new-generation more complex large models is higher than that of previous generations, breaking the long-held industry assumption that "scaling drives cost reduction". The simultaneous explosion of data volume and AI costs has become a new pain point for industrial development, which is a new topic worthy of in-depth research.

3. New business models: The differentiated, vertically specialized startup paradigm has been validated. Founders do not need to pursue outperforming giant products across the board; by proactively giving up some general capabilities and focusing on deepening strengths in a niche area, they can achieve rapid growth. This new startup model is worthy of further research.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

2026年6月23日,成立仅8个月的AI记忆初创公司Engram公布最新融资信息,总融资金额达9800万美元,投资方涵盖General Catalyst、Kleiner Perkins、红杉,还包括近期加入Anthropic的OpenAI联合创始人Andrej Karpathy。

Engram将自身定位为AI的习得性记忆,其模型可记忆企业专属工作流与上下文,预判用户问题并输出成本更低的精准响应。官方披露的数据显示,其模型效果可匹配甚至超越前沿实验室产品,运行所需token数量最高可减少100倍。token是运行AI查询时的计价单位。

当前新一代复杂度更高的AI模型,实际成本高于前代产品,打破了规模扩大将带动成本下降的行业普遍认知。Kleiner合伙人Leigh Marie Braswell的公开表述显示,当前数据量与相关AI成本均处于爆发阶段,Engram的方案可梳理企业的整体流程,将输出成本降低数个量级。

成立不足一年,这家仅13名员工的公司已经积累了包含微软、Notion、法律AI初创公司Harvey在内的客户名单。本次融资所得资金,将用于算力支持与人才招募。Engram的名称来源于神经科学术语,指代大脑中储存记忆的痕迹。

Engram联合创始人兼CEO Dan Biderman对记忆领域有长期研究积累,童年时期他就尝试帮助失忆的祖母回忆与家人相关的细节,后续他在哥伦比亚大学取得计算神经科学博士学位,后进入斯坦福大学AI实验室工作。在斯坦福期间,他发现现有AI模型存在记忆局限,过多上下文输入会提升模型负载,带来更高的运行成本。

Dan Biderman的公开表述显示,Engram的模型并非全面优于OpenAI、Anthropic等厂商的产品,其优势集中在垂直专精领域,部分通用能力会相应做出取舍。团队研发目标是在现有笔记类工具基础上,构建现有AI模型不具备的、类似人类的直觉层。

【版权提示】亿邦动力倡导尊重与保护知识产权。未经许可,任何人不得复制、转载、或以其他方式使用本网站的内容。如发现本站文章存在版权问题,烦请提供版权疑问、身份证明、版权证明、联系方式等发邮件至run@ebrun.com,我们将及时沟通与处理。

已点赞

+1

{{ praise_text }}

好文章,值得被更多人看到

广告

下一篇

黄小卫披露麒盛科技出海铁律:美国市场一把手全用本地人

用长期主义的办法,回答一个企业如何活得久、走得远这一终极问题。

跨境电商

2026-06-24 14:24:29