system
Integration, uptime, service. The unglamorous layer that decides whether a deployment survives year two.
集成、在线率与服务。决定一次部署能否活过第二年的不起眼的一层。
Bottlenecks瓶颈 P17 Integration, uptime, service集成、在线率与服务
What it isIntegration is the systems layer that turns a working robot into a working deployment: fitting the hardware into a customer's existing workflow, holding manufacturing yield (the share of units that come off the line actually working) as output scales up, keeping fleets online, and servicing them after the sale. It's the unglamorous work that decides whether a deployment survives past its pilot.
是什么“集成”（Integration）是把一台能跑的机器人变成一次真正部署的系统层：将硬件嵌入客户现有的工作流、在规模化生产中守住制造良率、让车队保持在线，并在交付后持续提供服务——这是决定一次部署能否活过试点期的、不起眼的一层。
ThesisOn CE-N, P17, Integration, uptime, service ranks 3rd of 18 bottlenecks, one of the strongest positions on the board, just behind Work. This is the layer where a workflow that looks profitable on paper either turns into a real annuity (steady recurring revenue) or quietly becomes a cost centre (a division that keeps costing money without paying for itself), and the two public proxies this taxonomy tracks sit at opposite ends of that outcome. Teradyne, maker of Universal Robots' collaborative arms and MiR's autonomous mobile robots, is the live test of the cost-centre case: FY2025 robotics revenue $308m against a $99m pre-tax loss. Intuitive Surgical, maker of the da Vinci and Ion surgical robots, is the answer if it converges toward the annuity case instead: consumables $1.73bn vs systems $685m.
论点在 CE-N 上，P17，集成、在线率与服务位列 18 个瓶颈中第 3，仅次于“工作”——这正是一条在纸面上算得通的工作流，最终要么在这一层变成真正的经常性收入，要么悄悄沦为成本中心的地方；而本表追踪的两个上市代理公司，恰好分处这一结果的两端。泰瑞达是成本中心这一情形的现实试金石：2025 财年机器人业务营收 3.08 亿美元，税前亏损 9900 万美元。直觉外科则展示了如果这条路转向经常性收入模式会是什么样子——耗材收入 17.3 亿美元，对比系统收入 6.85 亿美元。
Products mapped to this system映射到本系统的产品
Companies whose shipped product is a complete humanoid, quadruped, or other general-purpose embodied robot -- the OEM layer that integrates Muscle, Senses, Brain, Heart and Nerves into one machine, per the conference doc's 'Humanoid / general embodied OEMs' category.
17 companies 17 家公司 5 manufacturers 5 家制造商 +10 named in conference materials, unverified +10 项来自会议材料，未经核实
Mobile-robot, AMR, and diversified industrial-robot/servo OEMs whose portfolio mixes arms, mobile bases, and research or education platforms rather than a single complete humanoid, per the conference doc's 'General robot / research platforms' category.
Industrial Robot & Integration工业机器人与系统集成 · Servo Motors & Drives伺服电机与驱动
10 companies 10 家公司 6 manufacturers 6 家制造商 +16 named in conference materials, unverified +16 项来自会议材料，未经核实
Companies whose entire shipped product is the manipulator arm itself -- research, collaborative, or teleoperated surgical arms -- rather than a mobile base or full humanoid, per the conference doc's 'Complete manipulator platforms' category.
Industrial Robot & Integration工业机器人与系统集成
3 companies 3 家公司 +5 named in conference materials, unverified +5 项来自会议材料，未经核实
Contract manufacturers, EMS/factory-automation integrators, and turnkey production-line engineers that assemble, integrate, and service complete robot deployments rather than designing the robot itself, per the conference doc's 'Systems engineering / deployment' category.
2 companies 2 家公司 4 manufacturers 4 家制造商 +5 named in conference materials, unverified +5 项来自会议材料，未经核实
Companies mapped to this system映射到本系统的公司
Manufacture mapped to this system映射到本系统的制造
Teradyne Robotics is the live test of the cost-centre reading. Its Q2 2026 revenue of $100m, up 33% year over year, was the division’s first-ever $100m quarter and its fifth consecutive quarter of growth — but that growth arrived in the same quarter Teradyne’s Semiconductor Test segment grew 128%, and Robotics fell from 19% of group revenue in Q4 2023 to just 8% in Q2 2026 while growing.
Revenue actually declined from a 2022 peak of $326m to $293m in 2024 before recovering to $308m in 2025, and getting there took roughly a 24% cut to the division’s headcount across two rounds of layoffs. Teradyne has never disclosed the segment’s pre-tax income or its service/attach revenue, and did not discuss robotics profitability at all on the Q2 2026 call — the single most decision-relevant number on this beat sits in a 10-Q segment note that public search tools could not retrieve.
Intuitive Surgical is the counter-case: proof that an integration-and-service model can compound. Q2 2026 revenue was $2.89bn, up 19%, with instruments and accessories at $1.73bn against systems revenue of $685m — a 2.53x consumables-to-systems ratio that Intuitive discloses as a line item every quarter, precisely because the ratio is the moat.
And the related-party pattern already visible on Work shows up here too: Schaeffler is simultaneously an equity investor in Agility Robotics and Agility’s flagship deployment site; Foxconn leads Agility’s PIPE while Foxconn Industrial Internet is itself a contract-manufacturing node on this site’s own map.
Three independently commissioned analyst surveys (Wells Fargo, BTIG, China Merchants Securities) converge on the same finding about Intuitive’s moat: it holds roughly 81% of the global soft-tissue surgical-robotics installed base, and around three-quarters of surgeons on the existing platform are pursuing the next-generation upgrade rather than switching away — an independent confirmation of the consumables-ratio evidence above, not a restatement of it.
That said, sell-side coverage also surfaces three low-cost challengers gaining real if small traction: MedBot’s Toumai (23 units installed, 86 in-hand orders, J.P. Morgan’s China top pick), SS Innovations’ Mantra (revenue ramping from $5.9m to $43m over 2023-2025 at roughly half da Vinci’s price), and Medtronic’s Hugo — at a fraction of da Vinci’s RMB18-30m list price.
On the contract-manufacturing side of Integration, Morgan Stanley names Foxconn Industrial Internet the leading AI-server assembler (over 40% supply share, roughly 48% LTM revenue growth) and an explicit proxy for robot-body assembly capacity — the same Foxconn already flagged above as a related party to Agility.
And in a rare counter-example to the Teradyne cost-centre reading, Morgan Stanley upgraded KION Group to Overweight specifically on warehouse-automation attach revenue, a genuine instance of an integrator being re-rated up on service rather than down on cyclicality — though KION’s business is structurally closer to Symbotic’s than to Teradyne’s, and shouldn’t be over-generalised into a verdict on the whole system.
“Integration” is not one business model, it is two, and the public proxies span the whole distance between them. Teradyne looks like a company that needed a 24% headcount cut to hold its ground through a cycle — the profile of a cost centre, not an annuity. Intuitive looks like the annuity: it discloses the ratio that proves it, every quarter, and three independent sell-side surveys corroborate that the moat is holding.
Before sizing any private P17, Integration, uptime, service name, ask which of these two shapes it actually resembles — and separately, whether its named customers are arm’s-length.
A material share of what reads as third-party integration demand in the private book, from Agility’s Schaeffler and Foxconn relationships to whatever KION’s warehouse-automation attach turns out to be worth, may be the same balance sheet paying itself or a genuinely different, service-driven business — the two are easy to conflate and worth pulling apart name by name.
泰瑞达机器人（Teradyne Robotics）是”成本中心论”的现实检验对象。其 2026 年第二季度营收 1 亿美元，同比增长 33%，是该部门历史上首个破亿美元的季度，也是连续第五个增长季度——但这一增长恰逢泰瑞达半导体测试部门当季增长 128%，机器人业务占集团总营收的比例反而从 2023 年第四季度的 19% 降至 2026 年第二季度的 8%（增长的同时占比下降）。
营收实际上从 2022 年峰值的 3.26 亿美元下滑至 2024 年的 2.93 亿美元，之后才回升至 2025 年的 3.08 亿美元，而实现这一回升，靠的是两轮裁员、累计约 24% 的部门人员削减。泰瑞达从未披露该分部的税前利润或服务/附加收入，在 2026 年第二季度财报电话会上也完全没有讨论机器人业务的盈利能力——这条赛道上最具决策价值的数字，就藏在一份公开检索工具无法调取的 10-Q 分部附注里。
直觉外科（Intuitive Surgical）是反例：证明”集成加服务”模式确实可以产生复利。2026 年第二季度营收 28.9 亿美元，同比增长 19%，其中耗材与配件收入 17.3 亿美元，对比系统销售收入 6.85 亿美元——耗材与系统之比达 2.53 倍，直觉外科每个季度都把这一比率作为单独条目披露，正因为这个比率本身就是护城河。
上文“工作”系统中已经出现的关联方模式，在这里再次出现：Schaeffler 同时是 Agility Robotics 的股权投资者，也是 Agility 的旗舰部署场地；富士康领投 Agility 的 PIPE 融资，而富士康工业互联网本身就是本站地图上的一个代工节点。
三份独立委托的分析师调查（富国银行、BTIG、招商证券）在直觉外科护城河这一问题上得出了同一个结论：其占据全球软组织手术机器人存量装机的约 81%，现有平台上约四分之三的外科医生倾向于升级到下一代平台，而非转投他家——这是对上文耗材比率证据的独立印证，而非重复。
不过，卖方研究同样揭示了三家正在获得真实但规模尚小进展的低成本挑战者：微创机器人（MedBot）的图迈（Toumai，已装机 23 台，在手订单 86 台，摩根大通中国区首选标的）、SS Innovations 的 Mantra（2023 至 2025 年营收从 590 万美元爬升至 4300 万美元，定价约为达芬奇的一半），以及美敦力的 Hugo——定价均仅为达芬奇 1800 万至 3000 万元人民币标价的一小部分。
在”集成”的代工制造一侧，摩根士丹利将富士康工业互联网称为领先的 AI 服务器组装商（供应份额超过 40%，近十二个月营收增速约 48%），并将其明确视为机器人本体组装产能的代理指标——正是上文已被标记为 Agility 关联方的同一个富士康。
而作为泰瑞达”成本中心论”的罕见反例，摩根士丹利将 KION Group 上调至增持评级，理由正是仓储自动化的附加收入——这是集成商因服务而非因周期性被上调评级的真实案例，不过 KION 的业务结构上更接近 Symbotic，而非泰瑞达，不应被过度推广为对整个系统的定论。
“集成”并非单一商业模式，而是两种，本表的两个上市代理公司正好横跨这两种模式的两端。泰瑞达看起来像一家需要裁员 24% 才能在周期中守住阵地的公司——这是成本中心的画像，而非经常性收入。直觉外科看起来才是真正的经常性收入：它每个季度都披露那个能证明这一点的比率，三份独立的卖方调查也印证了这条护城河依然坚固。
在为任何私募 P17，集成、在线率与服务 标的定规模之前，应先判断它更接近这两种画像中的哪一种——并另行判断其具名客户是否为独立第三方。
私募名录中相当一部分看似第三方的集成需求——从 Agility 与 Schaeffler、富士康的关系，到 KION 仓储自动化附加收入究竟价值几何——可能只是同一张资产负债表在为自己买单，也可能是一门真正由服务驱动的不同生意；二者很容易被混为一谈，值得逐个标的地拆开来看。