Humanoid Robotics: What to Watch in the Next 90 Days [March 14, 2026]

by

The 90-Day Window That Will Define Humanoid Robotics in 2026 → 将定义2026年人形机器人格局的90天窗口期

From March 14 through mid-June 2026, the humanoid robotics industry faces one of its most consequential calendar stretches to date. Four confirmed production milestones, two major academic conferences, and at least one rumored mass-production launch are converging in a window that will stress-test every claim made at CES 2026. This is not a period of announcements — it is a period of delivery. Here is what to watch, why it matters, and which companies are making the boldest moves. → 从2026年3月14日到6月中旬，人形机器人行业正面临迄今为止最重要的一个时间段。四个已确认的生产里程碑、两个主要学术会议，以及至少一个尚未确认的大规模生产发布都将在这个窗口期汇聚，这将是CES 2026上所有承诺的一次压力测试。这不是发布公告的时期，而是交付成果的时期。以下是值得关注的焦点、其重要性，以及哪些公司正在采取最大胆的行动。 I'm analyzing the key timeline events in the humanoid robotics industry. On April 1st, two critical developments will occur. Boston Dynamics Atlas will be deployed at Google DeepMind's AI research facility, representing a significant shift in evaluating the company's commercial potential. Simultaneously, XPENG Robotics will announce full-scale production of its IRON humanoid robot at its Guangzhou manufacturing center, marking a pivotal moment in the company's robotic development strategy. The IRON robot showcases impressive technical specifications, including 82 active degrees of freedom and 200 total degrees of freedom. XPENG is strategically mirroring Tesla's approach by first deploying the robot internally in their factory before expanding to external commercialization. With an estimated price point around $150,000, these initial units have been operational on XPENG's electric vehicle assembly line since January 2025. The April launch focuses not just on proving the robot's capabilities, but demonstrating advanced manufacturing processes. XPENG's expertise in electric vehicle production - including battery management, precision assembly tools, and rapid supply chain capabilities - provides a distinctive manufacturing advantage that specialized robotics companies struggle to replicate. I'm closely monitoring potential external enterprise partnership announcements, which would signal XPENG's readiness to compete beyond its internal ecosystem. In May, the scaling of Digit marks a pivotal moment in warehouse automation. The robot's proven 16 kg payload and established track record in material handling position it competitively in a market projected between $15-50 billion. The critical question remains whether the Robot-as-a-Service (RaaS) model can sustain its economic viability at scale, especially with emerging lower-cost alternatives. Boston Dynamics strategically launches its commercial deployment program just nine days later, on May 10. This program represents a structural evolution, moving beyond pre-allocated anchor customers to a broader customer engagement model. The timing, two weeks before ICRA 2026, appears intentionally strategic, positioning the company to demonstrate an active commercial program to the industry's most significant academic-commercial conference. The enterprise contract units for Atlas will command six-figure pricing, indicating this program won't generate mass volume in 2026. However, it will establish the premium tier of the industrial humanoid market, signaling a deliberate market positioning approach. The ICRA 2026 Humanoid Robotics Summit on May 25, followed by the full International Conference on Robotics and Automation on June 1, represents the industry's most important intellectual gathering of the next 90 days. Unlike trade shows, ICRA produces peer-reviewed research that shapes development roadmaps 12 to 24 months out. With Atlas entering a DeepMind deployment just weeks prior and multiple companies mid-ramp on production, ICRA 2026 will function as a real-time performance review. Researchers and engineers attending will have access to deployment data that has never been available at previous conferences. Expect significant technical papers on bipedal locomotion in unstructured environments, dexterous manipulation benchmarks, and multi-robot fleet coordination — all areas where the gap between demo and deployment remains wide. On June 1st, Tesla is confirmed to begin production of Optimus 3, a potentially transformative event in the robotics industry. At AWE Shanghai in early March, Tesla positioned Optimus 3 as the world's most advanced robot, with ambitious production targets at the Fremont Factory aiming for up to 1 million units annually. While this scale requires significant near-term skepticism - as no humanoid robot manufacturer has yet produced more than a few hundred units - the directional ambition is clear. The current Gen 2 model, priced between $20,000-$30,000, has already generated strong market sentiment, scoring 8.2 out of 10. The upcoming production start of Gen 3 is expected to trigger an intense competitive response across the industry. The market landscape is becoming increasingly dynamic. Companies like Figure AI are directly challenging Tesla's positioning, with their Figure 03 targeting consumer-scale pricing below $20,000. This platform features a comprehensive 44 DoF full-body design and has garnered an impressive 8.5/10 market sentiment. As these developments unfold, the summer of 2026 is shaping up to be the first real test of whether the humanoid robotics market will be dominated by a single winner or if multiple platforms can successfully coexist at scale. Fourier Intelligence's potential GR-2 mass production launch on June 1 adds another layer of complexity. With their GR-1 already commercially available at $150,000-$170,000 and featuring a robust 40 kg payload, this development could significantly reshape market dynamics. The GR-2's potential simultaneous announcement with Tesla's Optimus 3 production start would force the market to absorb two substantial supply signals concurrently. Industry watchers should closely monitor Fourier Intelligence's communications through May for any confirmation or delay signals. Beyond these product launches, two critical sentiment indicators merit ongoing attention. First, the Unitree G1's concerning 2.5 out of 10 negative sentiment score - triggered by a public incident involving an elderly woman requiring police intervention - underscores the emerging risk of consumer perception in humanoid robotics development. As companies like 1X Technologies prepare to enter consumer home deliveries in 2026 at $20,000 or $499 per month, and Figure AI targets sub-$20,000 consumer pricing, the industry lacks a comprehensive playbook for managing public safety narratives at scale. Second, UBTECH's manufacturing acquisition, reflected in its strong 8.5/10 positive sentiment, signals a strategic shift toward vertical integration in production. This approach is becoming a competitive necessity, providing structural advantages to Tesla, XPENG, and UBTECH while potentially disadvantaging companies reliant on contract manufacturing models. The upcoming 90-day period represents a critical inflection point for humanoid robotics. More deployment data will emerge in this short timeframe than in the entire previous three years of the industry's development. Key milestones include: - Atlas at Google DeepMind deployment - XPENG IRON mass production confirmation - Digit's market availability expansion - Boston Dynamics' commercial program launch - ICRA 2026 Humanoid Summit - Potential Tesla Optimus 3 production start - Possible Fourier GR-2 mass production launch I'll track active platform comparisons, review detailed specifications, and monitor deployment milestones as they develop through the /compare/ and /robots/ resources. The critical 90-day window reveals multiple humanoid robot production and deployment events. Key launches include XPENG IRON's mass production on April 1st at $150,000 with advanced degrees of freedom, Tesla's Optimus 3 targeting $20,000-$30,000, and potential Fourier GR-2 mass production. Agility's Digit transitioning to broader market availability on May 1st signals significant industry progression. </think> [N] 人形机器人：未来90天值得关注的焦点... | HumanoidApplications.com [N] 最后更新：2026年3月15日 [N] 将定义2026年人形机器人格局的90天窗口期 [N] 从2026年3月14日到6月中旬，人形机器人行业正面临迄今为止最重要的一个时间段。四个已确认的生产里程碑、两个主要学术会议，以及至少一个尚未确认的大规模生产发布都将在这个窗口期汇聚，这将是CES 2026上所有承诺的一次压力测试。这不是发布公告的时期，而是交付成果的时期。以下是值得关注的焦点、其重要性，以及哪些公司正在采取最大胆的行动。 [N] 4月1日：两大催化剂同日落地 [N] Boston Dynamics Atlas进入Google DeepMind实验室 [N] 确认的4月1日Atlas在Google DeepMind的部署，代表着评估Boston Dynamics商业轨迹的质量性转变。这不是一个仓库取放应用场景。DeepMind是一个AI研究环境，这意味着Atlas将受到工厂部署刻意回避的那种严格、开放式的任务评估。如果Atlas在一个非结构化的研究环境中表现出色——面对关于精细运动控制和日常任务可靠性持续存在的问题——这将实质性地强化该平台长期差异化定位的论证。Boston Dynamics的2026年生产批次已全部预订完毕，分配给现代的Metaplant应用中心以及现在的DeepMind。这种供应限制本身就是一个信号：该公司不追求销量，而是在建立具有最高机构可信度的参考部署案例。 [N] XPENG IRON确认大规模生产 [N] 同样在4月1日，XPENG机器人确认其IRON人形机器人在广州工厂的全面大规模生产发布。IRON的规格是业内最具雄心的之一——82个主动自由度关节和200个总自由度——XPENG正在直接借鉴特斯拉的做法：先在内部工厂部署，然后才对外商业化。以约150,000美元的价格估算，自2025年1月以来，首批机组已在XPENG自己的电动汽车装配线上运行，4月的发布与其说是为了证明机器人，不如说是为了证明制造流程。在电池管理、精密装配工具和供应链速度方面的电动汽车生产专业知识，给XPENG带来了纯机器人公司无法轻易复制的结构性制造优势。密切关注4月份是否有任何外部企业合作公告——这将表明XPENG已准备好在其 captive 市场之外竞争。 [N] 5月：规模成为主题 [N] Agility Robotics Digit转向通用市场供应 [N] 5月1日，Agility Robotics将Digit扩展至通用市场供应——这是一个值得关注但未得到足够关注的里程碑。Digit在此期间已成为真实仓库运营中商业验证最充分的人形机器人，其机器人即服务订阅模式的定价约为250,000美元购买等效值。5月的规模化不是发布；这是解除企业独家瓶颈。对于一直在等待更广泛供应的物流运营商、零售履约中心和第三方物流提供商的采购团队来说，这个日期打开了大门。Digit的16公斤承载能力和在物料搬运方面的成熟记录，在150亿至500亿美元可寻址仓库自动化市场中占据了可防御的地位。关键的分析问题是，Agility的RaaS经济模式在规模化后能否维持，还是来自更低成本平台的竞争压力会在年底前迫使定价调整。 [N] Boston Dynamics于5月10日启动商业部署计划 [N] Boston Dynamics仅九天后的5月10日紧随其后，正式启动Atlas商业部署计划。这具有结构性重要意义。迄今为止，Atlas部署都是预先分配给锚定客户——现代和DeepMind。商业计划意味着更广泛的客户接纳流程、定价结构和支持基础设施。时机——ICRA 2026前两周——几乎可以确定是故意的，让Boston Dynamics以一个活跃的商业计划而非路线图幻灯片出席业内最重要的学术-商业会议。鉴于Atlas机组处于六位数企业合同级别，该计划在2026年不会产生大规模销量，但它将定义工业人形机器人市场的高端层级。 [N] ICRA 2026：汇聚点 [N] 2026年5月25日的ICRA 2026人形机器人峰会，随后是6月1日的完整机器人与自动化国际会议，代表了未来90天内该行业最重要的知识盛会。与贸易展会不同，ICRA产生的是塑造12至24个月后发展路线图的同行评审研究。随着Atlas在数周前进入DeepMind部署，以及多家公司处于生产爬坡中期，ICRA 2026将充当实时性能评估。参会的研究人员和工程师将获得以往会议从未公开的部署数据。预计将有大量关于非结构化环境中双足运动、灵巧操作基准测试和多机器人机队协调的技术论文——这些都是演示与部署之间差距仍然很大的领域。 [N] 6月1日：Tesla Optimus 3生产启动 [N] 未来90天内最具影响力的日期是6月1日，届时Tesla确认开始生产Optimus 3。Tesla在3月初的AWE上海展会上将Optimus 3描述为世界上最先进的机器人，弗里蒙特工厂产能目标是在规模化后达到每年100万台。这个数字在近期需要极大的怀疑——没有人形机器人制造商生产过超过几百台——但发展方向是明确的。目前定价目标为20,000至30,000美元的Optimus Gen 2，在8.2/10的市场情绪评分中已经产生了强劲的市场情绪。Gen 3的生产启动将引发该行业所见过的最激烈竞争反应。像Figure AI这样的公司，其Figure 03以低于20,000美元的消费者规模定价、44个自由度全平台和强劲的8.5/10市场情绪为目标，将在Tesla发出任何产量加速信号时直接面临价格压力。2026年夏天正在成为人形机器人市场是否赢者通吃还是多个平台能否大规模共存的首次真正考验。 [N] 传闻中的变数：Fourier Intelligence GR-2 [N] 关于Fourier Intelligence GR-2大规模生产的传闻（定于6月1日发布）的数据很少，应相应对待。已确认的是GR-1已以150,000至170,000美元的价格商业化，提供40公斤有效载荷——在重载细分市场中与Unitree H1-2的90,000至150,000美元具有竞争力。如果GR-2与Tesla的Optimus 3生产启动同时发布大规模生产公告，将迫使市场同时处理两个重要的供应信号。通过5月监测Fourier Intelligence的沟通以获取任何确认或延迟信号。 [N] 值得关注的情绪暗流 [N] 除了日程之外，过去七天出现的两个情绪信号值得持续关注。首先，Unitree G1的负面情绪评分仅为2.5/10——主要由涉及一位老年妇女和警方干预的公开事件驱动——这提醒我们，消费者认知风险现在是人形开发者面临的实质性业务风险。随着1X Technologies NEO在2026年以20,000美元或每月499美元进入消费者家庭交付，以及Figure AI以低于20,000美元的消费级定价为目标，该行业没有大规模管理公共安全叙事的既定 playbook。其次，UBTECH的制造收购反映在其8.5/10的正面情绪中，表明生产垂直整合正在成为竞争必需品而非可选项——这种动态对Tesla、XPENG和UBTECH有利，对依赖合同制造的的公司不利。 [N] 您的90天关注清单 [N] 4月1日： [N] Google DeepMind的Atlas部署——首个非结构化AI研究部署；以及XPENG IRON大规模生产确认 [N] 5月1日： [N] Digit通用市场供应——首个脱离企业独家状态的人形机器人 [N] 5月10日： [N] Boston Dynamics商业部署计划启动——高端层级定价和接纳结构 [N] 5月25日： [N] ICRA 2026人形机器人峰会——2027路线图的基准数据和研究信号 [N] 6月1日： [N] Tesla Optimus 3生产启动——本季度需求和价格压力事件；以及Fourier GR-2大规模生产（等待确认） [N] 持续关注： [N] Unitree G1公开事件余波——更广泛消费人形细分市场的监管和消费者情绪风险领先指标 [N] 未来90天将产生比人形机器人行业此前三年加起来更多的硬性部署数据。好好利用它。 [N] 在 /compare/ 进行活跃平台并排对比 [N] 在 /robots/ 查看完整规格 [N] 在 /deployments/ 追踪已确认的部署里程碑 [N] 未来90天人形机器人领域有哪些关键日期需要关注？ [N] 4月1日带来Atlas在Google DeepMind的部署和XPENG IRON大规模生产。5月1日标志着Digit通用市场供应。5月10日启动Boston Dynamics的商业计划。5月25日是ICRA 2026人形机器人峰会。6月1日是Tesla Optimus 3生产启动，以及可能的Fourier GR-2大规模生产发布。 [N] 此期间有哪些人形机器人发布？ [N] XPENG IRON于4月1日在其广州工厂开始大规模生产，配备82个主动自由度，预计价格为150,000美元。Tesla Optimus 3于6月1日开始生产，目标价格为20,000至30,000美元。Fourier Intelligence GR-2大规模生产传闻定于6月1日但尚未确认。Agility Digit于5月1日脱离企业独家状态。

April 1: Two Catalysts Land Simultaneously

Boston Dynamics Atlas Enters the Google DeepMind Lab

The confirmed April 1 deployment of Atlas at Google DeepMind represents a qualitative shift in how we should evaluate Boston Dynamics’ commercial trajectory. This is not a warehouse pick-and-place application. DeepMind is an AI research environment, which means Atlas will be subjected to the kind of rigorous, open-ended task evaluation that factory deployments deliberately avoid. If Atlas performs well in an unstructured research setting — against a backdrop of persistent questions about fine motor control and everyday task reliability — it materially strengthens the case for the platform’s long-term differentiation. Boston Dynamics’ 2026 production run is already fully committed, with fleets allocated to Hyundai’s Metaplant Application Center and now DeepMind. That supply constraint is itself a signal: the company is not chasing volume, it is building reference deployments with maximum institutional credibility.

Watch : Xpeng Unzips Humanoid Robot To Prove It's Not Human ? — OTOFOOTAGE → 观看：Xpeng展示人形机器人以证明其非人类？ — OTOFOOTAGE

XPENG IRON Confirms Mass Production

Also on April 1, XPENG Robotics confirms the full mass production launch of its IRON humanoid at its Guangzhou facility. IRON’s specifications are among the most aggressive in the field — 82 active degrees of freedom and 200 total — and XPENG is deploying a playbook borrowed directly from Tesla: internal factory deployment first, then external commercialization. With an estimated price point around $150,000 and initial units already running on XPENG’s own EV assembly line since January 2025, the April launch is less about proving the robot and more about proving the manufacturing process. EV production expertise in battery management, precision assembly tooling, and supply chain velocity gives XPENG a structural manufacturing advantage that pure-play robotics companies cannot easily replicate. Watch April closely for any announcements about external enterprise partnerships — that would signal XPENG is ready to compete beyond its captive market.

May: Scale Becomes the Story

Agility Robotics Digit Moves to General Market Availability

On May 1, Agility Robotics scales Digit to general market availability — a milestone that deserves more attention than it has received. Digit entered this period as the most commercially validated humanoid robot in real warehouse operations, with its Robot-as-a-Service subscription model priced at approximately $250,000 purchase equivalent. The May scale-up is not a launch; it is the removal of the enterprise-only bottleneck. For procurement teams at logistics operators, retail fulfillment centers, and third-party logistics providers who have been on the sideline waiting for broader availability, this date opens the door. Digit’s 16 kg carry payload and proven track record in material handling give it a defensible position in the $15–$50 billion addressable warehouse automation market. The key analytical question is whether Agility’s RaaS economics hold at scale, or whether competitive pressure from lower-cost platforms forces a pricing adjustment before year-end.

Boston Dynamics Launches Its Commercial Deployment Program on May 10

Boston Dynamics follows just nine days later on May 10 with the formal launch of its Atlas Commercial Deployment Program. This is structurally significant. Until now, Atlas deployments have been pre-allocated to anchor customers — Hyundai and DeepMind. A commercial program implies a broader customer intake process, pricing structure, and support infrastructure. The timing — two weeks before ICRA 2026 — is almost certainly deliberate, allowing Boston Dynamics to arrive at the industry’s most important academic-commercial conference with an active commercial program rather than a roadmap slide. Given that Atlas units are in six-figure enterprise contract territory, this program will not generate mass volume in 2026, but it will define the premium tier of the industrial humanoid market.

ICRA 2026: The Convergence Point

The ICRA 2026 Humanoid Robotics Summit on May 25, followed by the full International Conference on Robotics and Automation on June 1, represents the industry’s most important intellectual gathering of the next 90 days. Unlike trade shows, ICRA produces peer-reviewed research that shapes development roadmaps 12 to 24 months out. With Atlas entering a DeepMind deployment just weeks prior and multiple companies mid-ramp on production, ICRA 2026 will function as a real-time performance review. Researchers and engineers attending will have access to deployment data that has never been available at previous conferences. Expect significant technical papers on bipedal locomotion in unstructured environments, dexterous manipulation benchmarks, and multi-robot fleet coordination — all areas where the gap between demo and deployment remains wide.

June 1: The Tesla Optimus 3 Production Start

The single most consequential date in the next 90 days is June 1, when Tesla is confirmed to begin production of Optimus 3. Tesla described Optimus 3 at AWE Shanghai in early March as the most advanced robot in the world, with Fremont Factory capacity targeting up to 1 million units annually at scale. That figure requires enormous skepticism in the near term — no humanoid robot manufacturer has produced more than a few hundred units — but the direction of travel is unambiguous. Optimus Gen 2, currently priced at a target of $20,000–$30,000, already generates strong market sentiment at 8.2 out of 10. Gen 3’s production start will trigger the most intense competitive response the industry has seen. Companies like Figure AI, whose Figure 03 targets below $20,000 at consumer scale with a 44 DoF full-body platform and strong 8.5/10 market sentiment, will face direct pricing pressure the moment Tesla signals any acceleration in volume. The summer of 2026 is shaping up to be the first real test of whether the humanoid market is winner-take-most or whether multiple platforms can coexist at scale.

The Rumored Wildcard: Fourier Intelligence GR-2

Data on Fourier Intelligence’s GR-2 mass production launch, rumored for June 1, is sparse and should be treated accordingly. What is confirmed is that the GR-1 is already commercially available at $150,000–$170,000 with a 40 kg payload — competitive with the Unitree H1-2 at $90,000–$150,000 in the heavy-payload segment. A GR-2 mass production announcement coinciding with Tesla’s Optimus 3 production start would force the market to process two significant supply signals simultaneously. Monitor Fourier Intelligence communications through May for any confirmation or delay signals.

Sentiment Undercurrents Worth Tracking

Beyond the calendar, two sentiment signals from the past seven days merit ongoing attention. First, the Unitree G1’s negative sentiment score of 2.5 out of 10 — driven primarily by a public incident involving an elderly woman and police intervention — is a reminder that consumer perception risk is now a material business risk for humanoid developers. As 1X Technologies NEO enters consumer home deliveries in 2026 at $20,000 or $499 per month, and as Figure AI targets sub-$20,000 consumer pricing, the industry has no established playbook for managing public safety narratives at scale. Second, UBTECH’s manufacturing acquisition, reflected in its 8.5/10 positive sentiment, signals that vertical integration in production is becoming a competitive necessity rather than an option — a dynamic that advantages Tesla, XPENG, and UBTECH and disadvantages companies dependent on contract manufacturing.

Your 90-Day Watchlist

  • April 1: Atlas at Google DeepMind — first unstructured AI research deployment; and XPENG IRON mass production confirmation
  • May 1: Digit general market availability — first humanoid to exit enterprise-only status
  • May 10: Boston Dynamics Commercial Deployment Program launch — premium tier pricing and intake structure
  • May 25: ICRA 2026 Humanoid Summit — benchmark data and research signals for 2027 roadmaps
  • June 1: Tesla Optimus 3 production start — the demand and pricing pressure event of the quarter; and Fourier GR-2 mass production (watch for confirmation)
  • Ongoing: Unitree G1 public incident fallout — a leading indicator of regulatory and consumer sentiment risk for the broader consumer humanoid segment

The next 90 days will produce more hard deployment data than the entire preceding three years of the humanoid robotics industry combined. Use it. Compare active platforms side-by-side at /compare/, review full specifications at /robots/, and track confirmed deployment milestones as they land at /deployments/.

常见问题

What are the key dates to watch in humanoid robotics over the next 90 days?

April 1 brings Atlas deployment at Google DeepMind and XPENG IRON mass production. May 1 marks Digit general market availability. May 10 launches Boston Dynamics’ commercial program. May 25 is the ICRA 2026 Humanoid Summit. June 1 is Tesla Optimus 3 production start and possible Fourier GR-2 mass production launch.

Which humanoid robots are launching in this period?

XPENG IRON begins mass production April 1 at its Guangzhou facility with 82 active degrees of freedom and an estimated $150,000 price point. Tesla Optimus 3 starts production June 1 targeting $20,000-$30,000. Fourier Intelligence GR-2 mass production is rumored for June 1 but unconfirmed. Agility Digit exits enterprise-only status May 1.

2026年春季人形机器人有哪些会议安排？ ---

ICRA 2026人形机器人峰会将于5月25日举行，随后于6月1日举办完整的International Conference on Robotics and Automation会议。与贸易展会不同，ICRA发布经过同行评审的研究成果，这些研究将塑造未来12-24个月的发展路线图。预计将有关于双足运动、灵巧操作和多机器人舰队协调的重要论文。 ---

2026年人形机器人的预期价格是多少？ ---

Tesla目标价格为Optimus Gen 2为20,000-30,000美元，Gen 3生产将于6月1日启动。Figure AI目标价格为Figure 03消费者规模版本低于20,000美元。XPENG IRON企业版预计为150,000美元。Agility Digit维持在约250,000美元的采购等价。Fourier GR-1维持在150,000-170,000美元，有效载荷40 kg。 ---

未来90天将如何影响人形机器人市场？ ---

这一时期将产生比过去三年总和更多的硬核部署数据。Tesla Optimus 3生产将引发激烈的竞争性定价压力。Digit的全面上市将测试RaaS经济模型在规模化下的可行性。Boston Dynamics的商业计划将定义高端市场层级。结果将决定市场是赢家通吃还是多平台共存。