Google absorbs Intrinsic robotics: why Alphabet is merging AI with physical automation
Alphabet's robotics software company Intrinsic joins Google. What this means for AI-powered robots and Google's automation strategy.
Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchAI, startups & growth insights. No spam.
TL;DR: Alphabet announced on February 25, 2026 that Intrinsic, its robotics software subsidiary, is being folded into Google. The move ends Intrinsic's nearly five-year run as a standalone "Other Bets" company and signals that Google is now treating physical automation as a core product priority, not a moonshot side project. Intrinsic will continue operating its Flowstate platform and work directly with Google DeepMind and Gemini.
On February 25, 2026, Alphabet confirmed that Intrinsic Innovation LLC is joining Google. The announcement ended Intrinsic's existence as a discrete Alphabet subsidiary inside the "Other Bets" portfolio and brought it directly under Google's corporate umbrella.
Intrinsic will remain a distinct team inside Google, but it will now build in direct coordination with Google DeepMind and tap into Google's Gemini AI models and Cloud infrastructure. Alphabet declined to share financial terms related to the integration. There was no acquisition price because there was no acquisition -- Intrinsic was already wholly owned by Alphabet. This is an internal restructuring that changes where Intrinsic sits on the org chart, not who owns it.
CEO Wendy Tan White framed the move in terms of scale:
"Combined with Google's incredible AI and infrastructure, we're going to unlock the promise of physical AI for a much broader set of manufacturing businesses and developers. This will fundamentally shift production, from its economics to operations, and enable truly advanced manufacturing."
The timing is deliberate. It comes weeks after Intrinsic's Foxconn joint venture went live, months after Google DeepMind released Gemini Robotics 1.5, and at a moment when the broader robotics industry is attracting the most concentrated capital and attention in its history.
Intrinsic spent five and a half years inside Alphabet's X division, the "moonshot factory" where early-stage bets on transformative technology are incubated. It graduated into a standalone Alphabet company in July 2021, with a clear mission: make industrial robotics easier to program and deploy at scale.
The company does not make robots. It makes the software that makes robots useful.
Its flagship product is Flowstate, a web-based developer environment for building robotic applications without requiring deep expertise in robotics programming. Historically, configuring an industrial robot for a new task required months of engineering work -- custom code, specialized integrations, and significant re-engineering every time a product changed. Flowstate reduces that barrier substantially by providing a visual programming interface built on top of powerful AI capabilities.
On top of Flowstate, Intrinsic shipped the Intrinsic Vision Model (IVM) in late 2025. IVM is a foundation model for machine vision tasks in manufacturing settings. It handles pose detection, object tracking, segmentation, and point cloud generation using a suite of specialized transformers. Critically, it achieves sub-millimeter accuracy with standard RGB cameras and requires no training pipelines -- give it a CAD file, and it can start working. That removes one of the biggest friction points in deploying computer vision on a factory floor.
Before IVM, Intrinsic also moved aggressively on acquisitions. In 2022 it acquired several for-profit divisions of Open Robotics, the nonprofit behind the Robot Operating System (ROS). ROS is the closest thing robotics has to a universal middleware layer, and picking up those assets gave Intrinsic deep roots in the developer ecosystem.
The company also built deep integrations with NVIDIA. Flowstate connects to NVIDIA's Isaac foundation models for advanced robotic grasping and to the Omniverse platform for physically-based digital twin simulation. Developers using Flowstate can run a full simulation of a robotic workcell before touching hardware.
The simplest answer: Intrinsic and Google are more valuable combined than separate.
As an "Other Bets" company, Intrinsic operated somewhat independently of Google's core engineering and AI infrastructure. That autonomy allowed it to move fast and make product decisions without corporate overhead. But it also meant Intrinsic had to build or license capabilities that Google already had -- large foundation models, cloud compute, enterprise sales relationships, and global infrastructure.
Folding Intrinsic into Google eliminates that friction. Flowstate can now be built natively on top of Gemini models rather than integrating with them through external APIs. Intrinsic's enterprise customers can access its platform through Google Cloud rather than a separate sales motion. And Google DeepMind's robotics research -- which has been progressing rapidly -- can feed directly into Intrinsic's product roadmap.
There is also a competitive pressure dimension. Amazon has deployed hundreds of thousands of robots in its warehouse network. Microsoft is investing heavily in OpenAI, which has its own physical AI ambitions. Tesla is running Optimus through its paces in its Fremont factory. If Google wants to be a meaningful player in industrial automation, it needs to move faster and with more coordination than a structure of separated subsidiaries allows.
The "Other Bets" portfolio was always intended to house experiments. Once an experiment matures into something with real revenue potential and clear strategic alignment, it makes sense to bring it inside. Google Maps, Android, and YouTube all started as external acquisitions or separate initiatives that eventually became core to Google's product strategy. Intrinsic appears to be following a similar path.
The connection to Google DeepMind is where this move gets technically interesting.
DeepMind has been building toward what it calls "embodied AI" -- models that do not just process text or images but understand and act in the physical world. The trajectory from RT-2 (2023) through Gemini Robotics (early 2025) to Gemini Robotics 1.5 (September 2025) shows rapid progress.
Gemini Robotics 1.5 introduced what DeepMind calls "Chain-of-Action" planning. The idea is that a robot should reason through a sequence of decisions before acting, the same way language models reason through tokens before producing an answer. Rather than reacting to each input in isolation, the robot builds a plan and executes it. That is a qualitatively different capability from earlier systems.
DeepMind also shipped Gemini Robotics On-Device in June 2025, a lightweight version of the model optimized to run locally on robot hardware with low latency. That matters for factory settings where cloud connectivity cannot be assumed and where milliseconds of response time make a difference.
Intrinsic's Flowstate platform is a natural deployment layer for these models. DeepMind produces the underlying AI; Intrinsic provides the developer tools, simulation environment, and hardware integrations that make it possible to actually deploy that AI on a factory floor. The two pieces fit together in a way they could not when they were in separate parts of the organization.
Intrinsic CEO Wendy Tan White has previously described the goal as building "the Android of robotics" -- a platform layer that standardizes how industrial robots are programmed and deployed, the same way Android standardized mobile software across hundreds of hardware devices. Combining Intrinsic's platform with DeepMind's models and Google's cloud infrastructure makes that vision more plausible than it was a year ago.
One of the clearest signals of where Intrinsic was headed before this announcement was its joint venture with Foxconn, which launched in November 2025 at Foxconn's Hon Hai Tech Day.
The US-based joint venture focuses on electronics assembly -- one of the most labor-intensive and difficult-to-automate manufacturing domains. Electronics components are small, varied, and fragile. Assembly requires precise manipulation and rapid adaptation as product designs change across generations. That has made it resistant to traditional fixed automation.
The Foxconn-Intrinsic JV uses Flowstate as the common development environment and IVM as the vision backbone. Target applications include parts insertion, inspection, machine tending, and logistics within Foxconn's US facilities. The longer-term goal is "full factory orchestration and automation" -- a phrase that captures the ambition even if the timeline remains open.
Foxconn is one of the largest contract electronics manufacturers in the world. Using Intrinsic's platform at Foxconn's scale would be a significant proof point for Google's physical AI push. Under the new structure, that joint venture is now a Google partnership, not just an Intrinsic one.
Intrinsic joining Google does not happen in a vacuum. The broader race to own the infrastructure layer of physical AI is accelerating, with well-capitalized competitors across hardware, software, and the hybrid space between them.
Tesla Optimus is the highest-profile humanoid program in the market. As of early 2026, Tesla has begun Gen 3 production at its Fremont factory, though the robots are currently in a learning phase rather than performing productive work. CEO Elon Musk has guided for consumer availability by end of 2027 at a price point of $20,000 to $30,000. Optimus runs on xAI's Grok for language reasoning and FSD-derived neural networks for physical movement.
Figure AI launched Figure 03 in late 2025, targeting high-volume manufacturing. The robot uses the company's Helix model for full-body autonomy and incorporates safety-focused hardware design with soft textile components. Figure is in commercial alpha testing with selected partners.
Boston Dynamics is the most operationally advanced of the humanoid builders. Its Atlas robot is now in commercial production, with tens of thousands of units planned for Hyundai Motor Group manufacturing facilities. Boston Dynamics is deploying Atlas in Hyundai's Georgia plant in 2026. Pricing is in the $140,000 to $150,000 range per unit. Atlas supports teleoperation via VR, tablet control, and autonomous operation.
Amazon is not building humanoids but has deployed more robots than any other company in the world through its warehouse network. Its Proteus, Sequoia, and Digit programs span mobile robots, picking systems, and humanoid prototypes in partnership with Agility Robotics.
Google's angle is different from all of these. Google is not building robots. It is building the AI software layer that runs on robots from any manufacturer, similar to how Android runs on phones made by dozens of hardware vendors. That is a different bet with a different competitive position.
| Company | Robot hardware | Software platform | Foundation AI model | Current deployment |
|---|---|---|---|---|
| Google (Intrinsic) | ✗ | Flowstate | Gemini Robotics | Industrial (via partners) |
| Tesla | Optimus (humanoid) | FSD-derived | Grok (xAI) | Internal (testing) |
| Boston Dynamics | Atlas (humanoid) | Orbit | Custom | Hyundai factories |
| Figure AI | Figure 03 (humanoid) | Helix | Custom | Alpha testing |
| Amazon (Agility) | Digit (humanoid) | Custom | Custom | Amazon warehouses |
| NVIDIA | ✗ | Isaac / Omniverse | Custom VLAs | Platform (via partners) |
The table highlights something important. Google and NVIDIA are both playing the platform game -- selling AI and software tools that hardware makers and manufacturers use, rather than competing directly in the robot hardware market. That approach has a different margin profile and a different adoption path. If Intrinsic's Flowstate becomes the standard developer environment for industrial robotics, Google earns a cut of every deployment regardless of which robot company's hardware is running.
For manufacturers and factory operators, Intrinsic joining Google means the platform will have significantly more resources behind it. Google Cloud integration should make it easier to deploy Flowstate at enterprise scale, and access to Gemini models means the vision and reasoning capabilities of the platform will improve faster than Intrinsic could have driven alone.
For robotics developers, the picture is similar. Flowstate built on top of Gemini and backed by Google's developer ecosystem is a more compelling platform than Flowstate as a standalone product from a mid-stage Alphabet subsidiary.
For the industry, the move confirms a pattern that is becoming clear across multiple companies: the AI software layer, not the hardware, is where the most durable competitive positions are being built. Intrinsic's absorption into Google is Google staking a claim to that layer for industrial robotics, the same way it staked a claim to mobile computing through Android.
The global physical AI market was valued at $5.1 billion in 2025 and is projected to reach $83.6 billion by 2035, a 34% compound annual growth rate. Industrial robotics alone is a $50 billion market growing at 11% per year. Google's move into this space is not a small bet.
The question the industry will spend the next 18 months answering: can Flowstate plus Gemini plus Google Cloud become the de facto platform for deploying AI in physical manufacturing environments? Google has the distribution, the AI research capability, and now the organizational alignment to find out.
"Alphabet's Intrinsic Joins Google: reabsorbing a moonshot to accelerate physical AI" -- @SiliconANGLE
"The 'Android moment' for robotics? Google reabsorbs Intrinsic -- accelerating the push for physical AI" -- @futunn_en
Intrinsic is a robotics software company that was previously a standalone Alphabet subsidiary. It builds software and AI models that make industrial robots easier to program and deploy. Its flagship product is Flowstate, a web-based developer environment for building robotic applications without deep robotics expertise. It also built the Intrinsic Vision Model (IVM) for machine vision tasks in manufacturing.
Alphabet announced on February 25, 2026 that Intrinsic would be joining Google. The company had operated as an independent Alphabet "Other Bets" subsidiary since July 2021, when it graduated from Alphabet's X moonshot division after five and a half years of incubation.
Yes. Intrinsic will continue operating as a distinct team within Google and will maintain its product line including Flowstate. The integration with Google means Flowstate will have access to Gemini models and Google Cloud infrastructure, which should strengthen the platform's AI capabilities.
After the integration, Intrinsic will work closely with Google DeepMind. DeepMind has been building robotics-specific AI models including the Gemini Robotics series. Intrinsic's Flowstate provides the developer tooling and deployment layer that can run on top of these models. Together they represent a full stack: foundation models from DeepMind, developer platform from Intrinsic, and cloud infrastructure from Google.
The joint venture launched in November 2025 to deploy AI-driven robotics in Foxconn's US electronics manufacturing facilities. It uses Flowstate and the Intrinsic Vision Model to automate assembly, inspection, and logistics tasks. That partnership continues under the new structure and is now effectively a Google-Foxconn collaboration.
Google, through Intrinsic, is building the software and AI platform layer rather than the robot hardware itself. Tesla is building Optimus, its own humanoid robot. Boston Dynamics builds and sells Atlas. Google's bet is that the platform -- like Android for mobile -- is more valuable than any individual hardware product. Intrinsic's Flowstate is designed to run on robots made by many different manufacturers.
Intrinsic's CEO is Wendy Tan White, who has led the company since before it graduated from Alphabet's X division. She has previously described the company's mission as building "the Android of robotics" -- a platform layer that standardizes industrial robot programming across hardware from different manufacturers.
The Intrinsic Vision Model (IVM) is a foundation model released in late 2025 for machine vision tasks in manufacturing environments. It handles pose detection, object tracking, segmentation, and point cloud generation using specialized transformer architectures. It achieves sub-millimeter accuracy with standard RGB cameras, requires no training pipelines, and only needs a CAD file of the target object to operate. It is integrated into Flowstate.
The global robotics market was valued at approximately $50 billion in 2025, growing at 11% per year, with a projected value of $111 billion by 2030. The physical AI segment specifically was valued at $5.1 billion in 2025 and is projected to reach $83.6 billion by 2035, representing a 34% compound annual growth rate. AI-powered industrial robots represent a $16.8 billion market in 2025.
This move positions Google to compete in enterprise automation markets where it previously had limited presence. Industrial manufacturing is a massive sector that has been slow to adopt AI compared to knowledge work. By bringing Intrinsic inside Google and connecting it to Gemini and Google Cloud, Google now has a credible offering for manufacturers who want to automate production with AI. The strategic parallel is Google Cloud's competition with AWS and Azure -- a late start, but strong underlying technology and a clear will to compete.
German robotics startup Neura Robotics closed approximately €1 billion in funding from Tether Holdings, valuing the company at €4 billion as it prepares to fill nearly €1 billion in existing orders for cognitive humanoid machines.
NVIDIA GTC 2026 runs March 16-19 in San Jose with the Rubin platform, Jetson T4000, physical AI breakthroughs, and robotics partnerships with Boston Dynamics and more.
While the DoD blacklisted Anthropic as a supply chain risk, Microsoft Azure and Google Cloud continue offering Claude to commercial enterprise clients — creating a two-tier AI reality.