GPT-56T 861 —
MUSE-SPK 835 -0.7%
GPT-56SC 827 -5.3%
QWEN-38X 824 —
CL-OP55X 820 —
GPT-6A 820 —
GROK-46H 820 -5.2%
GLM-5 784 -8.4%
KIMI-K3X 742 -8.4%
CL-FAB5H 742 -5.7%
CL-OP5H 718 -6%
CL-OP5X 708 -18.2%
CL-OP46H 696 -6.2%
CL-OP47H 688 -6.1%
GEM-38FH 677 +0.1%
GEM-37FH 655 -24.3%
GPT-56S 619 —
GPT-55H 580 —
CL-OP47 579 -0.7%
INKL 531 —
GEM-31P 512 —
GEM-3P 498 —
CL-OP46 496 —
CL-OP48 489 -0.2%
GPT-56T 861 —
MUSE-SPK 835 -0.7%
GPT-56SC 827 -5.3%
QWEN-38X 824 —
CL-OP55X 820 —
GPT-6A 820 —
GROK-46H 820 -5.2%
GLM-5 784 -8.4%
KIMI-K3X 742 -8.4%
CL-FAB5H 742 -5.7%
CL-OP5H 718 -6%
CL-OP5X 708 -18.2%
CL-OP46H 696 -6.2%
CL-OP47H 688 -6.1%
GEM-38FH 677 +0.1%
GEM-37FH 655 -24.3%
GPT-56S 619 —
GPT-55H 580 —
CL-OP47 579 -0.7%
INKL 531 —
GEM-31P 512 —
GEM-3P 498 —
CL-OP46 496 —
CL-OP48 489 -0.2%
← Back to feed

Meta Acquires ARI to Build the Android of Humanoid Robots

Meta acquired Assured Robot Intelligence (ARI) on May 1 and folded both co-founders directly into Meta Superintelligence Labs. The deal closed the same day it was announced. Financial terms were not disclosed.

ARI was founded a year ago by Lerrel Pinto and Xiaolong Wang. Pinto had previously co-founded Fauna Robotics — Amazon acquired Fauna in March 2026, along with its bipedal Sprout robot and 50 employees. Wang came from Nvidia research and UC San Diego, where he won the MLSys 2024 Best Paper Award for AI model optimisation work.

Together they built two things worth acquiring: whole-body humanoid control models trained on reinforcement learning from demonstration, and a tactile sensor called e-Flesh that gives robots fine-grained pressure and texture feedback — a capability gap Pinto and Wang described as critical to “physical AGI.”

The Android Play

Meta’s robotics ambition is now explicitly articulated: it wants to do for humanoids what Google Android and Qualcomm did for smartphones. Build the intelligence layer — sensors, software, control models — and let hardware manufacturers build the machines. The company launched Meta Robotics Studio in 2025 to anchor the software platform. ARI’s technology, team, and IP now sit inside that effort.

The comparison to Android is not casual. Meta missed mobile. It acquired Instagram and WhatsApp to survive the smartphone era without owning the OS. The company is structurally determined not to repeat that error in the physical AI era, where the early platform bet may prove as durable as Android’s lead.

What ARI Built

The core ARI system trains robot control policies through demonstration rather than programmed instruction. A human operator shows a robot a task; the model generalises from that demonstration to novel scenarios. This matters because humanoid deployment in homes and high-value labour markets requires adaptability to unpredictable environments — fixed programming breaks on contact with real-world variability.

The e-Flesh sensor is ARI’s proprietary hardware contribution. It measures surface texture and contact pressure at a resolution comparable to human fingertip sensitivity. Robotic manipulation without fine-grained tactile feedback has been a persistent limitation in grasping objects with varied surfaces: eggs, textiles, circuit boards. e-Flesh addresses that directly.

Context

Meta’s humanoid push puts it into direct competition with its own portfolio dependencies. Its Llama models run on some of the same robotics stacks that are now competing with Meta’s in-house platform. 1X Technologies (OpenAI-backed) opened its NEO consumer robot factory in Hayward, California on April 30 — the same day ARI closed. The humanoid manufacturing race is accelerating faster than the intelligence layer beneath it, which is precisely the gap Meta is trying to own.

ARI joins an acqui-hire pattern building across big tech: Amazon bought Fauna Robotics in March, Tesla has a reported $2B hardware acquisition buried in its Q1 10-Q, and Google’s Gemini Robotics-ER 1.6 shipped earlier this year for industrial deployments. The physical AI layer is becoming the next platform war.