PokeBot has reportedly raised $100 million in just four months to attack what Tech Times calls "robotics' last hard problem": manipulation. The headline is easy to treat as another sign of 2026's overheated robotics funding market, but the focus is exactly right. Walking robots, warehouse robots, and factory cobots all become more valuable when they can handle the physical world with less scripting and fewer fixtures.
This post contains affiliate links. We may earn a small commission at no extra cost to you.Manipulation is where robotics stops looking like software and starts colliding with everything messy about reality. A robot has to identify an object, understand its shape and material, choose a grasp, control force, compensate for slippage, recover when the object moves, and complete a useful downstream task. That is hard enough with rigid parts in a factory. It becomes much harder with bags, cables, clothing, food, tools, retail items, medical supplies, and household clutter.
That is why a manipulation-focused funding round matters even before the company proves it can turn capital into deployments. The robotics market is full of impressive locomotion demos and foundation-model videos, but the commercial bottleneck is often the hand, gripper, perception stack, and policy that let a robot reliably do the thing a customer actually pays for.
Why Manipulation Is Still the Prize
Robots have been moving through the world for a long time. Automated guided vehicles, autonomous mobile robots, drones, quadrupeds, and robotaxis all show different flavors of mobility. Manipulation is a different layer of value. A mobile robot that cannot interact with objects is mostly a sensor platform or transport platform. A robot that can manipulate objects can pack boxes, load machines, stock shelves, prep food, sort recycling, support lab workflows, tend equipment, or help in care settings.
The challenge is that manipulation does not scale neatly from one object to the next. A suction cup that works on a cardboard box may fail on a mesh bag. A parallel gripper that handles a metal part may crush fruit. A dexterous hand that looks humanlike may be too fragile, expensive, or difficult to control for production use. Even the same object can behave differently depending on lighting, orientation, packaging, wear, moisture, and the surface it sits on.
This is the reason companies like Covariant, Ambi Robotics, RightHand Robotics, Dexterity, Figure, Physical Intelligence, Google DeepMind, NVIDIA, and many others keep circling the same problem from different angles. Some start with warehouse picking. Some start with humanoid hands. Some start with general robotics models. Some start with teleoperation and data collection. The commercial logic is the same: whoever makes object handling more general and more reliable expands the addressable market for robots.
Funding Is Following the Bottleneck
The timing also fits the broader investment pattern. Robotics venture funding has surged in 2026 as investors search for the physical-AI equivalent of the model-platform boom. Capital is flowing toward companies that can collect real-world data, train policies, and deploy robots into revenue-generating work. Manipulation sits at the center of that thesis because it produces valuable data and creates visible customer ROI.
There is a useful warning here too. A large raise does not solve the physics. Manipulation startups can burn enormous sums on hardware iterations, gripper design, robot fleet operations, labeled data, simulation, and integration work. Customers also care less about whether a system is "general" than whether it completes their specific job consistently and cheaply. The winners will probably look less like demo labs and more like operations companies with deep robot-learning teams inside.
For engineers and buyers trying to understand the category, a practical grounding in robot grippers, tactile sensing, and robotics manipulation textbooks is still useful. The field's frontier is AI-heavy, but the failures are often mechanical, tactile, and operational.
The Broader Robotics Context
PokeBot's reported raise lands as humanoid companies are racing to prove that general-purpose robots can do real work, not just walk convincingly. But the manipulation problem cuts across form factors. A wheeled mobile manipulator may be the right answer in a warehouse. A fixed arm may be the best answer in a factory cell. A humanoid may be useful where the environment is already designed around human reach and tools.
That means the best manipulation stack may become a platform layer used by many robot bodies rather than a single product category. Vision-language-action models, force control, tactile sensing, task planning, and data pipelines could move across arms, hands, cobots, and humanoids. If PokeBot is truly targeting the "last hard problem," the test will be whether it can make manipulation repeatable across enough objects and tasks to matter commercially.
The next signal to watch is customer evidence: cycle times, success rates, intervention rates, object variety, cost per task, and deployment time. Robot manipulation is finally getting the capital attention it deserves. Now it has to earn the uptime.
Source: Tech Times via Google News, "PokeBot Lands $100M in Four Months to Attack Robotics' Last Hard Problem: Robot Manipulation", August 3, 2026.