May 22 (Reuters) - These robots don't break into a jig

or jump, but their ability to do single tasks cheaply and

efficiently is attracting investor dollars as focus shifts to

function from flash.

Far from the sleek humanoids of science fiction that are

meant for complex and adaptive work, boxy and utilitarian robots

- some the size of industrial tool chests - are built to handle

tasks such as hauling parts, collecting trash or inspecting

equipment.

Interest has been rising in such specialized robots as they

offer a clear path to profitability, given the stress on

automation across industries, including retail, defense and

waste management, company executives, analysts and investors

told Reuters.

In contrast, makers of general-purpose humanoids are still

grappling with technical challenges, including limited training

data and difficulty operating in unpredictable real-world

environments, to make them viable.

Data from PitchBook shows robotics companies globally raised

$2.26 billion in the first quarter of 2025, with more than 70%

of that capital funneled into firms making task-focused

machines.

The funding reflects a broader global race to robot

supremacy. From Shenzhen to Silicon Valley, companies are trying

to develop machines that can take on physical work, with China

emerging as a major player thanks to ample government support.

The push has been fueled by advances in chip technology,

which has enabled more sophisticated AI models that allow robots

to perceive, process and react without needing remote servers.

"With Nvidia's ( NVDA ) Orin NX, we were able to put far more AI

models on the edge than we could earlier," said Saurabh Chandra,

CEO of Ati Motors, referring to Nvidia's ( NVDA ) high-performance AI

chip designed to run multiple machine learning models directly

on edge devices such as robots, without relying on the cloud.

Based in India's tech capital of Bengaluru, Ati Motors makes

robots that can tug around more than 1,000 kilograms on factory

floors and industrial sites. Ati Motors has deployed hundreds of

robots across more than 50 factories globally, including at

Hyundai, Forvia, and Bosch, with its

flagship Sherpa Tug logging over 500,000 kilometers in

operation.

In healthcare, Austin-based Diligent Robotics is seeing

traction with Moxi, a robot that handles non-patient-facing

tasks such as delivering supplies, medications and lab samples.

"We've found that by solving a very specific problem in a

high-need area like healthcare, we can create a sustainable

business model," Diligent Robotics CEO Andrea Thomaz said,

adding Moxi has reached product-level profitability.

HUMANOID CHALLENGES

The interest comes as general-purpose humanoids face

challenges such as teaching machines to navigate unpredictable

environments and developing sophisticated reasoning abilities.

Unlike generative AI, which is trained on vast online

datasets of text, images and audio, the data available to

develop humanoid robots is far more limited. These machines must

learn by interacting with the physical world and training on

datasets focused on tasks such as stacking boxes.

Firms like Figure AI, which aims to ship 100,000 humanoid

robots over the next four years, rely on advanced AI to process

real-time sensory data. That means such robots are mostly

confined to controlled environments such as car factories.

The cost of humanoids is also far higher than task-specific

robots. Components such as cameras and lidar sensors can push

manufacturing costs for humanoid robots to between $50,000 and

$200,000 per unit, compared with $5,000 to $100,000 for

task-specific machines, according to industry executives and a

market study by startup Standard Bots.

"(True) general-purpose robots have not really been invented

yet," said Marc Theermann, strategy chief at Boston Dynamics,

adding that "if somebody claims that they are commercially

finding a general-purpose robot, they are over-promising and

they will under-deliver."

While the company's Atlas robot has made headlines with

viral clips of it leaping, flipping and dancing, Theermann said

the value currently lies in targeted designs. Its four-legged

robot, Spot, excels in hazardous industrial inspections, a

narrow but lucrative niche.

Era Ventures has backed ViaBot, whose autonomous machines

manage trash collection in parking lots. Parkway Venture Capital

has diversified its bets between Siera AI's forklift automation

and Figure AI's humanoid ambitions.

"You'll see a transition where there will be robots built

for a task doing something very useful, very cost-effectively,"

said Raja Ghawi, Partner at Era Ventures. "And as that gets

better, people will realize there is a good reason to have a

full humanoid."