> ## Content Index
> Fetch the complete content index at: https://www.tao.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# Dyna Robotics Crosses the ROI Threshold as Din Tai Fung Scales Its Robot Deployment
- URL: https://www.tao.media/dyna-robotics-crosses-the-roi-threshold-as-din-tai-fung-scales-its-robot-deployment/
- Published: 2026-08-31T11:49:37.000Z
- Updated: 2026-08-31T11:49:37.000Z
- Description: Dyna Robotics says its robots have crossed commercial ROI. Din Tai Fung, the highest revenue-per-location restaurant chain in the US, is rolling the fleet out network-wide, with hundreds of robots targeted across hotels, logistics, and data centers by 2027.
- Author: Antonio Verrico
- Tags: Dyna Robotics, Robotics

[Dyna Robotics](https://www.dyna.co/?ref=tao.media) has crossed the commercial return-on-investment threshold for its manipulation robots, producing enough at high enough quality that customers economically benefit from buying one. Din Tai Fung, the highest revenue-per-location restaurant chain in the US, is rolling the robots out across its full network of locations and commissaries.

The Din Tai Fung expansion runs alongside Dyna's new deployments in hotels, logistics centers, and data centers, with the combined fleet expected to reach hundreds of robots by the first half of 2027\. Dyna calls the milestone the first case of a robot foundation model maker taking a manipulation pilot all the way to scaled commercial deployment.

> After so many demos, models, pilots, robots are still struggling to land real deployments with real customers. Until now.  
>  
> Today we’re excited to share that our robots have successfully crossed the ROI threshold, and Din Tai Fung, one of the highest revenue per location… [pic.twitter.com/CAWz7dKkPh](https://t.co/CAWz7dKkPh?ref=tao.media)
> 
> — Dyna Robotics (@DynaRobotics) [August 27, 2026](https://x.com/DynaRobotics/status/2093039146357178475?ref%5Fsrc=twsrc%5Etfw&ref=tao.media)

Waymo co-CEO Dmitri Dolgov, quoted in the announcement, described the frequently talked about 'robotics gap' cleanly: "The demo took 18 months; the product took about 15 years." For Dyna, the gap is everything beyond a good demo video. 

Recovery from failure, uptime, integration with a customer's workflow, and reliability sustained across months of daily use all sit on the far side of the line. Dyna calls itself a research-driven product company built around that gap, turning each field fix into reusable tooling or a more general model instead of a one-off patch.

At Din Tai Fung, each robot has to produce 1,500 table-ready napkins across an 18-hour shift. Dyna-1, released in June 2025, folded roughly 35 napkins an hour with a 75% pass rate against the customer's quality standard, totaling about 480 napkins a day. Dyna-2, the current model, folds 95 napkins an hour at a 93% pass rate, reaching 1,590 table-ready napkins a day, 2.7 times the earlier model's hourly rate.

The earlier model also struggled once a napkin left the folding stage, dropping each one into whichever bin slot happened to be open and leaving staff to re-stack them. The ten bin positions look nearly identical to a camera, so fixing this required the model to follow an explicit instruction rather than infer placement from appearance. Dyna-2 now places each napkin into one of ten designated stacks based on the assigned instruction, generalizing across different bin sizes without a separate model for each site.

A lab test lasting an hour confirms a model works, but it won't catch a model that fails once every two hundred trials, degrades gradually over a week, or performs well on one robot and badly on the next. Dyna instead grades every episode a deployed robot produces against the customer's own quality bar, continuously, at every site. Its deployed fleet generates more than a terabyte of camera and sensor data every day, run through an automated system labeling which step of a task failed and why.

0:00 

/0:13 

1× 

Factory & Assembly Mode

At one Din Tai Fung site, throughput dropped for no reason visible in the logs. The labeling system traced the cause to the step where the robot pulls a single napkin off the stack. Total failures at the site roughly doubled over two weeks, and missed grabs accounted for most of the increase, climbing from 108 to 380 incidents in the same window. The culprit turned out to be a worn gripper rather than a model regression.

The same instrumentation tracks mechanical wear across the fleet over time, flagging a degrading component before it fails and letting Dyna schedule maintenance instead of waiting for a breakdown mid-shift.

The rollout builds on Dyna-1, which was the most reliable robot foundation model in production at the time of its release. The gaps exposed during the pilot, on throughput and on bin placement, are what Dyna-2 was built to close.

Dyna treats each fix as reusable rather than site-specific. A failure mode found at one deployment becomes a check the company runs across the whole fleet, and a hardware issue diagnosed on one robot protects every other robot running the same components. That compounding effect is why a new site now goes from setup to meeting Dyna's production ROI bar in as little as three days.

Read their full announcement below:

[Not Just a Model, But a Product](https://www.dyna.co/research/scaling-customer-deployments?ref=tao.media)