Individuals, educators, makers
Creator
The BENTO IDE with Flow, Blocks, Python and the emulator. Models from the Store installed on your own board, each licensed to that board.
Sovereign Edge AI
Intelligence you own, where it runs.
BENTO covers the whole path for secure edge AI. With it you train on your own site and install models into running devices from a browser, while raw data stays where it was made. The same control holds for one device and for a whole enterprise.
Why now
AI is moving into the things around us.
Microcontrollers now carry neural processors; Infineon's PSOC™ Edge E84, for example, pairs an Arm® Cortex®-M55 with an Ethos™-U55 NPU. Regulation now asks device makers to account for the software they ship. The tools for building, delivering and governing AI on these devices have not kept up.
TinyML devices shipped in 2030
2.5B
ABI Research forecast, 2021
Connected IoT devices at the end of 2025
24.1B
IoT Analytics, 2026 outlook
EU Cyber Resilience Act applies in full
11 Dec 2027
Regulation (EU) 2024/2847; reporting duties from 11 Sep 2026
Sovereignty at the edge
Sovereignty that starts at the device.
Most sovereign-AI programmes focus on data centres. Ours begins with the device itself and keeps the data, the model and the decision with their owner.
Data stays where it is made
Inference runs on the device and training on your own server. Raw data stays on site, and only results travel, as Highlights. Data residency is a property of the design, so it does not depend on a contract clause.
Models you own, trained on your own data
Train on your own sounds, machines, rooms and languages. Every model still has to pass quality gates that cannot be waived silently.
Independent of any one cloud
Models run offline on the chip's own NPU, so there is no per-inference cloud bill and no API key at runtime. Installs come from a browser.
Built to be audited
Models are signed and sealed to one board on the BENTO HMI Kit, every install is tied to its board, and every model passes its gates before release. Designed to support the duties of the EU Cyber Resilience Act.
Sovereignty here means control of data, models and deployment. The silicon and toolchains, from Infineon and Arm, come from a global supply chain, as they do for everyone.
Installs like a song
Models are songs. Scenes are playlists.
Choose a model in the Store and it installs into the running firmware from your browser, without a reflash. Put several together into a scene for a room or a production line, much as you would make a playlist.
- Today
A model installs from the browser into the running firmware, with no reflash.
- Today
Several models make up a scene for one room or one line, like a playlist you put together yourself.
- Today
The board reports what it can run, and the Store explains why when a model cannot.
How value is exchanged
How BENTO is licensed.
Like silicon IP, BENTO is licensed where it runs, per device and per model, with prices in US dollars. Enterprise and Sovereign deployments are set by agreement.
Product teams, integrators
Studio
Everything in Creator, plus scenes (playlists of several models) across the boards your team manages.
Manufacturers, operators
Enterprise
The Edge AI Trainer on your own site, so raw data stays there. Installs across your fleet, Highlights, and custom models and scenes by agreement.
Governments, regulated sectors, national programmes
Sovereign
The whole path under your roof: the Trainer and a local Store on your premises, your own model catalogue, models trained on your languages and environments, and a record of what is installed where. By agreement.
Prices are in US dollars, per device and per model, and Enterprise and Sovereign deployments are by agreement. Purchases are made in each product; until online checkout opens, every licence starts with a conversation.
How it scales
From one room to a whole operation.
The same path that installs one model on one board also delivers scenes to every line in a plant and every site in a network.
- 01One deviceA model installed from the browser.
- 02One roomA scene of several models, composed like a playlist.
- 03One lineThe same scene on every board along a line.
- 04A fleetInstalls across sites, each tied to its board.
- 05A sector or a nationThe whole path under one roof: Trainer, Store and catalogue.
What runs today
Every claim names where it runs.
Each line below is in the products today, scoped to the board or the service it applies to.
- Today
Models install from the browser into the running firmware with no reflash (Edge AI Store, over USB).
- Today
On the BENTO HMI Kit, models are sealed to one board and stored encrypted, with board identity over mutual TLS.
- Today
Secure boot and a closed debug port, set on every BENTO device we deliver.
- Today
Training behind fourteen quality gates on your own server (Edge AI Trainer, in pilot, by invitation).
- Today
The BENTO IDE was used by about 200 students in a single session, September 2026.
Origin
Designed in Thailand for markets worldwide.
BDH Corporation is the parent company behind BENTO. Design and engineering are headquartered in Thailand; regional hubs for markets and partners are planned around the world. BENTO is the first venture, and the company is set up to take on more.
For investors and partners
Build the sovereign edge with us.
We are building BENTO for the long term with a small number of partners. They share our view that AI should belong to the people and places that create it. If you invest in deep technology, ship devices at scale or run a national or sector AI programme, we would like to talk.
A briefing is available on request.