From General AI to Embedded AI
Why intelligence is moving out of the data centre and onto the device
What Embedded AI changes
AI is now part of daily life — ChatGPT, DeepSeek, intelligent assistants, sales and support bots. All of them run on massive infrastructure and enormous compute, hosted in the cloud. This is what we call General AI (GAI).
But as AI moves into real products, the requirements change: lower cost, lower power, smaller footprint, faster response. Sending every single inference back to the cloud stops making sense — technically or commercially. That is where Embedded AI (EAI) comes in.
Embedded AI (EAI) moves inference down onto the device: a microcontroller performs feature extraction and state judgement locally, and returns only conclusions and key features to the network instead of continuously uploading raw waveforms.
01
Not just better metrics — a different cost structure
Where Alinket fits
The marginal cost of cabling, connectivity and long-term operations falls with it. Equipment that was previously written off because the installation cost was too high becomes viable to monitor after all.
Alinket has long built IoT wireless connectivity products and solutions, with end-to-end delivery capability from wireless modules, gateways and bridges through to the cloud platform. On-device intelligence is the last piece we are adding to that chain — we are building an integrated smart sensing terminal that combines sensors, edge inference, wireless connectivity and the cloud.
Latency
Data compliance
Offline availability
Bandwidth & storage cost
Queueing and round-trip delay make deterministic
real-time control impossible
Queueing and round-trip delay make deterministic
real-time control impossible
Streaming raw waveforms makes link and storage cost scale linearly with device count
No network means no capability — remote sites and mobile assets are the most exposed
|
|
02
The Basics: From Cloud AI to Embedded AI
2.1 What is Cloud AI?
2.2 Key Characteristics of Cloud AI
2.3 What is Embedded AI?
2.4 Key Characteristics of Embedded AI
2.5 Building Blocks of Embedded AI
2.6 Cloud AI vs Embedded AI
Typical applications today:
Smart cameras &
appliances
Smart cameras &
appliances
Electronics
Communication protocols
Languages & tooling
AI frameworks
Microcontroller (MCU)
Sensors & actuators
Smart home & IoT
Agriculture
Retail & POS
Wearables & health monitoring
Industrial automation
ADAS & smart vehicles
Virtual assistants and service robots (Siri, XiaoAi, Tmall Genie, Alexa, Google Assistant)
Search, recommendation and content distribution systems
Generative AI (large language models such as ChatGPT, DeepSeek and Gemini)
Cloud-based medical imaging assistance and large-scale data mining
In the same way we distinguish general-purpose computers from dedicated devices, Cloud AI refers to large-scale AI systems running on high-performance servers, usually inside cloud compute centres. These systems are built to process massive datasets, generate insights, and drive applications for large enterprises and entire industries.
Cloud AI is powerful, but it has clear limits wherever real-time response, offline availability, tight power budgets or data compliance are required. And it is exactly at those limits that Embedded AI becomes decisive.
Embedded AI combines resource-constrained electronics, embedded software and artificial intelligence so that models run on small hardware: microcontrollers (MCUs), neural processing units (NPUs), digital signal processors (DSPs) or dedicated SoCs. Unlike Cloud AI, which depends on a compute centre, Embedded AI delivers local intelligence — the device can think, learn and decide with no internet and no cloud access.
It is not a future concept. It already runs the devices around us, and it is moving fast into industry and infrastructure:
Facial recognition, voice control, anomaly detection, personalised suggestions
Vital-sign tracking, fall detection, arrhythmia detection, sleep monitoring
Predictive maintenance, visual inspection, smart metering, energy optimisation
Lane assist, collision detection, AI dashcams, driver-state monitoring
Lane assist, collision detection, AI dashcams, driver-state monitoring
Microcontroller and sensor sets — the physical base for sensing and compute
I2C, SPI, UART and SDIO for integration and data return; Wi-Fi, BLE, Wi-Fi HaLow and cellular on the wireless side
C / C++ for embedded development, paired with Python-side model training and data analysis
TinyML, TensorFlow Lite for Microcontrollers, and automated on-device model generation tools
Core compute for signal acquisition, feature extraction and model inference
Input/output interface to the physical world (vibration, current, temperature, sound, pressure)
Security systems, energy optimisation, gas and water leak detection
Security systems, energy optimisation, gas and water leak detection
Smart vending machines, anomalous transaction detection, inventory tracking
Not a replacement — a complement
Embedded AI does not replace Cloud AI; it completes it. The device side handles real-time, low-power, privacy-sensitive judgement. The cloud side handles global modelling, long-term trend analysis and cross-device knowledge. Together they form a complete intelligent system — and the application space behind them is enormous.
03
Embedded AI at Alinket
A worked example: the ALX856B Wi-Fi 6 dual-band controller
In a real embedded AI product, on-device compute is only half the problem — connectivity is the other half. A TinyML model can be perfectly tuned and still fail as a product if results cannot get off the device, if the battery drains in two days, or if data is exposed on the link.
Our approach is to move the complex parts of connectivity — protocol stacks, certificate handling and hardware crypto — inside the wireless module. That frees the host MCU from networking overhead, so its compute, flash and RAM go to inference instead.
ALX856B · Wi-Fi 6 Dual-Band Controller
Frees up the host MCU. The protocol stack is on-board: TCP/IP, TLS and EAP-TLS handshakes are handled by the module, so the host's compute and memory go entirely to TinyML inference.
Makes always-on AI battery-viable. Wi-Fi 6 Target Wake Time (TWT) lets the station negotiate wake slots with the AP and sleep in between — continuous local inference no longer means continuous radio draw.
Handles density and heavy payloads. OFDMA and uplink/downlink MU-MIMO cut queueing latency in dense deployments; 1024-QAM lifts short-range throughput. 2×2 dual-band MIMO leaves headroom for medical imaging and machine vision.
Keeps privacy on the device. A hardware crypto engine with WPA3 and 802.1X/EAP means the device uploads inference results, not raw sensor data — compliance by design.
Four things it does for embedded AI:
Wi-Fi 6
802.11 ax/ac/a/b/g/n
2.4 GHz / 5 GHz
On-module application core
Hardware crypto engine
Dual-band 2×2 MIMO
WPA3 / EAP
Cortex-M4
04
Why Build with Alinket
Let's Talk About Your Embedded AI Project
Years in IoT
The last piece: on-device intelligence
Built for volume production
Full in-house chain from wireless modules, gateways and bridges to the cloud platform
An integrated smart sensing terminal — sensors, edge inference, connectivity and cloud in one
Alinket delivers the complete embedded AI stack — sensors, on-device inference, wireless connectivity and cloud platform. Talk to us about your project.
From module selection to volume delivery — our engineers can join your design review.
Protocol stacks, certifications, configuration tools and hands-on support to shorten time to market
Address: Room 221, No. 1, Lane 706, Wuxing Rd, Pudong New District, Shanghai, China
Website：http://www.alinket.com https://b2b.alinket.com
Partners
Contact us
Email: marketing@alinket.com
Facebook
LinkedIn
Twitter
Follow us
Tel：+86 152 1433 1412 (Phone / WeChat / WhatsApp)