Skip to content

Topic

Edge AI

Running machine-learning inference directly on local devices—phones, sensors, robots—rather than in remote data centres, under tight power and latency limits.

Current stories

buildOne report1 publisher

Scaleout's loitering munition spent most of its 320-second mission choosing a target

In BAE Systems Bofors' January demonstration, the Swedish startup's onboard stack ranked objects, picked an armored engineering vehicle and released an explosive with no external link. The board details come from a different flight.

Reality

Evidence40
Adoption15
Hype gap+30
Incentives60
Confidence55
buildOne report1 publisher

GlobalFoundries pins an early-stage edge-AI process to its 1.1 billion euro Dresden expansion

GlobalFoundries topped out its 1.1 billion euro Dresden expansion on October 9, aiming for more than one million 300mm wafers a year by end-2028. FDX Fusion, the edge-AI process it announced alongside, is a far earlier bet that still needs German federal approval and customer qualification.

Publishers:runtimewire.com

Reality

Evidence45
Adoption15
Hype gap+20
Incentives65
Confidence50
productOne report1 publisher

Lola Vision Systems targets the 200 hours of setup between an AI model and a new chip

Lola Vision Systems will license its compiler on existing chips, aiming at what its founder says can be roughly 200 hours of manual setup per new chip. Cheaper porting would let edge teams test a second board, though the company has not said how many of those hours its software saves.

Publishers:techcrunch.com

Reality

Evidence25
Adoption10
Hype gap+30
Incentives80
Confidence35
investConfirmed5 publishers

Onsemi's $5.7 billion cash bid for Synaptics lands 16% above the market price

Onsemi will pay $123 a share in cash, about $5.7 billion, for Synaptics after a rival bid, replacing a June stock deal valued near $7 billion. The cash price sits 16% above Synaptics' Thursday close, so holders gain on the market even as the headline total shrinks.

Perspective Coverage

6 publishers
Builder
Builder 17%
Operator
Operator 23%
Investor
Investor 60%

Reality

Evidence72
Adoption
Not scored
Hype gap+15
Incentives70
Confidence70
scienceOne report1 publisher

Prototype shoe runs its onboard AI on electricity harvested from footsteps

Simiao Niu's team built a battery-free shoe that runs AI on footstep power, sorting four activities with 95.4 percent accuracy. Charge-free gait monitoring for conditions like Parkinson's is the goal, though so far the shoe only labels activity, counts steps and estimates calories.

Reality

Evidence45
Adoption
Not scored
Hype gap+30
Incentives
Not scored
Confidence40
investConfirmed4 publishers

SiMa.ai raises $150M at a $1.45bn valuation with Fidelity and Michigan backing edge silicon

SiMa.ai closed an oversubscribed $150M Series C co-led by Fidelity and Amplify at a $1.45bn post-money valuation, with J.P. Morgan and the State of Michigan among new backers. The round bets on chips that run AI on devices rather than in data centres.

Perspective Coverage

4 publishers
Builder
Builder 29%
Operator
Operator 20%
Investor
Investor 51%

Reality

Evidence50
Adoption35
Hype gap+45
Incentives70
Confidence60
productOne report1 publisher

Nvidia's entry-level robot brain doubles to 78 TOPS and keeps its 8 GB ceiling

Jetson Orin Nano 2 takes the entry seat in edge robotics to 78 trillion operations per second, but not until 2027, and the memory figure is what decides which small models ship.

Publishers:siliconangle.com

Reality

Evidence30
Adoption10
Hype gap+35
Incentives65
Confidence40
scienceOne report1 publisher

Nanjing's LightTok chip produces AI tokens inside the sensor pixel

The LightTok chip senses, stores and computes in the same pixel, and its authors report 87.3% recognition accuracy with a tenfold energy gain on turning light into tokens across an array of 1,024 pixels.

Publishers:livescience.com

Reality

Evidence46
Adoption7
Hype gap+32
Incentives62
Confidence44
productOne report1 publisher

Raytron puts the thermal alarm's interpretation step inside the camera

At CIOE 2026 the company showed an infrared vision-language model trained on more than five million of its own thermal samples, built to describe gas plumes and electrical faults on the camera's own chip.

Reality

Evidence28
Adoption12
Hype gap+35
Incentives70
Confidence42
scienceOne report1 publisher

Retina-inspired sensors are the part of neuromorphic computing that has already shipped

The 20-watt brain frames the ambition of brain-inspired hardware. The working examples are event cameras tracking debris in orbit and a low-power processor sold by Australia's BrainChip, running alongside GPUs.

Reality

Evidence45
Adoption32
Hype gap+25
Incentives70
Confidence50
scienceOne report1 publisher

A Nature Perspective puts sixteen years of self-assembling memristor work under one name

The Perspective calls them self-organizing memristive networks and applies mean-field theory, graph theory and disordered-systems physics to their conductance dynamics, with the energy saving still stated as a motivation.

Publishers:nature.com

Reality

Evidence58
Adoption
Not scored
Hype gap+25
Incentives40
Confidence60
scienceOne report1 publisher

Bristol mathematicians prove random yes-or-no questions can sort millions of classes

Each classifier answers one randomly chosen question and trains in minutes on a laptop; the Bristol team says enough of them can pick one item out of millions, with no coordination between them.

Publishers:phys.org

Reality

Evidence34
Adoption
Not scored
Hype gap+34
Incentives62
Confidence56
leadershipOne report1 publisher

The Army is buying GPS-free navigation from sensors its vehicles already carry

TERN's $11.26 million contract puts offline positioning and fleet-health monitoring on sensors the Army already owns, on a premise its president states plainly, that battlefield GPS can no longer be assumed.

Reality

Evidence34
Adoption33
Hype gap+27
Incentives78
Confidence56
investOne report1 publisher

Analog Devices is paying $1.35bn in cash for a neural front end already in production

Analog Devices has disclosed a price for Alif Semiconductor, $1.35 billion of upfront cash with up to $200 million contingent, without disclosing anything about what Alif earns, so the comp everyone copies has no denominator.

Publishers:prnewswire.com

Reality

Evidence48
Adoption30
Hype gap+35
Incentives85
Confidence55
buildOne report1 publisher

Analog Devices pays up to $1.55B to put an Arm NPU next to its own sensor front ends

The Ensemble parts run Arm's Cortex-M55 cores and Ethos-U55 NPUs, so what Analog Devices is buying for up to $1.55B is the integration around them and a distribution channel Alif could not build alone.

Publishers:runtimewire.com

Reality

Evidence60
Adoption28
Hype gap+22
Incentives72
Confidence56