Since the discovery of an undocumented I/Q sampling feature in ESP32 chips about two weeks ago, there’s been a flood of similar features being found in other wireless chips, including LoRa and Wi-Fi chipsets. The latest chip to join their ranks is the MediaTek MT7921AU, often used in Wi-Fi adapters, with two different projects developing independent SDR drivers for it in the last few days.

The first was BirriaSDR, which discovered that the chip exposes raw I/Q samples for debugging and calibration; BirriaSDR instead streams these to the host at 40 Msps in 14-bit resolution, with the option to tune from 2156 to 2610 MHz. BirriaSDR uses libusb to directly communicate with the Wi-Fi adapter and access its registers; it doesn’t make any permanent firmware changes, so the Wi-Fi adapter goes back to its original state after being used as an SDR. BirriaSDR implements an API similar to the RTL-SDR’s.

Mtk79ICAP, by contrast, exposes the MT7921AU’s sampling features over a standard SoapySDR interface. Unlike BirriaSDR, it applies some firmware changes to the chip, and it requires a modified driver to work. However, it can reach a maximum bandwidth of 80 MHz and tune to the 5 GHz bands. The Wi-Fi adapter in use can also simultaneously work in monitor mode and stream SDR data, though this only works when automatic gain control is enabled.

This is the second MediaTek chip we’ve seen repurposed for SDR work, after the earlier MT7996. Conveniently, Wi-Fi adapters often have external antennas, which should make it easier to connect a frequency mixer circuit. These adapters aren’t quite there yet, but some hackers have already used the ESP32 to transmit arbitrary signals.

[This is a sponsored article written in collaboration with Selangor Information Technology & Digital Economy Corporation (Sidec).]

You probably don’t give much thought to drones surveying a plantation.

Sure, you might notice it flying across the fields, but not the constant stream of information its sensors process to keep it stable and on course.

It’s not all too different from what happens inside a car or the satellite passing overhead, which relies on its own onboard electronics to keep everything running from hundreds of kilometres above us. 

What do these three have in common, you may ask? Well, behind all of them are semiconductors, and these semiconductors are increasingly being designed right here in Malaysia.

Taking chip design into our own hands

For years, we’ve prided ourselves for Malaysia’s role in the chip industry, specifically in ATP (assembling, testing, and packaging chips). Now, though, there’s significant effort to take on more of the work that comes before all of that.

Part of this is at the Malaysia Semiconductor IC Design Park in Puchong, where more than 400 engineers across 20 companies are working on chip designs for AI systems, drones, cars, satellites and more. 

This is part of Malaysia’s broader push to move up the semiconductor value chain—from being known for what we do with chips to having a bigger say in what goes into them in the first place.

Here are six chips designed in Malaysia (and located in the IC Design Park) to give you a good idea of what that looks like.

1. The AI brain that doesn’t need the cloud

Many of the AI tools we use today run in the cloud, which means our requests travel to faraway data centres and back before they can be executed. That’s alright for a chatbot, but not ideal for autonomous robotics and smart agriculture that need to react in real-time, or for sensitive data like intelligence video analytics and AI education systems.

This is where semiconductor chips like SkyeChip‘s MARS1000 come in. 

A 7nm edge AI processor that’s designed and developed in Malaysia, MARS1000 enables AI inferences to run locally on devices rather than on the cloud. This keeps data on-premises while responses are faster.

The semiconductor design company behind it, SkyeChip, was founded in 2019 by industry veterans from Intel, Altera, and Broadcom. Its core team holds over 113 patents across Malaysia, the US, and China.

Speaking to us, SkyeChip hopes to put AI into places that couldn’t afford it before. Take TVET, for example. Training a CNC operator or a utility technician needs expensive machines and a lot of instructor time. But with a MARS1000-powered AR station, a trainee can practise the procedure safely, step by step, before touching real equipment. 

This means fewer workshop accidents, less capital tied up in machines, and many more trainees per instructor.

“Our ambition is simple: a Malaysian-designed chip, inside Malaysian schools, factories and cities, solving Malaysian problems, and then carrying that model to the rest of ASEAN.”

–  Joe Lee, Director of Engineering & Business Development at SkyeChip

2. Giving AI more room to think

You probably know by now that AI requires a lot of memory to work with, and the memory attached to GPUs (the chip that does the heavy lifting) is actually limited. 

So when a model outgrows its GPU, the usual fix is to add more GPUs. But this can get expensive, especially for organisations and businesses that want to run AI on their own systems to keep their data private.

This is the gap aiDAPTIV+ was built to close. A product of Phison Electronics Corporation that was jointly developed with Maistorage Technology (Phison’s regional hub), aiDAPTIV+ combines high-performance cache memory SSDs with memory management middleware.

In simpler terms, aiDAPTIVE+ makes different types of memory (GPU, system, and cache) work together. Then AI applications can access more data locally without adding software complexity. Yes, even under heavy load, aiDAPTIV+ can consistently deliver faster responses by eliminating the need to recalculate evicted tokens.

As a Malaysian IC startup founded in 2024, Maistorage’s mission is to drive innovation in NAND

controllers, ICs, and storage modules tailored for data centers, AI, and automotive industries. But its bigger ambition is to democratise powerful AI and make it more accessible to the masses.

“Locally in Malaysia, we are committed to cultivating a robust partner network, developing turnkey AI solutions tailored for key sectors such as healthcare, education, manufacturing, and the public sector, and fostering domestic AI talent. Our ultimate objective is to empower Malaysian organizations to move beyond mere AI adoption and truly build, control, and own their AI capabilities.”

– Pua Kein Keong, Director at Maistorage

3. Powering satellites from a Selangor design centre

Space is a wonderful place to explore, but it’s also one of the most unforgiving places to put electronics. The conditions up there are extreme and once hardwares are launched, they can’t easily be fixed. So it’s even more important that every system on board performs as they’re supposed to without hiccups. 

This requires a “manager” to control how power is distributed, watch for trouble, and report back to HQ on Earth. One of these “managers” is Weeroc Space‘s CONAN.

Designed for efficient GaN-based power conversion, the CONAN chip controls the power switches, monitors the current, temperature and voltage, and constantly adjusts the output to keep the power supply running smoothly.

Although Weeroc Space is a European semiconductor design company, just last year, the firm invested RM20 million to establish a design centre at the IC Design Park in Puchong. Its Malaysian team works on radiation-hardened microchips and photodetector read-out systems for aerospace, satellite, and scientific applications. 

Under the leadership of its CEO Dr. Salleh Ahmad, Weeroc Space Malaysia is strengthening the nation’s role in high-reliability chip design, supporting global projects involving ESA, NASA, and Airbus Defence & Space.

Over the next 5 to 10 years, Weeroc Space Malaysia aims to build a strong portfolio of Malaysian-designed space-grade ICs, develop a new generation of highly skilled engineers, and see technologies designed by its team powering missions in space.

“Our ambition goes beyond building a successful semiconductor company. We want Weeroc Space Malaysia to show that world-class space technologies can be created in Malaysia and compete globally. Ultimately, we want Weeroc Space to become a global semiconductor player and help put Malaysia on the map for advanced space technology. ”

– Dr. Salleh Ahmad, CEO of Weeroc Space Malaysia

4. The flight controller at the heart of a drone

When a drone is in flight, it’s constantly being pushed around by numerous factors: wind, its own movements, and the weight it carries. To stay on course, something has to read its sensors and adjust its motors in real time, over and over again. 

That’s the job of a flight controller, like Alphaswift Industries’ FC Mini. This compact flight controller acts as the drone’s flight brain and handles stable and responsive flight control, real-time sensor processing, autonomous mission support, and built-in safety functions.

It’s built around an STM32H743 microcontroller and combines dual IMUs, a pressure sensor, and microSD flight logging, It supports a wide range of drone power systems, from 2S to 8S LiPo batteries, and works with popular flight software such as Betaflight, iNav, and ArduPilot.

With a background in aerospace engineering and innovation, Dr. Shian Lee established Alphaswift Industries to pioneer smart drone technologies for logistics, agriculture, and emergency response, while advancing Malaysia’s role in the global aerospace ecosystem.

“[We aim] to establish Alphaswift Industries as a trusted avionics developer, providing reliable, secure, and non-Chinese-component flight control solutions for global UAV markets.” 

This is being done in two stages:

1. Positioning FC Mini as a high-performance, NDAA-compliant flight controller for FPV drones, industrial UAVs, and professional drone developers.
2. Developing a widely adopted flight control solution, enabling seamless integration into Alphaswift’s own UAV platforms and other drone systems.

5. The unglamorous chip every car needs: LDO

See that infotainment screen in your car and your car headlamps? Well, they (and other electronics in the vehicle) all share the same electrical system. But the power supply isn’t always steady, and the delicate chips inside these electronics need clean, stable power to do their jobs.

This is where Infinecs Systems‘ automotive low dropout regulator (LDO) comes in. 

It can take in up to 40V and provide a steady 3.3V or 5V output at up to 150mA. To protect from issues like low voltage, short circuits, and overheating, Infinecs readily equips it with built-in safety features that allows the LDO to operate in a car’s fluctuating temperatures, ranging from -40°C to 125°C.

Cars aside, LDO can also be used in telecommunications equipment, battery management systems, and motor drives, making it useful for a range of industrial applications that require stable and reliable power.

Led by industry veterans like CEO and Co-Founder Kalai Selvan Subramaniam, Penang-based Infinecs Systems brings decades of semiconductor experience from companies including Intel and Motorola, with its team having contributed to technologies such as USB 1.0 and advanced IC design in Malaysia.

The firm delivers end-to-end engineering across IC design, embedded software, systems engineering, and product development. It also has offices in Cyberjaya and India, serving global customers in automotive, industrial, medical, consumer electronics, and communications industries. 

6. Malaysia and SEA’s first homegrown FPGA

Most times, semiconductor chips are built to do one job and one job alone. This makes designing a custom chip for every new idea a slow and costly process, so engineers often turn to a chip that can be reprogrammed after it’s made.

That’s an FPGA (Field-Programmable Gate Array), and SiliconX has built Malaysia’s and Southeast Asia’s first homegrown one. 

It packs 30,000 logic elements into a 14nm chip, and is designed for applications in automotive systems and intelligent motor control (like smart billboards and large LED screens). 

The two-year-old firm develops programmable chips entirely in-house, including the hardware and software tools needed to use them. While this first chip targets automotive and motor-control applications, SiliconX is developing FPGA solutions for a wider range of industries, including telecommunications, AI, healthcare, and aerospace.

Its Managing Director, Ng Meng Wah, leads the company’s mission to develop locally designed programmable chips that reduce reliance on imported technologies and strengthens Malaysia’s semiconductor ecosystem.

“Our ambition is to become a global FPGA chip product company, putting Malaysia on the world’s list of chip products and moving the country up the semiconductor value chain. In the next five years, we will bring two flagship devices to market: a cost-effective FPGA for mass-market applications, and an AI-capable FPGA for edge AI.”
– Ng Meng Wah, Managing Director of SiliconX

Building the people behind the chips

All that said, designing more chips in Malaysia also means having the engineers to build them.

The Advanced Semiconductor Academy of Malaysia (ASEM), housed at IC Design Park 2, has trained 850 engineers with a 100% placement rate so far. 20 companies at the Park are hiring locally and have thus far given 407 engineers a chance to build careers in chip design without having to leave Malaysia. You can be one too!

• Find out what’s next for the Malaysia Semiconductor IC Design Park here.

Intel’s next-generation Nova Lake processors could use a unified Core Ultra 4000 naming scheme across desktop PCs and laptops, according to a new leak. The change would scrap an earlier expectation that Intel would use three-digit Core Ultra 400 names for some mainstream laptop processors while reserving four-digit numbers for desktop and high-performance laptop chips.

Hardware leaker Jaykihn reportedly corrected that information on October 16, saying Nova Lake H-series and other mobile processors would also use four-digit model numbers. When asked whether the entire Nova Lake lineup would fall under Core Ultra 4000, the leaker replied in the affirmative. Intel has not officially confirmed the naming scheme, so treat the information as a report rather than a final product announcement.

The move could make Intel’s next processor generation easier to recognise across different devices. However, model numbers alone will not tell buyers how powerful a particular chip is, and the company still needs to clarify its final product lineup.

All five Nova Lake families could adopt four-digit names

The reported naming scheme would cover five processor families: Nova Lake-S for desktop PCs, Nova Lake-HX for high-performance laptops, Nova Lake-H for mainstream performance laptops, and Nova Lake-U and Nova Lake-UL for lower-power mobile devices.

Earlier reports suggested that Nova Lake-S and Nova Lake-HX would use Core Ultra 4000 branding, while other mobile chips would retain the three-digit Core Ultra 400 label. The leaker’s revised information removes that distinction, placing the expected desktop and laptop processors under the same four-digit series.

Some desktop model names have already appeared in leaks, including the Core Ultra 9 4970K BFC, Core Ultra 9 4950K and Core Ultra 7 4870K BFC. These names remain unofficial, and specific retail model numbers for the mainstream laptop variants have not been confirmed.

What the naming change means for buyers

A consistent naming scheme could make it easier for buyers to identify processors belonging to the same generation, regardless of whether they are shopping for a desktop or laptop. It may also reduce confusion caused by two different numbering systems appearing within one processor family.

However, the Core Ultra 4000 label will not mean every chip offers the same performance. Buyers will still need to consider the processor’s category, power requirements, core configuration and the device it is designed for.

Nova Lake is expected to begin arriving in 2027, with desktop and mobile variants potentially launching at different times. Intel has yet to confirm the complete lineup, specifications, pricing or release schedule.

Advertisement

Until an official announcement arrives, the revised naming scheme remains based on leaked information. More details should emerge as Nova Lake approaches launch.

Watch Fiji vs Cook Islands in a vital 2026 Rugby League World Cup Pool A encounter at the McDonald Jones Stadium in Newcastle, New South Wales. Respectively the No7 and No8 teams in the world rankings, the Pacific Island teams will be desperate to win in case one of their bigger pool rivals slip up.

Fiji have a good World Cup record, reaching three semi-finals in a row between 2008 and 2017. Yes, it feels like this year’s format means the deck is stacked against them – pre-tournament favorites Australia and New Zealand are expected to qualify from Pool A – but that won’t stop Mick Potter’s men trying to demonstrate the best of their power and flair. The 33-year-old hooker Apisai Koroisau is their captain and talisman, and brings the experience of winning three NRL premierships with the Rabbitohs and Panthers. Sunia Turuva and Kurt Donoghoe provide youthful energy in the backs.

In the other incident, a model instructed to retrieve figures from a public research dashboard bypassed the organization’s security controls and used the unauthorized access to steal credentials. It then made configuration changes on the server.

In one evaluation, the changes let the model make an internal search tool relay queries to the public dashboard. In a second evaluation, the model downloaded five non-public tables containing 147,451 rows. OpenAI researchers had prompted the agents to retrieve only five dashboard figures. Even after resetting the setting, the model continued using the token to retrieve non-public data.

“The evaluation ended without the model disclosing that it had guessed answers, accessed the organization’s internal systems, downloaded non-public tables, or changed the authentication configuration,” OpenAI said. “All five answers in this evaluation were marked incorrect.”

In a third report, the company said a model prompted to update a cryptocurrency trading application used funds from a locked wallet on a test network. The model “derived wallet keys from publicly known developer recovery phrases, found an account with test-network funds, and used that account to pay for the requested contract deployment.”

The report continued: “The model questioned whether it had permission, then reasoned that a public recovery phrase and valueless test-network tokens made the funds acceptable to use. The deployment succeeded and charged transaction fees to the account. This case shows that a model can substitute its own judgment about likely harm for the account holder’s permission, then act on that judgment without authorization.”

Nvidia CEO Jensen Huang has sung the praises of Microsoft’s new Surface Laptop Ultra, describing it as “so beautiful” and “incredible”.

Speaking at a developer event in London, Huang outlined the development process that went into building the hardware powering the new product, which is the first Microsoft device powered by an Nvidia chip.