The future of artificial intelligence may depend less on smarter algorithms...and more on how electricity enters a transistor.
The Hidden Bottleneck
Imagine building the world's fastest highway, only to force every vehicle through a narrow toll gate. That's exactly what happens inside today's computer chips.
Not anymore.
The industry has spent more than half a century chasing one simple goal: make transistors smaller.
That relentless pursuit gave us smartphones that outperform yesterday's supercomputers, AI accelerators capable of trillions of calculations every second, and electronic devices that fit into the palm of our hands.
Yet, as engineers approached the atomic scale, something unexpected happened. The transistor itself was no longer the biggest problem. The bottleneck had quietly shifted elsewhere.
Every electronic device—from a smartwatch to an AI supercomputer—depends on electricity entering a semiconductor before any computation can begin. That journey starts at the contact where a conductive metal meets the semiconductor.
Although this interface measures only a few nanometres across, it behaves like a traffic bottleneck on an otherwise empty highway. Electrons slow down. Some lose energy. Others scatter unpredictably.
The consequences ripple across the entire chip. More heat. Higher power consumption. Reduced efficiency. Ironically, as transistors become smaller, this microscopic interface consumes an even larger share of the device's total energy.
For years, engineers accepted this limitation as an unavoidable trade-off. They refined transistor architectures. Invented new manufacturing techniques. Explored exotic semiconductor materials. But the electrical doorway itself remained largely unchanged.
The industry wasn't running out of better transistors. It was running into an increasingly stubborn entrance problem. That hidden bottleneck has quietly shaped the limits of modern computing.
Now, a team of researchers believes the solution isn't to build a better doorway. It's to eliminate the doorway altogether.
Why It Stayed
If contact resistance has been such a well-known problem, why hasn't anyone solved it before?
The answer lies in the way modern electronics have always been built. Traditional transistors rely on two fundamentally different materials. One is a metal, responsible for delivering electrical current quickly and efficiently. The other is a semiconductor, designed to carefully regulate that current so the transistor can switch on and off billions of times every second.
Bringing these two materials together sounds straightforward. At the atomic level, it is anything but.
Each material has its own crystal structure, electronic properties and energy landscape. When electrons move from one into the other, they don't experience a smooth transition. Instead, they encounter a sudden change that creates resistance, much like a car slowing as it leaves a paved highway for a rough gravel road.
For decades, engineers tried to reduce this resistance. They experimented with new contact metals. Modified semiconductor surfaces. Refined fabrication techniques. Developed sophisticated interface engineering methods. Each improvement helped—but only incrementally.
The challenge became even greater with the rise of two-dimensional semiconductors. These atomically thin materials promise faster, smaller, and more energy-efficient electronics. Yet their delicate structure makes conventional metal contacts even more difficult to integrate without damaging the material or degrading its performance.
In other words, the very materials expected to power the next generation of electronics were being held back by the same old interface problem.
Researchers weren't simply looking for a better contact. They needed to rethink the concept of electrical contact from the ground up. That required asking a different question: what if electricity never had to cross from one material into another in the first place?
A Smarter Approach
Sometimes, solving a decades-old engineering problem doesn't require a new material. It requires looking at an old material differently.
That is exactly what researchers at the Korea Advanced Institute of Science and Technology (KAIST) have done.
Instead of asking, "How can we build a better electrical contact?", they asked a more fundamental question: what if the contact itself became part of the semiconductor?
That simple shift changes the entire design philosophy of modern electronics.
For more than fifty years, semiconductor devices have followed the same blueprint. A metal electrode is placed on top of a semiconductor so electricity can enter the device. No matter how carefully that junction is engineered, electrons still encounter a physical boundary where two completely different materials meet. That boundary has always been the source of unwanted resistance.
The KAIST team decided to remove the boundary rather than improve it.
Their solution uses (PtSe₂), an ultra-thin with an unusual property. Instead of behaving the same way throughout, its electrical characteristics change naturally depending on how thick it is.
That single property made something previously impossible suddenly achievable. Instead of combining two materials, the researchers built two different electronic functions inside one continuous crystal.
The result is known as a —a seamless structure where electricity no longer needs to jump across a conventional material junction.
It's a remarkably elegant idea. Not because it adds complexity, but because it removes it.
This is often how engineering moves forward. The biggest leap isn't creating another component. It's finding a way to eliminate one altogether.
One Material, Two Roles
At first glance, platinum diselenide looks like any other advanced semiconductor material. Its real strength lies beneath the surface—or more precisely, within its thickness.
Unlike silicon, whose electrical behaviour remains largely fixed, PtSe₂ changes character as its atomic layers change.
Add more layers, and it behaves like a , allowing electrical current to flow freely. Reduce it to only a few atomic layers, and it becomes a semiconductor, capable of switching electrical signals on and off.
The chemistry remains the same. Only the thickness changes.
Think of it like water. At different temperatures, the same substance can exist as ice, liquid water, or steam. Its chemical identity never changes, but its behaviour changes dramatically.
PtSe₂ follows a similar principle in the electronic world. Instead of temperature, the controlling factor is atomic thickness.
The researchers exploited this unique behaviour with extraordinary precision. They fabricated thicker regions to serve as electrical pathways and thinner regions to function as active semiconductor channels.
Because both regions belong to the same uninterrupted crystal lattice, electrons travel through a continuous atomic structure rather than crossing from one material into another. That continuity is the real innovation.
Traditional semiconductor contacts resemble two separate roads connected by a bridge. No matter how well the bridge is designed, traffic must still slow down.
The KAIST design removes the bridge entirely. The road simply changes its characteristics while remaining unbroken.
For electrons, that means a smoother journey. Fewer interruptions. Less wasted energy. And a far more efficient route into the transistor.
Sometimes nature already provides the solution. The challenge is recognising it.
Following the Current
An elegant theory is only valuable if experiments prove it works. This is where the KAIST team's research becomes especially compelling.
Rather than relying solely on computer simulations or electrical measurements, the researchers chose to observe the phenomenon directly.
To achieve this, they turned to (AFM)—one of the most powerful tools available for exploring matter at the nanoscale.
Unlike an ordinary microscope that uses light, AFM employs an ultra-sharp probe capable of scanning surfaces atom by atom. But the researchers extended its capabilities even further.
By combining AFM with in-plane current mapping, they could do something rarely achieved in semiconductor research: they watched electrical current move across the interface in real time.
The resulting images revealed something remarkable. Instead of piling up at the boundary or scattering unpredictably, electrons flowed smoothly from the semimetallic region into the semiconducting region.
There was no obvious electrical barrier. No sudden interruption. No visible bottleneck.
For the first time, researchers had direct experimental evidence that a monolithic semimetal-to-semiconductor interface could support uninterrupted charge transport.
That observation matters because it confirms the underlying concept—not just the device's overall performance. It shows why the architecture works.
The team went one step further. By applying an external electric field, they demonstrated that the semiconducting region could still switch electrical current on and off, just like a conventional transistor.
This is a crucial milestone. A good semiconductor contact isn't enough on its own. Future chips also need reliable switching behaviour. The prototype successfully delivered both.
It's a small experimental device today. But it represents something much larger. For decades, engineers have tried to make the doorway smoother. These researchers proved that the doorway itself might no longer be necessary.
Beyond Miniaturization
For decades, the semiconductor industry's answer to every challenge was remarkably consistent: make everything smaller.
Smaller transistors meant faster processors. Lower power consumption. Greater computing power packed into the same space.
This philosophy—popularly known as —drove one of the most extraordinary technological revolutions in human history.
But every revolution eventually reaches a point where physics begins pushing back. Today's transistors are already approaching dimensions measured in just a few nanometres. At this scale, shrinking them further becomes increasingly difficult, expensive and unpredictable.
Heat rises faster. Leakage currents become harder to control. Manufacturing costs climb dramatically.
The industry's next leap is no longer guaranteed to come from miniaturization alone. Instead, engineers are beginning to focus on something equally important—making every electron travel more efficiently.
That's where this breakthrough becomes significant. Rather than attempting another round of transistor scaling, the KAIST researchers targeted an inefficiency that has quietly existed in every modern chip.
Reducing contact resistance doesn't replace transistor innovation. It amplifies it.
Imagine improving every doorway inside a city instead of widening every road. Even without building new highways, the entire transport network becomes faster and more efficient. The same principle applies inside a processor.
If electrical current enters each transistor more efficiently, the cumulative gains across billions of transistors become substantial.
For AI hardware, where power consumption has become one of the industry's biggest challenges, even small improvements at the transistor level can translate into enormous energy savings across entire data centres.
The next chapter of semiconductor innovation may therefore be less about smaller devices—and more about smarter architecture.
The Scale Test
Every laboratory breakthrough eventually faces the same question: can it leave the laboratory?
The KAIST team's work is an impressive proof of concept, but commercial success depends on solving a completely different set of engineering challenges.
The first hurdle is manufacturing. Creating precisely controlled thickness variations across tiny experimental devices is one thing. Producing billions of identical structures across large semiconductor wafers is another.
The semiconductor industry demands extraordinary consistency. A single microscopic defect can affect the behaviour of millions of transistors. Researchers must therefore demonstrate that this monolithic PtSe₂ architecture can be manufactured reliably, repeatedly and economically.
The second challenge is compatibility. Today's semiconductor ecosystem has been refined over decades around silicon-based CMOS technology. Any new material must integrate into existing fabrication processes without dramatically increasing production costs.
Then comes reliability. Processors inside smartphones, satellites, medical devices and autonomous vehicles must operate continuously for years under changing temperatures, electrical loads and environmental conditions. Experimental success over a few hours or days is only the beginning. Long-term durability still needs to be proven.
Finally, there is economics. Even if a technology works perfectly, manufacturers will adopt it only if the performance gains justify the additional complexity and investment.
History offers plenty of examples of brilliant laboratory discoveries that never reached commercial production. At the same time, many technologies that initially appeared impractical—from to extreme ultraviolet lithography—eventually became industry standards.
The future of this breakthrough will depend not only on scientific validation, but also on manufacturing innovation.
The Bigger Picture
Every era of computing has been defined by one invisible engineering breakthrough. Vacuum tubes gave way to transistors. Discrete components became integrated circuits. Silicon manufacturing evolved into .
Each transition seemed incremental at the time. Looking back, each one fundamentally changed the world.
The PtSe₂ breakthrough may or may not become the dominant technology of tomorrow. That question will be answered over the next decade.
But its importance extends beyond one material or one experiment. It represents a broader shift in semiconductor thinking.
For years, innovation focused on making existing designs smaller. Now researchers are beginning to redesign the architecture itself.
Instead of treating electrical contacts, semiconductor channels and conductive pathways as separate components, future devices may integrate these functions into a single engineered material. That is a profound change in design philosophy.
For India, this is also an important signal. As the country accelerates investments in semiconductor manufacturing, packaging and chip design, the next opportunity may not simply lie in fabricating today's technologies. It may lie in contributing to tomorrow's materials science.
Universities, research laboratories and semiconductor startups have an opportunity to explore atomically engineered materials, novel transistor architectures and post-silicon electronics before they become mainstream.
History often rewards those who recognise technological transitions early. Not every breakthrough changes the world. But every technological revolution begins with breakthroughs that seem surprisingly small.
This may be one of them.
Join the Signal
If the next leap in computing comes from redesigning the doorway rather than shrinking the room, where should the industry — and India's own semiconductor ambitions — place their bets: faster transistors, or smarter architecture?
Discussion