HONG KONG — As artificial intelligence data centers threaten to overwhelm regional power grids and push current microchips to their physical limits, researchers in Hong Kong have engineered a new type of transistor that sidesteps the primary thermal barrier that has hindered computing for decades [1, 2].
The breakthrough, published in Science by a team at The Hong Kong Polytechnic University, introduces a quantum tunneling field-effect transistor built from atomically thin sheets of bismuth and indium selenide [1, 3]. The device flips cleanly between on and off states using just 160 millivolts of gate voltage [1, 2].
Because conventional silicon transistors require between 700 and 800 millivolts to do the same job, this new architecture slashes the electricity wasted during computing by roughly 95 percent [1, 2]. Put into practical terms: a standard 60-watt high-performance computer chip could perform the exact same work while drawing less than 3 watts of power, eliminating noisy cooling fans, boiling heatsinks, and massive data center energy bills.
Escaping the “Boltzmann Tyranny”
Every conventional microchip in use today relies on thermionic emission [1]. Inside standard silicon transistors, electrical current is controlled by raising and lowering an energy barrier [1]. To turn the switch on, electrons rely on ambient heat to jump over that barrier [1].
Because heat naturally spreads energy unevenly among particles, physics dictates a fundamental law known as the Boltzmann limit: at room temperature, it takes a fixed minimum electrical push to boost current tenfold [1, 3]. This means classical silicon switches require at least 700 to 800 millivolts to shut off completely without leaking electricity [1, 2].
Years ago, chip designers ran straight into this wall [1, 2]. They could make transistors smaller, but they could no longer lower the voltage without causing the chips to leak current and overheat [1, 3]. The result was the modern “thermal wall” and the emergence of “dark silicon”—a reality where large sections of your laptop or phone processor must remain turned off at any given moment to keep the chip from melting itself.
Tunneling Through the Wall
Instead of forcing electrons to jump over a barrier, the new device relies on quantum tunneling [1, 3].
According to quantum mechanics, subatomic particles can behave like waves and pass straight through solid physical boundaries if the barrier is narrow enough [1, 3]. By stacking two-dimensional layers of bismuth and indium selenide, the researchers created an atomic gateway [1, 3]. Applying just 160 millivolts opens that gateway, letting electrons tunnel horizontally across the channel [1, 2].
Conventional Transistor (Thermionic)
Gate Voltage: ~800 mV
Electrons must hop OVER a high energy barrier
Result: Massive heat, requires fans and heatsinks
Quantum Tunneling Transistor (TFET)
Gate Voltage: ~160 mV
Electrons tunnel straight THROUGH the thin barrier
Result: Cold operation, works on battery power
Earlier experimental tunneling transistors suffered from a fatal flaw: trickle currents [1, 3]. While they could switch with low voltage, the stream of electrons was too weak to quickly flip neighboring gates, making them too slow for real-world computers [1, 3].
The Hong Kong team bypassed this bottleneck by using a pristine 2D interface [1, 3]. The crystalline layers fit together naturally without broken atomic bonds or defects to snag moving charges, delivering robust, usable electrical drive currents at room temperature for the first time [1, 3].
What a 95% Power Cut Means in Practice
The energy used when a computer switches a transistor doesn’t drop evenly with voltage—it drops exponentially [1, 2]. Because energy scales so aggressively with voltage, cutting the operating push from 800 millivolts down to 160 millivolts results in a massive dividend [1, 2]:
| Device Specification | Standard Advanced Silicon | 2D Quantum Tunneling Chip |
| Operating Gate Swing | ~800 millivolts | 160 millivolts [1, 2] |
| Power of a 60 W Chip | 60 watts | ~2.5 to 3.5 watts |
| Cooling Method | Loud fans, copper pipes, water chillers | Silent chassis cooling (no fan) |
| Laptop Battery Life | ~1 hour under heavy gaming/AI load | All-day continuous run time |
| Primary Limitation | Overheats under high density | Needs new mass-fabrication equipment [1] |
For an everyday laptop, this means workstation-grade computing that stays cool to the touch and runs for days on a single battery charge. For massive AI data centers—which currently require entire cooling towers and power substations—it means slashing cooling infrastructure down to bare minimums.
The Road Ahead: Overcoming the Current Paradox
While the physics is proven on prototype wafers, bringing a 160-millivolt chip into commercial production presents major manufacturing hurdles [1].
- The Current Paradox: When you drop the voltage drastically, moving the same massive volume of raw computing work requires drawing higher overall current across the chip’s internal power lines. Engineers will need to adopt backside power delivery—routing power cables through the back of the silicon wafer rather than the top—to keep internal microscopic wires from overheating.
- Clock Speeds vs. Massive Parallelism: Because quantum tunneling delivers slightly lower peak current than the most heavily boosted silicon transistors, these chips are unlikely to break raw single-core speed records for desktop gaming. Instead, they are tailored for massively parallel processors, such as AI neural engines and graphics processors that run thousands of lightweight cores side-by-side at moderate speeds.
- Foundry Adoption: Moving this technology from research labs into major foundries like TSMC or Intel requires developing industrial coating tools that can spread two-dimensional bismuth and indium selenide across foot-wide silicon wafers without a single microscopic rip or fold [1].
If foundries can master that manufacturing process, the 160-millivolt switch will usher in a silent, low-energy era for high-performance computing, sidestepping silicon’s oldest thermal limit for good [1, 2].
References
- Bytivo Technology Editorial Board, “Hong Kong PolyU Quantum Tunneling Transistor Breaks MOSFET Limit,” Bytivo Tech News, September 1, 2026.
- Quantum Zeitgeist, “Hong Kong PolyU’s Quantum Leap Overcomes A Key Barrier To AI,” Quantum Zeitgeist Semiconductor Analysis, September 1, 2026.
- Hao, J., et al., “Tunnel field-effect transistors exhibiting sub-Boltzmann-limit switching via 2D bismuth/indium selenide heterojunctions,” Science, Vol. 393, Issue 6812, pp. 910–916, August 27, 2026.
- IEEE, “International Roadmap for Devices and Systems (IRDS): Beyond CMOS and Steep-Slope Logic Devices,” IEEE Standards Association, 2024–2026 Editions.
- Ionescu, A. M., & Riel, H., “Tunnel field-effect transistors as energy-efficient electronic switches,” Nature, Vol. 479, pp. 329–337.