New 3D silicon chip stacks circuits on top of each other to boost computing power 29%

By Rich McEachran0%

7/16/2026, 9:25:00 AM

BS Summary: This article contains 25 faulty reasoning types, including Ambiguity (Equivocation), False Dilemma, and Appeal to Authority, with Optimism Bias as the most egregious example at 22.3% saturation with 216 hits. Analysis detected 1,797 faulty-reasoning hits from 969 analyzed words, generating a BS Score of 39.1% and a BS Rank of 29% (15,716 of 21,887 articles). This article is better (less manipulative) than 71.80% of the article peer group.

The massive hardware demands of artificial intelligence (AI) applications are stretching the physical and structural limitations of semiconductors. 
But researchers have engineered a three-dimensional silicon chip that they propose as the solution. 
In a new study published May 27 in the journal Nature , scientists found a way to cram more computing power into a chip by stacking silicon circuits in multiple layers in a way that doesn't impact performance. 
Stacking chips vertically, known as 3D integration, is more efficient than traditional 2D chips, where silicon circuits are spread across a single surface. 
This is because stacking shortens the distance that data has to travel and reduces the power required for data transmission. 
The researchers' 3D chip uses ultrathin silicon membranes and low-temperature manufacturing techniques to overcome the challenges of current chip architectures. 
"Our method is not only easier to implement with lower cost, but it has several advantages over previous approaches to stack silicon wafers," Qing Cao , first author of the study and a materials science and engineering professor at the University of Illinois Urbana-Champaign, said in a statement . 
Extending Moore's law 
Since the 1960s, ensuring that electronics can handle more demanding applications has meant making transistors smaller so more can be packed onto a single chip. 
But, as Cao pointed out, doubling the number of transistors every couple of years  a principle known as Moore's law  is becoming less feasible. 
"If you look at the actual size of transistors, they're not getting smaller, especially in terms of their contacted gate pitch," Cao said in the statement  defined as the combined width of one transistor gate and the space needed to separate it from the next. 
"This is because we're becoming limited by the intrinsic material properties of silicon and the fundamental rules of quantum mechanics . 
If we're going to keep up the trend of increasing processing power of our microprocessors, we have to start thinking beyond just squeezing more devices on a single surface." 
The researchers think vertical integration across multiple layers is the best way to guarantee that engineers can continue to adhere to Moore's law, because this approach creates room for more transistors on a chip. 
"Today it takes six microelectronic devices called transistors on a single plane to store one bit of information," Cao explained, suggesting that just like in a densely populated city, the only way to solve overcrowding is to build upward. 
"You get the same functionality, but the spatial footprint is reduced while making communication between layers faster and more efficient." 
Scientist Gordon Moore seen with a graph representing Moore's Law. 
(Image credit: Intel ) Getting around the heat problem 
Stacking is nothing new, of course, but vertical integration  building layers directly on top of one another  can create thermally dense packages. 
In the study, the researchers noted that the fabrication of high-quality silicon chips demands temperatures up to 1,832 degrees Fahrenheit (1,000 degrees Celsius). 
However, once the first chip layer has been completed, the metal wiring introduced to connect further layers can be destroyed by such high temperatures. 
As a result, the "thermal budget"  the maximum amount of heat that can be endured before degradation starts to occur  for any additional layers is 752 F (400 C), said Cao. 
This can result in performance and reliability issues. 
When creating 3D stacked silicon chips, manufacturers have sought to avoid this problem by using alternatives to single-crystalline silicon for the upper layers, according to the researchers. 
These materials include amorphous and nanocrystalline metal oxides, carbon nanotubes and polycrystalline silicon, but they can lead to performance and reliability issues, the scientists said in the study. 
To overcome this challenge, Cao and his team adopted an approach called "monolithic integration"  a process in which all chip components are fabricated on a single piece of substrate, as opposed to making them separately and then bonding them together later. 
To build each chip, the researchers created ultrathin silicon nanomembranes that they then transferred, using a roll laminator, onto a substrate containing the bottom layer. 
Scientists say they've eliminated a major AI bottleneck  now they can process calculations 'at the speed of light' 
Scientists trained an AI model using an IBM quantum computer  and it answered questions correctly that the base model couldn't 
What's the biggest bottleneck to building better AI? 
It's no longer the lack of computing resources  it's generating enough energy to feed it 
The maximum temperature required to generate a strong bond using this method was just 392 F (200 C)  five times less than the heat normally required. 
The membranes they transferred were also just 10 nanometers thick or less  about the size of a protein  compared with the approximately 500-to-700-micrometer (500,000 to 700,000 nanometers) thickness of a typical wafer. 
Because they are thin, these membranes are mechanically flexible to conform to the underlying surface, Cao added. 
The result of this process was a 3D chip with three layers, each containing 625 transistors. 
This pales in comparison to the billions of transistors that can be crammed onto chips already on the market, but the researchers believe their technology boasts power efficiency benefits. 
The electrical current that can flow through the chip has proved to be at least three to four times greater than that of monolithic chips made from alternative materials. 
The big question is whether their 3D silicon chip can make the leap from the laboratory to commercial applications. 
While the research demonstrates the potential of a chip comprising three stacked layers, the scientists suggested that plenty more layers can be added in future iterations. 
Can you match these ancient devices to their pictures? 
Find out with our computing quiz! 
Article reasoning-pattern comparisonThis article: 1.4%Rich McEachran: 0.4%Live Science: 2.7%Confirmation Bias1.4%This article: 2.8%Rich McEachran: 0.7%Live Science: 1.2%Anchoring Bias2.8%This article: 8.8%Rich McEachran: 2.2%Live Science: 2.7%Availability Heuristic8.8%This article: 6.1%Rich McEachran: 1.5%Live Science: 1.4%Representativeness Heuristic6.1%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.5%Hindsight Bias0.0%This article: 9.0%Rich McEachran: 5.0%Live Science: 3.0%Overconfidence Bias9.0%This article: 11.8%Rich McEachran: 2.9%Live Science: 3.3%Framing Effect11.8%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.5%Loss Aversion0.0%This article: 0.3%Rich McEachran: 0.1%Live Science: 0.4%Status Quo Bias0.3%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.2%Sunk Cost Effect0.0%This article: 22.3%Rich McEachran: 11.5%Live Science: 3.5%Optimism Bias22.3%This article: 4.6%Rich McEachran: 1.2%Live Science: 1.2%Pessimism Bias4.6%This article: 11.0%Rich McEachran: 2.8%Live Science: 3.3%Negativity Bias11.0%This article: 5.1%Rich McEachran: 4.5%Live Science: 0.6%Self-Serving Bias5.1%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.4%Fundamental Attribution Error0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Actor-Observer Bias0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.3%In-Group Bias0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Out-Group Homogeneity Bias0.0%This article: 6.1%Rich McEachran: 1.5%Live Science: 1.3%Halo Effect6.1%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Horn Effect0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Dunning-Kruger Effect0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.9%Recency Bias0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.3%Primacy Effect0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Blind-Spot Bias0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Ad Hominem0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Straw Man0.0%This article: 15.7%Rich McEachran: 4.9%Live Science: 4.2%Appeal to Authority15.7%This article: 17.3%Rich McEachran: 7.9%Live Science: 1.1%False Dilemma17.3%This article: 2.7%Rich McEachran: 0.7%Live Science: 0.4%Slippery Slope2.7%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Circular Reasoning0.0%This article: 10.4%Rich McEachran: 3.9%Live Science: 3.8%Hasty Generalization10.4%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.3%Red Herring0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.3%Bandwagon0.0%This article: 2.0%Rich McEachran: 1.0%Live Science: 2.3%Appeal to Emotion2.0%This article: 5.7%Rich McEachran: 1.4%Live Science: 0.5%Begging the Question5.7%This article: 4.5%Rich McEachran: 1.1%Live Science: 2.3%Post Hoc (False Cause)4.5%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Tu Quoque0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.4%Burden of Proof0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.5%Appeal to Nature0.0%This article: 2.6%Rich McEachran: 0.6%Live Science: 0.3%Composition/Division2.6%This article: 6.2%Rich McEachran: 1.5%Live Science: 1.8%Anecdotal6.2%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%No True Scotsman0.0%This article: 20.4%Rich McEachran: 5.1%Live Science: 1.7%Ambiguity (Equivocation)20.4%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Gambler’s Fallacy0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Middle Ground0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Personal Incredulity0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Special Pleading0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.1%Genetic Fallacy0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 1.5%Unattributed Quote0.0%This article: 2.1%Rich McEachran: 0.5%Live Science: 1.0%Quote-first Misdirection2.1%This article: 5.1%Rich McEachran: 1.3%Live Science: 3.5%Biased Writer Voice5.1%This article: 0.0%Rich McEachran: 0.0%Live Science: 1.0%Indoctrination0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Politically Left Leaning Bias0.0%This article: 0.0%Rich McEachran: 0.0%Live Science: 0.0%Politically Right Leaning Bias0.0%This article: 1.5%Rich McEachran: 0.5%Live Science: 1.6%Attempt to Sell a Product or S…1.5%

969 words analyzed.

Speakers

2speakers30%attributed speech680writer words
Voice mapSelect a segment to jump to its words
Writer's voice • 15 words • 0.0% coverageWriter's voice • 18 words • 0.0% coverageWriter's voice • 14 words • 0.0% coverageWriter's voice • 38 words • 0.0% coverageWriter's voice • 23 words • 0.0% coverageWriter's voice • 20 words • 0.0% coverageWriter's voice • 20 words • 0.0% coverageQing Cao • 49 words • 100.0% coverageWriter's voice • 3 words • 0.0% coverageWriter's voice • 25 words • 0.0% coverageQing Cao • 26 words • 0.0% coverageQing Cao • 46 words • 0.0% coverageQing Cao • 21 words • 0.0% coverageQing Cao • 29 words • 0.0% coverageWriter's voice • 34 words • 0.0% coverageQing Cao • 39 words • 0.0% coverageQing Cao • 20 words • 100.0% coverageWriter's voice • 10 words • 0.0% coverageIntel • 9 words • 0.0% coverageWriter's voice • 24 words • 0.0% coverageWriter's voice • 23 words • 0.0% coverageWriter's voice • 24 words • 0.0% coverageQing Cao • 33 words • 0.0% coverageWriter's voice • 8 words • 0.0% coverageWriter's voice • 27 words • 0.0% coverageWriter's voice • 28 words • 0.0% coverageWriter's voice • 42 words • 0.0% coverageWriter's voice • 25 words • 0.0% coverageWriter's voice • 19 words • 0.0% coverageWriter's voice • 21 words • 0.0% coverageWriter's voice • 8 words • 0.0% coverageWriter's voice • 16 words • 0.0% coverageWriter's voice • 27 words • 0.0% coverageWriter's voice • 34 words • 0.0% coverageQing Cao • 17 words • 0.0% coverageWriter's voice • 16 words • 0.0% coverageWriter's voice • 29 words • 0.0% coverageWriter's voice • 29 words • 0.0% coverageWriter's voice • 19 words • 0.0% coverageWriter's voice • 26 words • 0.0% coverageWriter's voice • 9 words • 100.0% coverageWriter's voice • 6 words • 100.0% coverage
Selected voice

Qing Cao

94%flagged-word coverage
280 attributed words97% of attributed speech91% writer coverage
0%10.0%20.0%Biased Writer Voice+17.5 ptsWriter: 0.0%Qing Cao: 17.5%17.5%Quote-first Misdirection+7.1 ptsWriter: 0.0%Qing Cao: 7.1%7.1%Attempt to Sell a Product -2.2 ptsWriter: 2.2%Qing Cao: 0.0%0.0%

Attribution is sentence-level. Pattern percentages are calculated only from words assigned to that voice.

Loading…
Loading…
Loading…
Loading…

Analysis

Hover over highlighted words in the article to view the associated bias or fallacy analysis.