Mini Capstone
The Benefits of Early Research For Non-Classical Computing
Before graduating from John Paul the Great Academy, I will complete my Senior Capstone Project. This project will involve working throughout my senior year on a detailed research paper on a topic of my choice, as well as delivering an oral presentation in which I will publicly defend my work before faculty, students, family, and other community members. In preparation for this, I have completed a Mini Capstone project, which included an oral presentation in my Speech and Debate class.
1
Introduction
Everyone wants speed, efficiency, and innovation in technology. Every few years, most people upgrade their phones to get the latest and greatest. Their new phone is faster and has a few cool new features. We all know old devices slow down over time, and new devices are always better. But what if that was not the case? What if devices stopped improving? What if they needed to be bigger, and draw more power, just to be a little bit faster?
As most people know, computers have billions of tiny components that produce signals which communicate either “1” or a “0.” The tiny devices that produce these signals are called “transistors.” The iPhone 16 Pro released in 2024, for instance, has eighteen billion transistors. The current fastest consumer graphics card, which costs over two thousand dollars, has over ninety-two billion transistors.1 As computer chips get smaller, faster, and more power efficient, they rely on the continual development of smaller transistors. However, transistors cannot get much smaller; we are approaching the size limit. Right now they are as small as 2 nanometers wide, smaller than the wavelength of visible light.2 Since classical computers are reaching the theoretical size limit of transistors, technology companies must invest more heavily into researching alternate computing options.
1 “NVIDIA GeForce RTX 5090 Specs.” 2024. TechPowerUp. October 3, 2024.
https://www.techpowerup.com/gpu-specs/geforce-rtx-5090.c4216.
2 Loeffler, John. 2022. “No More Transistors: The End of Moore’s Law.” Interestingengineering.com. February 13, 2022. https://interestingengineering.com/innovation/transistors-moores-law.
2
What is Moore’s Law?
Most people may have a general notion that the price to performance of computing devices increases over time. This concept is predicated on the consistent shrinking of transistors, which has been accurate for decades. This phrase was coined by Gordon Moore in 1965 when he noted that the number of transistors in a dense integrated circuit will approximately double every two years.3 For over sixty years, this concept has remained remarkably accurate. It is why tech companies can consistently release faster devices every year, without needing more room for processing chips. The more transistors in a device, the faster and more efficient it will be.
There are several reasons why the trend described by Moore’s Law cannot continue, the most concerning reason as to why we cannot continue to shrink transistors for much longer is that they are approaching the size of an atom. While some people say that Moore’s Law is “dead,” it still holds true for the very near future. Moreover, due to natural restrictions on the way classical electron-based transistors work, computers cannot get much faster while remaining the same size. One of these restrictions is the speed of light, which remains a constant since calculations cannot be performed faster than the speed of light. The only way to increase the speed of a computer is to increase the number of transistors, and therefore increase the number of simultaneous calculations that it can perform. This means that if we are unable to make transistors smaller, we cannot make faster computers without making them larger. Heisenberg’s Uncertainty Principle states precision at a subatomic level is limited due to uncertainty about some aspects of the particles. James Powell calculated that by 2036, Moore’s Law will be obsolete solely from the Uncertainty Principle. Another issue with transistors at a subatomic level is quantum tunneling. Quantum tunneling is a physics phenomenon that can lead to small
3 Kasiorek, Przemysław. “Moore’s Law Is Dead. Now What? | Built In.” builtin.com, November 10, 2021.
3
particles phasing through walls, something that is becoming more relevant as transistors get smaller and the walls holding electrons shrink. The smaller and thinner these barriers get, the more often electrons may tunnel entirely out of the transistor holding them.
Why does it matter?
-
Rendering
One of the plainest reasons for developing faster computers is to speed up the time it takes to render images. Almost all large-budget Hollywood movies use some form of CGI, or computer-generated images. CGI allows for more spectacular special effects and can also be much cheaper than building large, detailed sets for scenes. Without these effects, movies would lack much of their current realism and immersiveness. Animated movies, such as Luca, are made entirely with computer animations. The difference between current animated movies and the ones from the late 20th century is how the animations are made. Older animated movies were entirely hand-drawn, with many pictures being strung together to create the video. Modern ones are “rendered,” meaning a computer takes in all of the information about the environment, and does millions of calculations to determine how light bounces throughout the scene. However, these modern animated movies can take extensive time and resources to render. For example, a single frame from the movie Luca took fifty hours to render.4 Movies need to have at least twenty-four of those frames each second to appear as a smooth video to the human eye. This would take over eight hundred years of a fast computer rendering for a one-hundred-minute movie. Because of this, the rendering calculations are performed simultaneously across thousands of computers that make up a “render farm.” This is why special effects for movies are
4 “Rendering.” n.d. Sciencebehindpixar.org. Accessed April 30, 2025. https://sciencebehindpixar.org/pipeline/rendering.
4
so expensive. Rendering films requires vast amounts of power and expense to use this many computers.5 For example, Avatar: The Way Of Water had a virtual effects budget of over three hundred fifty million dollars. Rendering is also required even to display videos onto screens. The graphics cards of computers are used to perform graphical calculations, but even when the computer is just playing a video, it has to determine what exact color each pixel has to be. Another more striking example of this is the Las Vegas Sphere. It has a staggering resolution of sixteen thousand by sixteen thousand pixels. That is two hundred and fifty-six million total pixels. For it to be capable of displaying extremely high-resolution video on this giant screen, it uses one hundred fifty Nvidia RTX A6000 graphics cards. One of these cards costs upwards of five thousand dollars. That is over seven-hundred and fifty thousand dollars worth of graphics cards just to be able to display a video to a screen. Since computer rendering is so slow and expensive, it is clear that an alternative solution is needed.
2. Artificial Intelligence
Perhaps the most relevant reason as to why we need faster computers is Artificial Intelligence. AI requires a staggering amount of processing power to run. In short, traditional large language models perform vast numbers of matrix math calculations to generate user requests. While some new research is being done to eliminate matrix multiplication to allow for greater scalability, all AI models still require some level of parallel processing power to run efficiently, and many “smarter” models demand even more computing power. Especially with recent advances in AI, and its continual growth, higher-level computing will be a necessity in the not-so-distant future. The larger and smarter AI models become, the more processing power and electricity they demand.
5 Branch Education. “How Does Ray Tracing Work in Video Games and Movies?” YouTube, August 17, 2024. https://www.youtube.com/watch?v=iOlehM5kNSk
5
3. Gaming
Gaming has become an increasingly popular pastime over the past few years. Video games use computer processors to calculate everything taking place, and graphics cards to render the scene. While movies and videos are typically viewed at twenty-four or thirty frames per second, video games are played at much higher frame rates. For a game to be considered “playable,” it typically has to run at a minimum of sixty frames per second. A very smooth gameplay experience typically demands higher, and competitive games need frame rates in the hundreds to give the best competitive advantage. The reason that games are run at much higher frame rates is because they require constant user input rather than passive viewing. Any delay between the user input and it being displayed on the screen can be easily perceived. While some games are not difficult to run, others require more processing power than what is currently available. One of the most difficult-to-run publicly available games right now is Cyberpunk 2077. With all graphics settings set as high as possible, the fastest and most expensive hardware available struggles to get above thirty frames per second. Even Minecraft, one of the least graphically demanding games you can think of, isn’t able to run smoothly under some conditions. With path-traced shaders installed, it is easy to drop below ten frames per second under certain conditions, even with the latest hardware. This is unacceptable to play with by anyone’s standards. Though upscaling and frame generation can help, they often inadvertently reduce image quality and create visual artifacts. This example demonstrates the need for faster computers, even for mass consumer use.
4. Energy Savings
The faster computer chips get, the more power-efficient they become. If easy calculations are performed on both a faster and slower chip, the faster one uses much less power than the
6
slower chip, because it doesn’t need to work as hard. Any power efficiency gains are also two-fold. The more power any chip consumes, the more heat it creates. This heat is dispersed through heatsinks and fans. Therefore, if a chip consumes less power, it also reduces the power draw from the fans.
5. Intensive Software
Over time, electronic devices age. They slow throughout their lifespan and eventually need to be replaced. Especially after operating system updates, old chips struggle to perform the same tasks they used to do without issues. This is because operating systems get more intensive as time progresses. As more features are added to improve the user experience, the chip has to juggle more background tasks. If computers were unable to continue getting faster while still staying small, eventually they wouldn’t be able to keep up with software updates.
6. Simulations
One of the most practical arguments for investing in advancing computing capabilities is for faster, more in-depth simulations. Faster computing could allow for faster simulation of particles, allowing for better understanding of how different chemical compounds interact with each other and with the human body. This could lead to a mass discovery of safer, more effective drugs for a variety of purposes. It could also enable improvements to pattern recognition AI, helping it to become much more accurate at a variety of predictions, such as diagnosing patients much earlier. This technology could have untold benefits, and save countless lives.
What now?
Now that it is clear why faster computers would be beneficial and why classical computers cannot get much faster, I will cover some of the alternate computing options that have the potential for further advancement.
7
1. Specialized Architecture
Many computers already have some form of specialized architecture. For a long time, all graphical computations were done directly on the processor. The problem with this is that the processor is more suited for more difficult, complex equations, while rendering often requires millions of simple equations. This issue was resolved with the invention of the graphics card. The graphics card performs all of the transformations, translations, and shading required for three-dimensional rendering. While most processors have several powerful cores, graphics cards have thousands of weaker cores. This allows them to complete rendering much more efficiently, freeing up the processor for where it is more needed. This is the first instance of specialized architecture in computers. Now, further specialized architecture might be needed to continue aiding computer performance and efficiency. When artificial intelligence became more widespread a few years ago, the calculations were performed on graphics cards because of their ability to perform many calculations quickly, such as the matrix multiplication used for most AI models. Today new cores are being used for AI, called tensor cores. Tensor cores are specifically optimized for deep matrix multiplication used in modern models. Creating even more specialized cores for different computer tasks could increase efficiency and speed, delaying the need for the development of transistor miniaturization. One example of where this is needed is memory. While permanent data is stored on hard drives and solid-state drives, computers also use memory for immediate access of temporary data by the processor. But an issue that is beginning to emerge is the speed. Even though it only takes nanoseconds for the processor to access information from the memory, it can still be bottlenecked if a computer is storing large amounts of information on the memory. Specialized architecture is a simple solution to
8
improving computational efficiency6 and speed while still allowing room for further miniaturization of computer chipsets. Engineering and producing such architecture would require minimal time and resources if the microchip industry as a whole were to fully pursue it.
The only issue with relying on specialized architecture is that it would only slow the approach of the oncoming blockade that is Moore’s Law. What is required to go beyond Moore’s Law is either subatomic transistors or an entirely new method of computing. Even if such microchips were produced, they can still only get so much smaller. On top of the impermanence of this approach, the microchips produced would also certainly be more expensive. The more separate components need to be produced for a computer, the more the cost of the computer’s production will go up. Because of this, a more long-term solution is required.
2. Spintronics
Another method of computing that is currently being researched is spintronics. These computers use the spin of electrons in a transistor to determine its state, rather than the quantity of electrons. These computers could also take advantage of nanomagnetism. By manipulating and observing the spin and charge of electrons, these computers would store data in a revolutionary way, allowing for subatomic computing to become a reality. These properties also have several features that show high potential for use in neuromorphic computing. Neuromorphic computing is an approach to computing that mimics the way the human brain works.7 It entails designing hardware and software that simulate the neural and synaptic structures and functions of the brain to process information. Such computing could be
6 Badham, James. “Specialized Hardware Solves High-Order Optimization Problems with In-Memory Computing.” Techxplore.com. Tech Xplore, January 8, 2025. https://techxplore.com/news/2025-01-specialized-hardware-high-optimization-problems.html
7 Marrows, Christopher H., Joseph Barker, Thomas A. Moore, and Timothy Moorsom. “Neuromorphic Computing with Spintronics.” NPJ Spintronics 2, no. 1 (April 29, 2024): 1–7. https://doi.org/10.1038/s44306-024-00019-2
9
accomplished by spintronics, allowing for the advancement of edge computing. Edge computing applications, which entail the processing and storage of data at the source of its production (i.e., near where it is created), are now being applied to a growing number of technologies. The application of edge computing translates into devices that can collect, store and process data, such as smart watches, computers that analyze utility grid data, computerized security technologies, and other systems. This means that each chip not only processes and refines data, but also can store the result at its point of origin. Spintronics would also allow the creation of chips that are a combination of processors and memory, something that has been sought after for years. AI macros can also be made much more effective at mapping their outputs correctly onto the inputs through spintronics.
Spintronics could be the unique approach to computing that accelerates it beyond the upcoming barrier caused by subatomic uncertainty. It also has several inherent properties that allow for neuromorphic computing, which would be a massive breakthrough for artificial intelligence and neuroscience. However, it still has several unresolved issues that need to be worked out before it can be used for mainstream computing. Mainly, it is still in very early development, and it could be years until it is ready for large-scale production and is stable enough to fully replace classical computers.
3. Optical Computing
Optical computing, also known as photonic computing, is a rapidly growing area of interest in computing. It has enormous potential to be the perfect solution to overcoming Moore’s Law. It uses photons, rather than electrons, to perform calculations. Optical neural networks could be used to physically perform calculations, similarly to analog computers. This results in exceptional power efficiency, as well as much greater speed than traditional digital neural
10
networks.8 This increase is even more significant in deep neural networks, which have extra layers. Some small groups are already developing early prototypes for optical computers, one of which is called Lightmatter. They take advantage of the speed of light, using fiber optic-like connectors to send information between chips.9 This would be crucial to high-speed connections throughout massive server farms, such as the ones used by ChatGPT. The excitement surrounding photonic computing caused Lightmatter to raise hundreds of millions of dollars from investors in a very short time. They are currently valued at 4.4 billion dollars. Even still, this is just one company among others that are all developing photonic computing chips. Others include ANELLO, Quandela, and Luminous. Within just a few years, optical computing will likely be ready for mass use. On top of rapid calculations, this technology has many other potential uses. It can also be used to increase the signal-to-noise ratio in 5G transmissions and help with internet security.10 Despite the growing interest in this new computing method, it still has several issues that need to be resolved before it can surpass classical computers. For starters, the speed of light in silicon photonics is only about forty percent of the speed of light in air. This is a significant issue because the speed of the calculations performed is directly related to the speed of light. Also, because some optical computers are analog, they are unable to perform a variety of different calculations and are more limited to matrix multiplication. However, because optical
8 Kazuhiro Gomi. “Optical Computing: What It Is, and Why It Matters.” Forbes, September 10, 2024. https://www.forbes.com/councils/forbestechcouncil/2024/09/10/optical-computing-what-it-is-and-why-it-matters/
9 Winn, Zach. “Startup Accelerates Progress toward Light-Speed Computing.” MIT News | Massachusetts Institute of Technology, March 1, 2024. https://news.mit.edu/2024/startup-lightmatter-accelerates-progress-toward-light-speed-computing-0301
10 Zhang, Weipeng, Joshua C Lederman, Thomas Ferreira, Jiawei Zhang, Simon Bilodeau, Leila Hudson, Alexander Tait, Bhavin J Shastri, and Paul R Prucnal. “A System-On-Chip Microwave Photonic Processor Solves Dynamic RF Interference in Real Time with Picosecond Latency.” Light Science & Applications 13, no. 1 (January 9, 2024). https://doi.org/10.1038/s41377-023-01362-5.
11
computers have so much potential to be the future of neural networks, more research needs to be done. For them to outperform classical digital neural networks, more effort needs to be spent now on research and development. We need to start preparing now if we want photonic computing to be the next step in advancing the technology age, especially if we want it to be viable by 2036, the expected end of Moore’s Law.
4. Quantum Computing
Quantum computers have been in development for decades. They have been seen as the ultimate endpoint in computing since before they were even created. They use special particles called qubits. If two qubits are supercooled, they can be linked together, and as long as they remain cold enough, they will remain linked at any distance. This state is called quantum entanglement. Any entangled particles will spin the same way, almost as if they are the same particle in two different places. These entangled particles can be harnessed in quantum computers to perform calculations. Because qubits act in different ways than electrons, they can spin forwards and in reverse, but also both at the same time. This unique property of qubits allows for multiple calculations to be performed simultaneously. Since qubits can attempt many solutions at once, this makes them particularly effective at solving specific complex equations much faster than classical computers.11 One example of an algorithm designed to be solved faster on a quantum computer is Shor’s Algorithm. It is an algorithm designed to efficiently factor large integers. This specific algorithm would take exponentially longer on a classical computer than on a quantum computer. They are also particularly effective at solving complicated equations, such as non-deterministic calculus, where the solution is different each time you solve it. There is
11 Arpan Kumar Kar, Wu He, Fay Cobb Payton, Varun Grover, Adil S Al-Busaidi, and Yogesh K Dwivedi. “How Could Quantum Computing Shape Information Systems Research – an
Editorial Perspective and Future Research Directions.” International Journal of Information Management, October 27, 2024, 102776–76. https://doi.org/10.1016/j.ijinfomgt.2024.102776
12
currently a common misconception that quantum computers try all possible solutions to an equation at once. This isn’t quite true though. A great analogy for how they work is to imagine the unknown part of an equation as a lake. Quantum computers would ‘drop’ stones into this allegorical lake, causing ripples. They then find where the ripples intersect, at the solutions to the equation. This completely different method of computing is extremely quick at performing some calculations, such as the ones used in the medical field to crack the genetics of organisms. Quantum computers can also improve artificial intelligence by helping determine what information is important while the model is trained, thus improving its reliability and reducing incorrect outputs. Finally, quantum computers’ most intriguing use is for simulations. They would be much quicker at performing detailed simulations that would be otherwise impossible on classical computers, such as chemical simulations on the atomic level. These could be used to discover stronger and lighter materials, as well as make batteries safer and more power-dense. Qubits still aren’t a miracle though. They are still unstable, and have an extremely high chance of being read incorrectly because of how they function.12 Since qubits are always in superposition, meaning they are in multiple states of motion at once, they are very difficult to accurately observe. This unreliability only scales higher the more qubits are on a chipset. However, emerging research from Microsoft has shown that qubits can be manipulated into a topological superconductor state, where they are a flat wave rather than a round particle.13 This also prevents qubits from being knocked out of an entangled state by other particles. This allowed Microsoft to create the Majorana chip, which could be much smaller and more reliable than previous quantum computers. Quantum computing still isn’t a complete godsend, however. It still has major
12 Katabarwa, Amara, Katerina Gratsea, Athena Caesura, and Peter D. Johnson. “Early Fault-Tolerant Quantum Computing.” PRX Quantum 5, no. 2 (June 17, 2024). https://doi.org/10.1103/prxquantum.5.020101
13 Nayak, Chetan. “Microsoft Unveils Majorana 1, the World’s First Quantum Processor Powered by Topological Qubits - Microsoft Azure Quantum Blog.” Microsoft Azure Quantum Blog, February 19, 2025. https://azure.microsoft.com/en-us/blog/quantum/2025/02/19/microsoft-unveils-majorana-1-the-worlds-first-quantum-processor-powered-by-topological-qubits/
13
drawbacks and issues, and it is unlikely to ever be used in home computers. The chips need to be cooled to a temperature less than that of deep space to function at all, and they will likely never be able to perform all calculations faster than a classical computer can. They are a very powerful tool, but only for certain tasks. Despite this, it still has huge potential for the advancement of computing, and it shouldn’t be disregarded.
Conclusion
Since classical computers will not be able to continue increasing in performance whilst still decreasing in size, alternate computing options must be much more heavily researched and developed. If transistors continue shrinking at the rate they are, it will be only a little more than a decade before they hit their limit, signaling the maximum limit for the speed of computers. While chips may continue to get optimized and get small boosts in performance, it won’t be long afterward until they can’t get any faster without getting bigger. Since classical computers are reaching the size limit of transistors, technology companies must invest more heavily into researching alternate computing options.
Though there are several options for dodging the approaching limits of Moore’s Law and quantum uncertainty, not all of them are viable. While specialized architecture is a perfect immediate fix for the short-term improvement of computers, it isn’t viable as a permanent solution, and can only act as a temporary boost to computing power. It should be implemented, as it will take minimal development time to provide a substantial increase in computational efficiency and speed for many devices. Spintronics also seems to be the logical next step after classical computers, as both rely on the manipulation and observation of electrons to perform calculations and store data. It could very well be the technology that is present in almost all home computers in a couple of decades. Because of this, it should still absolutely be researched, but it
14
may not be ready in time. This technology is powerful, but it is not worth our full research attention at the moment. Rather, it would be most prudent to invest in developing technologies that are the most improved over our current systems, that being optical and quantum computers. Optical computers provide an exciting potential future for the development of artificial intelligence. They are perfectly geared towards performing the calculations used in deep neural networks, and they do so in an analog way, providing a massive boost to power efficiency and speed. In an AI-driven society, optical computers are the most logical and prudent choice for investing massive amounts of money and time into research. Optical computers would provide the necessary push to overcome the boundaries that classical computers are forced to abide by. The final, most unique option is quantum computing. Quantum computing, if it continues advancing at its current pace, will completely obliterate the current capabilities of any classical computer. They are capable of things that humanity has previously only dreamed of being able to do. Atomic-scale simulations, instantaneous calculations, and highly accurate and fast genetic cracking are only a few of the incredible breakthroughs that would be made possible by a quantum age in computing. For all of these reasons, quantum and optical computing are the most likely to effectively surpass the boundaries set on classical computers by Moore’s Law.
15
Works Cited
Aasen, David, Morteza Aghaee, Zulfi Alam, Mariusz Andrzejczuk, Andrey Antipov,
Mikhail Astafev, Lukas Avilovas, et al. “Roadmap to Fault Tolerant Quantum Computation Using Topological Qubit Arrays.” arXiv.org, February 17, 2025. https://arxiv.org/abs/2502.1225
Aghaee, Morteza, Alejandro Alcaraz Ramirez, Zulfi Alam, Rizwan Ali, Mariusz Andrzejczuk, Andrey Antipov, Mikhail Astafev, et al. “Interferometric Single-Shot Parity Measurement in InAs–Al Hybrid Devices.” Nature 638, no. 8051 (February 19, 2025): 651–55. https://doi.org/10.1038/s41586-024-08445-2
Arpan Kumar Kar, Wu He, Fay Cobb Payton, Varun Grover, Adil S Al-Busaidi, and Yogesh K Dwivedi. “How Could Quantum Computing Shape Information Systems Research – an
Editorial Perspective and Future Research Directions.” International Journal of Information Management, October 27, 2024, 102776–76. https://doi.org/10.1016/j.ijinfomgt.2024.102776
Badham, James. “Specialized Hardware Solves High-Order Optimization Problems with In-Memory Computing.” Techxplore.com. Tech Xplore, January 8, 2025. https://techxplore.com/news/2025-01-specialized-hardware-high-optimization-problems.html
Binder, Kurt, and Erik Luijten. “Monte Carlo Tests of Renormalization-Group Predictions for Critical Phenomena in Ising Models.” Physics Reports 344, no. 4-6 (April 1, 2001): 179–253. https://doi.org/10.1016/s0370-1573(00)00127-7
Branch Education. “How Does Ray Tracing Work in Video Games and Movies?” YouTube, August 17, 2024. https://www.youtube.com/watch?v=iOlehM5kNSk
Caballar, Rina, and Cole Stryker. “What Is Neuromorphic Computing? | IBM.” www.ibm.com, June 27, 2024. https://www.ibm.com/think/topics/neuromorphic-computing
Cambridge University. In the Cambridge Dictionary. No Date Provided. Accessed February 21, 2025. https://dictionary.cambridge.org/us/dictionary/english/non-deterministic
Fadelli, Ingrid. “A CMOS-Compatible Spintronic Compute-In-Memory Macro to Secure AI Edge Devices.” Techxplore.com. Tech Xplore, July 31, 2023. https://techxplore.com/news/2023-07-cmos-compatible-spintronic-compute-in-memory-macro-ai.html
Chow, James. “Quantum Computing in Medicine.” Medical Sciences 12, no. 4 (November 17, 2024): 67–67. https://doi.org/10.3390/medsci12040067
Dargan, James. “3 Most Important Advantages of Quantum Computing.” The Quantum Insider, June 19, 2023. https://thequantuminsider.com/2023/06/19/advantages-of-quantum-computing/
Giovanni Finocchio, Massimiliano Di Ventra, Kerem Y. Camsari, Karin Everschor-Sitte, Pedram Khalili Amiri , Zhongming Zeng. “The Promise of Spintronics for Unconventional Computing.” Journal of Magnetism and Magnetic Materials 521 (March 1, 2021): 167506. https://doi.org/10.1016/j.jmmm.2020.167506
González, Victor H, Artem Litvinenko, Akash Kumar, Roman Khymyn, and Johan Åkerman. “Spintronic Devices as Next-Generation Computation Accelerators.” Current Opinion in Solid State and Materials Science 31 (August 1, 2024): 101173–73. https://doi.org/10.1016/j.cossms.2024.101173
Hu, Jingtian, Deniz Mengu, Dimitrios C Tzarouchis, Brian Edwards, Nader Engheta, and Aydogan Ozcan. “Diffractive Optical Computing in Free Space.” Nature Communications 15, no. 1 (February 20, 2024). https://doi.org/10.1038/s41467-024-45982-w
Kasiorek, Przemysław. “Moore’s Law Is Dead. Now What? | Built In.” builtin.com, November 10, 2021. https://builtin.com/hardware/moores-law.
Katabarwa, Amara, Katerina Gratsea, Athena Caesura, and Peter D. Johnson. “Early Fault-Tolerant Quantum Computing.” PRX Quantum 5, no. 2 (June 17, 2024). https://doi.org/10.1103/prxquantum.5.020101
Kazuhiro Gomi. “Optical Computing: What It Is, and Why It Matters.” Forbes, September 10, 2024. https://www.forbes.com/councils/forbestechcouncil/2024/09/10/optical-computing-what-it-is-and-why-it-matters/
Marrows, Christopher H., Joseph Barker, Thomas A. Moore, and Timothy Moorsom. “Neuromorphic Computing with Spintronics.” NPJ Spintronics 2, no. 1 (April 29, 2024): 1–7. https://doi.org/10.1038/s44306-024-00019-2
Nayak, Chetan. “Microsoft Unveils Majorana 1, the World’s First Quantum Processor Powered by Topological Qubits - Microsoft Azure Quantum Blog.” Microsoft Azure Quantum Blog, February 19, 2025. https://azure.microsoft.com/en-us/blog/quantum/2025/02/19/microsoft-unveils-majorana-1-the-worlds-first-quantum-processor-powered-by-topological-qubits/
Nvidia. “Ray Tracing.” NVIDIA Developer, August 14, 2018. https://developer.nvidia.com/discover/ray-tracing
Reddy Thamma et al., Sankara. “A Comprehensive Evaluation and Methodology on Enhancing Computational Efficiency through Accelerated Computing Introduction.” Journal of Science Technology and Research, no. 5 (2024): 517. https://philpapers.org/archive/SANACE-7.pdf
“Rendering.” n.d. Sciencebehindpixar.org. Accessed April 30, 2025. https://sciencebehindpixar.org/pipeline/rendering
Sciencedirect.com. “Dennard Scaling - an Overview | ScienceDirect Topics,” 2018. https://www.sciencedirect.com/topics/computer-science/dennard-scaling
Shen, Yichen, Nicholas C. Harris, Scott Skirlo, Mihika Prabhu, Tom Baehr-Jones, Michael Hochberg, Xin Sun, et al. “Deep Learning with Coherent Nanophotonic Circuits.” Nature Photonics 11, no. 7 (June 12, 2017): 441–46. https://doi.org/10.1038/nphoton.2017.93
Smith-Goodson, Paul. “Google’s 105-Qubit Willow Chip Achieves Major Quantum Milestones.” Forbes, January 28, 2025. https://www.forbes.com/sites/moorinsights/2025/01/28/googles-105-qubit-willow-chip-achieves-major-quantum-milestones/
Winn, Zach. “Startup Accelerates Progress toward Light-Speed Computing.” MIT News | Massachusetts Institute of Technology, March 1, 2024. https://news.mit.edu/2024/startup-lightmatter-accelerates-progress-toward-light-speed-computing-0301
Zhang, Weipeng, Joshua C Lederman, Thomas Ferreira, Jiawei Zhang, Simon Bilodeau, Leila Hudson, Alexander Tait, Bhavin J Shastri, and Paul R Prucnal. “A System-On-Chip Microwave Photonic Processor Solves Dynamic RF Interference in Real Time with Picosecond Latency.” Light Science & Applications 13, no. 1 (January 9, 2024). https://doi.org/10.1038/s41377-023-01362-5