The Capital Expenditure (CAPEX) of U.S. hyperscale and AI companies remains steady, driven by consistent demand and monetization. AI is enhancing efficiency across industries, driving financial returns that sustain strong demand. Lowering the cost of AI—much like the shift from mainframes to minicomputers and PCs—has not shrunk the market but has significantly expanded it. Microsoft’s CEO referencing Jevons Paradox was no coincidence. The paradox is named after William Stanley Jevons, a 19th-century English economist, who observed that technological improvements that increased the efficiency of coal use led to the increased consumption of coal in a wide range of industries. “It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption.” (W. Stanley Jevons, “The Coal Question,” 1865)

DeepSeek claims have not been independently verified but appear to be based on sound theoretical foundations. Their announcement was strategically timed, as the next generation of models is anticipated within approximately 12 months. DeepSeek claims that they have distilled existing models that are evolving toward more compact and efficient representations, thereby reducing computational needs. They claimed they spent around USD 6m to train their model, which is more than 10x lower cost than models from Meta (Facebook). They achieved this capability with lower performance NVIDIA GPUs which are also lower cost. Due to export control, they could not use high-end NVIDIA GPUs. Chinese companies have been forced to focus on efficiency, which is not a priority for U.S. companies that simply throw money at the problem. DeepSeek models roughly match the accuracy of ChatGPT-4o. Each new generation of ChatGPT models is about 10x bigger than the previous one and the new models are being released in about a year. The next generation ChatGPT-5 is expected in about 12 months. It will take DeepSeek time to come up with more compact and distilled models like ChatGPT-5.

The huge contribution of DeepSeek to humanity is that applying their techniques on next generation models to make them 10x more efficient will mean that they will need similar resources and have similar costs as today’s models instead of being 10x more costly. DeepSeek is open source, and all AI developers can adopt these efficient approaches.

Today’s generative AI falls far short of true Artificial General Intelligence (AGI). It is trained on a fraction of the text produced by humanity. If the current rate of model growth can be sustained it will be approximately 2032 before AI models can be trained on all text produced by humanity. Therefore, improvements and demand for more capacity should continue for at least 7-8 years. Likely, it would take longer because it is difficult to continue at the current growth rate due to financial, power and semiconductor capacity.

The models are like AI software running on AI hardware. DeepSeek models run on NVIDIA GPUs but can run on other hardware. Tachyum is porting DeepSeek on its hardware.

The industry is looking into various ways to improve efficiency. A key topic is that DeepSeek used FP8 data formats instead of BF16 for training, used by GOOGLE TPU for example. FP8 takes 2x less memory than BF16 and is about 4x more semiconductors and/or increased power efficiency. Tachyum announced FP8 support a few years ago. We believe that the whole AI industry will move toward FP8 training. Tachyum takes a similar approach to support shared exponents for FP8 as DEEPSEEK. Based on preliminary information, Tachyum FP8 matrix multiply-add operates at a higher precision than NVIDIA FP8, which should produce even better results for DeepSeek.

The final frontier is FP4 training which can be about 4x more efficient than FP8 training and will consume 2x less memory. The industry is making good progress towards FP4 training and Tachyum hardware provides superior support for FP4. The FP4 technology was unveiled by Tachyum several years before NVIDIA mentioned FP4. If DeepSeek and/or the industry masters FP4 training, Tachyum’s Prodigy hardware will be the superior platform for it.

Tachyum has previously published a paper on Tachyum AI (TAI) 4-bit and 2-bit effective weights for Large Language Models (LLM) and showed that 4-bit weights were fully viable, and for 2-bit weight, there was minor degradation of accuracy, to best of our knowledge, Tachyum’s achievement has not yet been surpassed by NVIDIA. For generative AI inference, it appears Tachyum chips will further improve the efficiency of DeepSeek.

Another significant DeepSeek innovation, which is hardware independent, was what mixture of expert models are activating in creating a response. Based on the preliminary information from DeepSeek, they need 10x less mixture of experts to load into memory and then activate. It is expected that such techniques will soon be adopted by U.S. AI companies, like OpenAI, Meta, Microsoft, Google and others.

Overall Tachyum’s view of the DeepSeek announcement and technology is extremely positive in many aspects, and it confirms that Tachyum chose the right processor development strategy. It will be explained below why we believe in such a positive impact.

Tachyum will enable the widespread implementation of the next generation models, like ChatGPT-5, sooner. Without Tachyum’s prodigy chips, the 10x larger models and their corresponding cost would constrain their deployment in markets sensitive to Return on Investment (ROI).

Because of chip export control to China, Chinese companies must focus on efficiency, and they have succeeded. Because DeepSeek is open source, U.S. companies will try to catch up to DeepSeek’s software technology, which will result in more cost and power efficient AI throughout the world.

Response from the U.S. administration to DeepSeek leadership has resulted in top officials calling it a “Sputnik moment”, when the U.S. realized they were falling behind other superpowers and responded with a dramatic focus on investment in AI in the U.S., and calls to create an AI Manhattan Project in the U.S. The result is the focus on a dramatic increase in AI investment in the US, on top of the StarGate project which intends to raise and spend $100-500 billion USD over the next 4 years. Other countries will also have to respond and increase spending on AI Research and Development (R&D).

Could you imagine such a technological hub with a similar reputation in Europe or even in Slovakia? It sounds like science fiction, but the key to success could be the world’s fastest AI supercomputer that Slovakia intends to build.

Why an AI supercomputer?

Supercomputing is something far removed from the average person’s thinking. Its benefits for society and its connections to everyday life can be hard to imagine, especially in our region. However, the reality is that the presence of high-performance computing (HPC) and artificial intelligence have expanded into areas of our lives we would not have expected before. Did you know that the design of Pringle’s potato chips was crafted by engineers using supercomputers, making chips easy to stack up and impossible to predict how they will break up when you eat them, creating a crunchier snack?

One of the sectors where the use of supercomputers is growing rapidly and becoming established is the automotive industry. The world’s car manufacturers use the capacity of supercomputers to design car bodies, improve aerodynamics, reduce fuel consumption and increase safety by simulations of collisions in crash tests. The automotive industry has built a strong tradition in Slovakia and the CEE region, so why not also move development here, in addition to assembly lines? With supercomputer capacity and performance, research and development in this region could move to a whole new level.

Another real-life application of supercomputers is healthcare. And this is for several reasons. One of them is vaccine and drug development, which can be attributed to high performance computing, which has shortened some vaccine development to one year, thus saving many human lives. The Covid-19 pandemic has shown the need to make this development more effective. Another aspect is the possibility of diagnostics improvements, such as cancer or other serious diseases, where, in particular, AI plays a critical role.

Today, there is no doubt that AI is increasingly becoming part of many aspects of our lives and societies. We use AI practically every day. It has changed the way we travel, communicate, and shop; it is used by hospitals to analyze images and symptoms, so they can provide better healthcare.

Most well-developed countries have understood this fact and reflected the critical role of AI in their national strategies and visions. Many developing countries are also beginning to realize this fact and are also working on their strategies, to not “miss the train”. According to PwC’s recent global AI study, AI will increase global GDP by 14%, or $15.7 trillion, by 2030¹. It is not a question of whether, but when the country should start preparing a national AI strategy, so that its economy does not stagnate and national GDP is not adversely impacted. AI is an opportunity for every country, because all countries today still stand on roughly the same starting line.

Slovakia would not have to start building its “Technological Valley” from scratch and on a “green field”. It may try to leverage one of its existing initiatives. Košice has good prerequisites in this regard, especially for their cooperation between universities, the connection of education with research and development, and the integration of industry into this ecosystem. An example is the Cassovia New Industry Cluster (CNIC) or IT Valley.

However, for these initiatives to succeed in attracting the world’s attention, and to get on the map of “technology valleys”, they must have something unique, something that’s the best in the world. That “something” could be a human brain-scale AI supercomputer. Such a project would certainly attract attention, and when combined with the right set-up, it can be a “magnet, not only for the region of Central and Eastern Europe but also for the European Union and the world. Slovakia would thus get on the “radars” of the world scientists, universities, global companies, and other relevant organizations.

Slovakia’s “Technological Valley” with the fastest AI supercomputer could play a key role in one of the main European projects for the next decade: “Destination Earth”. It is a digital twin of the Earth, a highly accurate model of the earth, powered by supercomputers, AI, Cloud computing and high-speed connectivity networks. Fed by various data sources, such as our eyes in the sky (Copernicus satellites) that will monitor and predict environmental change and its impact on human lives, it will give citizens the opportunity to critically assess policies and measures that affect them. The aim of this project is to analyze the past, monitor the present, and predict the future. Slovakia, with its human brain-scale supercomputer, could be at the center of this major inflection point,

The new Slovak supercomputer has been discussed for a long time, and plans to build it are finally becoming a reality. Slovakia has a unique opportunity to build the fastest AI supercomputer in the world. Let’s not miss this ”once-in-a-generation” opportunity.

There are multiple approaches to achieve human brain-like AI. These include machine learning, spiking neural networks like SpiNNaker, neuromorphic computing, bio AI, explainable AI and general AI. Multiple AI approaches require universal supercomputers with universal processors for humanity to deliver human brain-scale AI.

For example, the EU’s 1 billion euro Human Brain Project is a large, 10-year scientific research effort, and its goal is to provide the infrastructure tools to enable researchers across Europe to conduct research in the fields of neuroscience, computing and brain-related medicine.

Retired Manchester University professor Steve Furber, the inventor of the ubiquitous ARM processor and a leading researcher on the Human Brain Project, has been using Spiking Neural Nets to model high-level brain functionality.

AI is having a rapidly increasing impact on all of our lives, but there is still a great deal to learn from biology if we are to realize the full potential of AI in the future.

Steve Furber

The SpiNNaker machine, made up of 1 million ARM processors, is designed to mimic human brain functions and has modelled a mouse’s brain functionality in real time. In order to model the functionality of a human brain, it will take 1,000 times more computational power.

There were also experiments by IBM in 2009 in which it used 147,456 processors and 144 terabytes of DRAM at NNSA’s Lawrence Livermore National Lab and claimed to simulate a neural network the size of a cat brain — which is 4.5% of human brain capacity and runs 643 times slower than real time. Its model would require a machine with 23 times more capacity and be 643 times faster to approach human brain-scale AI.

Europe is dedicated to becoming an AI powerhouse. The EU, which uses 30% of the world’s computational resources, controls only 5% of those resources. To achieve technological sovereignty, the EU is now funding projects aimed at providing indigenous computational resources for the EU to move into an AI-intensive future.

Germany and France, along with other EU participants, have created a 7 billion euro HPC project and a 10 billion euro EuroCloud project — the goal of which is to design the next generation of a federated European “cloud” by specifying common requirements for an EU data infrastructure specified by Gaia-X.

While Europe is stepping into the AI supercomputing arena with significantly increased funding for several key projects, the U.S. is working on its own exascale machine. El Capitan, DOE’s $600 million, two-exaflop machine, is slated to come online at Lawrence Livermore Labs in 2023.

Dominance in the field of AI is not only crucial to GDP growth and our quality of life, but it’s also a national security priority. The U.S., China, Russia and the EU are in an AI arms race to see who will dominate the AI landscape.

Any way you slice it, the explosive growth in AI technology means that consumers, corporations and government agencies will soon be able to exploit highly advanced AI systems in order to create life experiences that we can only dream of today. Human brain-scale computing is now looming on the horizon, and it could change your life more than electricity.

In cooperation with the Ministry of Foreign and European Affairs of the Slovak Republic, Tachyum hosted the event featuring an in-depth discussion on the importance of AI and supercomputers, which will reach human brain-scale AI in 2022. Prominent guests included Ingrid Brockova, State Secretary of the Ministry of Foreign and European Affairs of the Slovak Republic, Radoslav Danilak, Founder and CEO Tachyum; and Steve Furber, ICL Professor of Computer Engineering at The University of Manchester, who was recently awarded the Charles Stark Draper Prize for contributions to the invention, development, and implementation of reduced instruction set computer (RISC) chips. The debate was moderated by Tarek Jundi, Founder and CEO MindMover.

Human brain-scale AI will enable artificial neural networks to be built that have a scale and complexity comparable with the human brain, which has about 10¹⁴ synapses. These are very large-scale AI systems.

Steve Furber

AI is very powerful technology, as many scientists say it is even more powerful than nuclear science. Like any technology you can use AI for good and bad. It is very important to use AI properly. Democratization of AI is very important and removes the monopoly of a very few rich nations with this technology and allows equitable access to technology by everybody throughout the world.

Radoslav Danilak

More than ever, we are witnessing the world change through digitization. An innovation-friendly economy is key. We have the potential to be an attractive incubator for innovative companies and Tachyum is another example of a unique innovative company from Slovakia.

Ingrid Brockova

The organization of the Expo 2020 Dubai in general has been fabulous and everyone must applause the UAE for the great organization of the event. Whilst a 3D presentation is a standard, many countries moved to 4D or even 5D experience including scents or vibrations to deliver the live experience to the visitors. Technologies connected to humanity and the importance of AI for humanity resonate in many pavilions such as Kazakhstan, Russia, Saudi Arabia, UAE and many others.

During development of a new processor, close cooperation of architects, hardware, software, and verification teams is necessary. Architects are responsible for the design of the chip, its instruction set, fetching and evaluation process of each instruction, cooperation, and data exchange between CPU cores and between the entire CPU in a multi-socket configuration. Many more of their activities define overall properties, performance, and the processor’s ability to communicate with external peripherals.

Software teams

One of the software teams modifies opensource GCC and LLVM compilers to match the instruction set, memory and peripheral design to generate a highly efficient assembly code, fully utilizing the power of a freshly-designed processor.

Building on their work, another software team modifies the operating systems to be able to seamlessly work on the brand new architecture. Besides operating systems, it is important to provide compatibility with widely used applications such as those which are intensively used in data centers: Database applications, webhosting and streaming tools, neural network training applications, tools for remote administration of servers, scientific packages…

You can find the list of currently supported software packages after logging into the Customer portal:

How can we check compatibility of such applications on our processor when it was not manufactured yet?

We are using a software emulator which precisely imitates the processes in our processor. This enables us to run and verify various software applications and operating systems even without having our hands on the Tachyum Prodigy processor. These emulators are running on our servers and our engineers are using them for verifying, testing and analyzing their performance.

Hardware teams

Hardware teams are working in parallel with software teams. They design the internal circuitry of the processor based on the architecture to match the required performance. This design is actually a huge schematic built of logic gates and function blocks which are used for interconnection with DDR5 memories or PCI express devices.

Such designs are subject to scrupulous checks done by the verification team. After this step, it is ready to be sent into the semiconductor fab for final design touches to match the production process. The same design can “run” on our FPGA prototype which uses one of the most advanced FPGA chips produced by Intel.

Our FPGA prototype is the final level in verification of our architecture and allows us to run complex algorithms and operating systems. Compared to software emulation, our FPGA based system uses real DDR RAM and real peripherals. This gives our engineers and partners opportunity to observe the system performance in real setup.

Emulation portal

The emulation portal grants our partners access to virtual machines running on the Tachyum Prodigy architecture as well as the FPGA prototype. That way they can test the compatibility of the tools they are using or ask our software team for assistance with porting these programs.

The following video depicts the process of logging in to the Emulation portal and using Linux 5.10.27 command line interface to:

• identify the system
• compile simple C application
• dump the assembly of this app
• clone and build apps from git repository
• run simple Java app
• list the attendees of ITAPA 2021 conference using Python

In this video following snippets of code were used:

• C/C++

#include <stdio.h>
int main()
{
  printf("\n\n\t\tTachyum zdravi Itapu 2021\n\n\n");
  return 0;
}

• Java

class ahoj {
  public static void main(String args[])
  {
    System.out.println("\n\n\t\tAhoj Itapa 2021!\n\n"):
  }
}

• Python

import re  
import urllib.request

speakri = []
for i in range(9):
  s = 'https://www.itapa.sk/9181-sk/spikri/?kongres=&page=%d' % i
  f = urllib.request.urlopen(s)
  html = f.read().decode('utf-8')
  speakri += re.findall('<h3 class=\"title\">(.*)</h3>',html)

speakri.sort()
print(speakri)

Photogallery

See photos from ITAPA 2021 International Congress:

Contact our team at the email below for evaluation access to the Customer Portal

Why is this important? What are the benefits to drivers and passengers?

Let’s look at history that shows that driving cars goes hand in hand with accidents. There are passive and active components to protect passengers in cars with passive ones focused on minimizing accident impact. These include seat belts, air bags and more. Active ones are designed to prevent accidents and make driving safer.

At slow speeds, such as parking, we use ultrasound sensors. In 1987, the Belgian Rudy Beckers invented the parking sensors that are now used in cars all over the world. He invented the sensors so that his wife would no longer have to get out of the car to give him directions while he was parking.

Cameras. More and more vehicles have cameras for all around view. At first, they were used to help parking, later for reading road signs and traffic lights or for lane recognition. More importantly, people and animals can be recognized by cameras in the dark, thanks to infrared technology to improve safety.

Radar. As we know, radar works with electromagnetic signal, so it is best to position distance between vehicles. Radar is mostly fitted in front of the car to work with adaptive cruise control, helping to keep a safe distance between cars in front. The backward-facing sensor records cars approaching from behind and potentially changing lanes to avoid a rear end collision. They are also important when moving in reverse and in parking.

All those sensors were initially used as separate items performing just one function. Nowadays all data from sensors can be interconnecting and processing in real time, resulting in automated motorway driving. Reliability is increasing but drivers still need to pay attention.

Lidar systems for real-time image recognition are the most discussed sensors in the car industry. Some in the car industry see lidar sensors as the future and some do not. Some because of price and some because they feel they can trust cameras which are cheaper and usable with AI (Artificial Intelligence) algorithms that help.

As the sensors are interconnecting, we can see interconnecting cars using WLAN or 5G to match behavior of cars close to each other.

Innovative technology definitely helps to protect passengers and save lives.

AI role in self-driving and why EU behind Tesla

Why is Tesla winning? Tesla is more of an innovative technology company than car manufacturer. They equipped all their cars with camera systems with a vision that create AI algorithms enabling the autonomous driving of the car. Their plan is very simple – they just upload and enable software to do it. All components are already in place.

In fact, what we see today is Tesla doing some adjustments in their plan uploading only software is not difficult, but their dream becoming reality. Tesla’s marketing KISS (Keep It Simple Stupid) works. While their customers may not understand all the technologies involved, but appreciate the advantages they get and believe in the dreams of Elon Musk.

In general, we can say that machine learning algorithms make it possible. Today most of electrical vehicles (EV) or other cars with driving assistance have cameras and lots of sensors. They can record driving in real time. Machine learning needs this recording for learning. This is how Tesla does it. In the EU we have GDPR and it is no problem for computers to anonymize the recordings. The more video and sensor recordings, the better algorithms we will get.

How Slovakia and Tachyum could help

A supercomputer, which could be built in Slovakia with Tachyum Prodigy^® chips is 10 times faster that Tesla has today and speeds up AI technology working over recordings. In the EU there are more car manufacturers other than Tesla being sold at the same level as Tesla or even better. In addition, EU car makers have the advantage of learning from Tesla.

Autonomous driving is perceived by society as a system that drives the same as a human driver. But this is not the case. The goal is to drive better and safer than a human. Most accidents happen because of human error. The driver is tired, drunk, not paying enough attention, does not have enough experience, or has overvalued his driving abilities. Computers will never have these problems. We can expect that these systems will get better and better and have even more advantages. We just need to continuously train the systems via AI as much as we can, and they will constantly improve.

Moreover, computers can focus with the same quality in many directions via cameras or sensors. A human driver can’t. The next level will car intercommunication. Vehicles will know about each other dynamically within a 1 km range and optimize the drive. Just imagine how many accidents will be avoided if we were able to see 4 or 5 cars in front of us.

Tachyum technology could be used to process all this data. Slovakia’s supercomputer is the perfect fit. As mass production of these chips hits the market, car makers can procure their own supercomputers.

With this achievement coming, we can be sure that future will be better. What the EU needs is a powerful AI system, and Slovakia and Tachyum can enable it.

Dr. Danilák and Professor Pavol Sovák, the Rector of Pavol Jozef Šafárik University signed a Memorandum of Understanding. The aim is to take Artificial Intelligence and Machine Learning teaching to a higher level and to enrich dissertations with a new element - Tachyum’s emulator for the future supercomputer. Interesting questions from a mathematical and theoretical-informatics point of view also arise in the field of new data structures created by Tachyum engineers. Here is a space for communication between science and engineering.

Dr. Danilák discussed with CNIC chairman Professor Pavol Miškovský the potential of the connection between Tachyum and CNIC, which will support the possibility of spin-offs in the field of AI and at the same time create a critical mass of innovators for cooperation on further Tachyum products.

The meeting of Dr. Danilák with his former teacher from the Department of Computers and Informatics of the Faculty of Electrical Engineering and Informatics at Technical University of Košice, Dr. František Jakab, Associate Professor, was very amiable. “It has been my genuine pleasure to watch the information in the media about Radoslav Danilak as one of the world’s most respected hardware architects in recent years,” said Dr. Jakab. “I see him standing together with me in front of the door of Professor Jelšina’s office, which led the subject of computer architecture. I remember how Rado held in his hands drawings with the architecture solution of a special arithmetic coprocessor, the solution which he wanted to show to the professor and which he had previously shown and described to me in detail. Therefore, I was really pleased when I had the opportunity to welcome Rado and his team in person at the TECHNICOM University Science Park of the Technical University in Kosice and discuss with him his future plans and intentions and interest in cooperating with the Technical University.”

If a country wants to be recognized as a supercomputing power, it’s not enough to have a supercomputer in the top 100. It’s not even enough to have a supercomputer in the top 50. There are nearly 200 countries in the world. If researchers want a supercomputer to solve the most complex and challenging calculations, they don’t want the 50th best supercomputer. They want the best, second best, or maybe the third best.

With Tachyum’s Prodigy as the heart of the NSCC Slovakia Supercomputer, expected to debut in the first half of 2022 – the world’s fastest AI supercomputer – the entire nation wins. A reference design featuring 256 Prodigy Processors per 54U rack delivers 1 AI ExaFLOP of compute power. This powerful AI performance can be used to address crucial international issues such as climate change, global health, and universal language translation. The eyes of the world will be on Slovakia, and Tachyum as the core of AI infrastructure.

Tachyum will be a highlight among Slovak exhibits at EXPO Dubai, Slovakia’s key event abroad in 2021-22, during Global Goals Week, January 16-22, 2022. Tachyum, in cooperation with Ministry of Foreign and European Affairs will host an event featuring a well-developed discussion on the importance of AI and supercomputer innovations in a knowledge society, and its concrete benefits for humanity.

It is also crucial for making our planet greener, at a time when the effects of climate change are ravaging our world. AI will spawn progress in weather and disaster forecasting, reducing industrial emissions, designing fuel-efficient engines, preserving natural habitats, and more.

These AI-based projects will benefit us all. Current supercomputers are somewhat less than super, and have limitations in speed, power consumption, space requirements, costs, access to technology and expertise. Put simply: AI is not for all.

The demand for computational power is unstoppable. AI voraciously consumes processing power, and that translates to electrical power. Energy consumption such as this is unsustainable – and ironic, considering we are trying to save the planet, not tax it further by increasing power demands. We must find another way. Constantly building new power plants to supply electricity to data centers is not the answer.

Worldwide, the competition to reap sustainable AI benefits has heated up, while the EU slept in. EU consumes 30% of world compute power resources, but operates only 5% of world’s compute resources. As much as 84% of data is processed outside the EU, so the actual functional amount of resources consumed by EU is staggering. Even more distressing, this means most AI in the EU today is in the hands of other countries and misaligned with EU interests.

We need to address this situation, for the sake of progress as well as national security. Tachyum’s Prodigy Universal Processor, a general-purpose chip equally adept in commercial and government applications, can help accomplish this and even position the EU as the leader in supercomputing and the data center market.

Tachyum uniquely combined GPU and CPU computational capabilities in Prodigy, resulting in dual-use hardware and unrivaled computational speed and efficiency. How efficient? Prodigy consumes one-tenth the electrical power of today’s chips in equivalent workloads – which in itself is enough to bring AI processing back home to the EU instead of relying on the resources of other nations.

At Tachyum, we have designed our Prodigy chip to be a core building block for the future of cloud and data center IT infrastructures in the EU, to protect the interests of the EU, and to address sustainability goals.

Building the Slovak AI on our platform, with an installation of 64 compute racks, delivers 64 AI ExaFLOPs and 515 DP PetaFLOPs. This far exceeds the standard for simulating the processing power of the human brain (between 1 and 100 ExaFLOPs, an ExaFLOP is 10¹⁸, or a billion billion, floating-point operations per second). The cost of building this human brain-scale is 70 million euros.

In contrast, the price tag of Fugaku, the world’s fastest, which began operation in May 2020, delivers 537 DP PetaFLOPs and costs about one billion U.S. dollars, or nearly 850 million euros. This is more than 12 times higher than the Slovak AI.

With the fastest AI in the world, Slovak and the entire EU will move from the back of the race to the front, and a world-leading position.

With its computing capacity, energy efficiency, and cost the Tachyum will democratize AI for all!

And we’re doing it soon: the fully functional Prodigy emulation system is now ready for carefully selected customers to perform early testing and software development.

As many the once-only-imagined projects, like autonomous vehicles, become reality, the coverage has been celebratory and optimistic about how they will benefit society.

But what is often underreported is the cost that such advancements impact on the world to achieve such results. The proliferation of internet-connected devices and the growing demand for compute-intensive applications require a massive expansion of data center capacity, which will require a massive increase in electricity consumption. Because such costs are absorbed by the huge tech companies leading these initiatives, the public has largely been unconcerned but the environmental impact of AI/ML workloads will be paid by all.

While the amount of electrical power required once followed Dennard’s scaling while processors followed Moore’s law of growth – doubling transistors every 2 years, there has been an exponential increase in power consumption since 2012 to satisfy the requirements for machine learning systems – doubling every 3.4 months, according to AI research group OpenAI. The largest data centers today are estimated by the U.S. Department of Energy to each require more than 100 megawatts of power capacity – enough to power around 80,000 households in the U.S. With the media buzz promising that AI will soon be everywhere, energy consumption such as this is unsustainable.

To contain the energy crisis facing the world with the inevitable increase in AI/ML applications, a solution like Prodigy, the world’s first universal processor, is needed to help offer significant improvements in performance, energy consumption, server utilization and space requirements in hyperscale data centers.

In provisioning Prodigy for use in current and future data centers in exchange for the slate of non-optimized chips now available, the industry will be able to achieve significant “green” benefits, such as:

• CO2 emission reduction of 600 million tons per year – the equivalent of eliminating the entire airline industry
• 10 years of data center expansion without the need to expand power usage at all
• Enablement of single hardware platforms that can handle multiple AI workloads at a time, eliminating redundant systems and avoiding the need to move massive data from one system to another
• Democratization of AI that will enable sustainable products and solutions across a broad spectrum of applications and markets

The need for energy-efficient data centers in not a new concept but with the rapid acceleration of power-hungry AI/ML workloads placing an increased strain on an already taxed electrical grid infrastructure, the costs are too high to continue to ignore. By enabling a 4x lower data center annual TCO, Tachyum’s Prodigy is an ideal solution to solving these issues and advancing the entire world to a greener era.

How do you eat an elephant? One bite at a time.

Though most people probably wouldn’t contemplate actually eating an elephant, the prospect of completing such a huge task would certainly be daunting for anybody. That’s why the approach of finishing the job one step at a time is an important consideration when faced with any Herculean task.

Here at Tachyum, we’ve been readying our Prodigy Universal Processor for quite some time. Our goal of getting it to market later this year is one that we have clearly stated, and we have been working diligently, step by step, to ensure that we accomplish this. The latest steps we have made towards production include the availability of our FPGA emulation prototype for Prodigy and the release of our motherboard emulation for Prodigy. Here’s what these important milestones mean and why they’re important …

FPGA stands for field-programmable gate array, which is basically some hardware circuits than can be programmed with complex functions that can act together as a more comprehensive multi-core processor. It is an important step before manufacturing our Prodigy chip because it allows us to verify Prodigy’s capabilities and design to ensure that the final product will perform as intended. Our FPGA emulation prototype was developed in house to offer the best possible accuracy and performance. Customers can begin benchmarking and porting their own application software to native Prodigy code, enabling them to use Prodigy’s fully functional FPGA emulation for product evaluation, performance measurements, as well as software development, debug and compatibility testing. This will also help customers smooth the adoption curve for Prodigy in their existing or new data center and/or HPC systems that demand high performance, high utilization and low power.

Our motherboard emulation prototype of the Prodigy Universal Processor FPGA prototype was recently sent to manufacturing, which, when returned, will be plugged into the completed FPGA emulation Prodigy Processor boards to create a fully functional system. This milestone brings Prodigy one step closer to production-ready status. With the motherboard back from manufacturing, which is expected in early May, progressing through each stage of testing and manufacturing so quickly stems from our meticulous design process and a thorough understanding of semiconductor engineering and the unbelievable talent of the Tachyum team. We look forward to continuing the journey to tape out later this year.

We are lucky to have been able to savor each “bite” in the process of developing Prodigy thus far. From the initial concept to engineering plans to seeing that first “blink” of the FPGA when it was plugged in, it has been a rewarding experience and one we cannot wait to share with the world when completed. Please continue to follow along with our journey to completion through this newsletter or via our social media platforms.