This is an AI-generated image created with Midjourney by Molly-Anna MaQuirl
Much of the AI news that dominates the conversation is about what AI can do, what it is doing or the threats it poses. Depending on the person’s feelings about AI, it is treated like a magic wand with the power to improve all our lives or like an all-powerful insidious monster that will destroy society. The truth, of course, is somewhere in between and much less sensational.
What is usually missing from these discussions is an insight into how AI actually works and the technology necessary for this modern marvel to function. AI chips are at the forefront of AI development but don’t typically receive the attention they deserve.
Today, we’ll discuss what AI chips are and why they’re important. Then, we’ll take a look at the leading AI chips companies and why who makes AI chips matters.
Today’s AI solutions are among the most highly advanced technologies available to consumers and businesses. Due to how advanced they are and the complicated procedures that they are asked to execute, they require a huge amount of processing power. That’s where specialized AI chips come in.
Tech experts don’t have an exact definition of what AI chips are as they’re still so new and the industry is moving so quickly, but the general consensus is that AI chips are the computing hardware that is capable of handling the amount of workload AI demands. One of the roles these chips are used in is in training AI systems.
The faster an AI system can be trained, the sooner it can become useful, and the creators can see how well their algorithms work. Since trial and error can often be involved in creating an AI system, the training process needs to be as streamlined as possible. Advanced AI chips make that possible.
One aspect of these chips that sets them apart from the average computer chip is the use of parallel processing. The name provides a clue to how they work. Instead of working on one operation at a time, a parallel processing chip performs multiple tasks at the same time. A simpler version of parallel processing is used in some video game chips, particularly graphics processing units (GPUs).
Top of the line AI chips are incredibly powerful, and they have a price tag to match. The H100, recently released by NVIDIA, makes use of 80 billion transistors and is currently retailing for $30,000. This sort of price tag gives a hint to who the audience for these top chips are. They are designed for businesses and governments to use, not for the average consumer.
Most consumers with an interest in AI will never have any need to interact with or think about the chips that make AI possible. This is because end users deal with the finished product, with AI programs such as Dall-E and ChatGPT, or with AI services such as Siri, Cortana and Alexa.
As the researchers at the Center for Security and Emerging Technology have explained, being at the forefront of AI chip production is essential for countries to maintain national security. This is because it is far easier to control “the computer hardware necessary to implement modern AI systems” than it is to control everything else related to AI.
This is why the companies making AI chips matter so much. Governments need to be able to trust the companies behind their AI chips. With the current tense state of global relations, and the climate of suspicion that exists between many nations, having some measure of control over where these chips come from is essential.
NVIDIA is the leading AI chip manufacturer in the world. They were the company behind the earliest GPUs, and they have used the last few decades to build on that genius innovation. While they are based in California, their chips are produced around the world in Taiwan and South Korea. This kind of outsourcing is partly what governments have concerns about because there is little to no way for governments to exercise any oversight on these overseas factories.
Intel is the AI chip manufacturer that many consumers will be the most familiar with, thanks to the fact that Intel chips power most home computers and laptops. Nearly all of Intel’s chips are fabricated in the United States. They also have their own neural network training accelerator, named Gaudi. This amount of control at all levels is ideal for those with security concerns.
It shouldn’t be a surprise that Google has their own AI chip. Google Alphabet is the wing of the company that produces their AI chips. These chips are used to power most Google programs and products. As such a powerful company, it shouldn’t be shocking that many governments are suspicious of the amount of autonomy Google has.
Not all the top AI chip manufacturers are long-established companies. Like all things tech, start-ups are an important part of the ecosystem. Among the top companies creating AI chips in 2023 are Cerebras Systems, SambaNova Systems, Graphcore, Groq and Mythic. All but Graphcore, which is from Bristol, are based in California. These are all considerably smaller and younger companies, but their success shows that AI is a developing field that rewards creativity and innovation.
AI chips are one of the parts of the entire AI system that everyone should be able to agree on. While the algorithms and software are the brains of the AI, the chip is the heart and nervous system — without the chip, there is no functional AI.
Having a large number of different companies making AI chips is equally important. Competition will drive forward innovation and keeping the industry based in multiple countries is the best way to ensure that things remain balanced around the world.
AI chips might not be as flashy a topic as the possibilities of AI or as dramatic as stories of AI missteps, but they’re just as important.
This is an AI-generated image created with Midjourney by Molly-Anna MaQuirl
Much of the AI news that dominates the conversation is about what AI can do, what it is doing or the threats it poses. Depending on the person’s feelings about AI, it is treated like a magic wand with the power to improve all our lives or like an all-powerful insidious monster that will destroy society. The truth, of course, is somewhere in between and much less sensational.
What is usually missing from these discussions is an insight into how AI actually works and the technology necessary for this modern marvel to function. AI chips are at the forefront of AI development but don’t typically receive the attention they deserve.
Today, we’ll discuss what AI chips are and why they’re important. Then, we’ll take a look at the leading AI chips companies and why who makes AI chips matters.
Today’s AI solutions are among the most highly advanced technologies available to consumers and businesses. Due to how advanced they are and the complicated procedures that they are asked to execute, they require a huge amount of processing power. That’s where specialized AI chips come in.
Tech experts don’t have an exact definition of what AI chips are as they’re still so new and the industry is moving so quickly, but the general consensus is that AI chips are the computing hardware that is capable of handling the amount of workload AI demands. One of the roles these chips are used in is in training AI systems.
The faster an AI system can be trained, the sooner it can become useful, and the creators can see how well their algorithms work. Since trial and error can often be involved in creating an AI system, the training process needs to be as streamlined as possible. Advanced AI chips make that possible.
One aspect of these chips that sets them apart from the average computer chip is the use of parallel processing. The name provides a clue to how they work. Instead of working on one operation at a time, a parallel processing chip performs multiple tasks at the same time. A simpler version of parallel processing is used in some video game chips, particularly graphics processing units (GPUs).
Top of the line AI chips are incredibly powerful, and they have a price tag to match. The H100, recently released by NVIDIA, makes use of 80 billion transistors and is currently retailing for $30,000. This sort of price tag gives a hint to who the audience for these top chips are. They are designed for businesses and governments to use, not for the average consumer.
Most consumers with an interest in AI will never have any need to interact with or think about the chips that make AI possible. This is because end users deal with the finished product, with AI programs such as Dall-E and ChatGPT, or with AI services such as Siri, Cortana and Alexa.
As the researchers at the Center for Security and Emerging Technology have explained, being at the forefront of AI chip production is essential for countries to maintain national security. This is because it is far easier to control “the computer hardware necessary to implement modern AI systems” than it is to control everything else related to AI.
This is why the companies making AI chips matter so much. Governments need to be able to trust the companies behind their AI chips. With the current tense state of global relations, and the climate of suspicion that exists between many nations, having some measure of control over where these chips come from is essential.
NVIDIA is the leading AI chip manufacturer in the world. They were the company behind the earliest GPUs, and they have used the last few decades to build on that genius innovation. While they are based in California, their chips are produced around the world in Taiwan and South Korea. This kind of outsourcing is partly what governments have concerns about because there is little to no way for governments to exercise any oversight on these overseas factories.
Intel is the AI chip manufacturer that many consumers will be the most familiar with, thanks to the fact that Intel chips power most home computers and laptops. Nearly all of Intel’s chips are fabricated in the United States. They also have their own neural network training accelerator, named Gaudi. This amount of control at all levels is ideal for those with security concerns.
It shouldn’t be a surprise that Google has their own AI chip. Google Alphabet is the wing of the company that produces their AI chips. These chips are used to power most Google programs and products. As such a powerful company, it shouldn’t be shocking that many governments are suspicious of the amount of autonomy Google has.
Not all the top AI chip manufacturers are long-established companies. Like all things tech, start-ups are an important part of the ecosystem. Among the top companies creating AI chips in 2023 are Cerebras Systems, SambaNova Systems, Graphcore, Groq and Mythic. All but Graphcore, which is from Bristol, are based in California. These are all considerably smaller and younger companies, but their success shows that AI is a developing field that rewards creativity and innovation.
AI chips are one of the parts of the entire AI system that everyone should be able to agree on. While the algorithms and software are the brains of the AI, the chip is the heart and nervous system — without the chip, there is no functional AI.
Having a large number of different companies making AI chips is equally important. Competition will drive forward innovation and keeping the industry based in multiple countries is the best way to ensure that things remain balanced around the world.
AI chips might not be as flashy a topic as the possibilities of AI or as dramatic as stories of AI missteps, but they’re just as important.