Hello everyone, the advancements in artificial intelligence (AI) and technology are remarkable these days, aren’t they? Today, we’re going to talk about ‘computation’. Specifically, where should computation take place – on edge devices or in the cloud? Let’s explore why this topic is important and what changes it can bring.
Edge vs. Cloud: Where Should Computing Happen?
When you think of computers, do you think of ‘semiconductor-based calculators’? In the past, all computations were done on centralized servers. However, with the advent of personal computers (PCs), individuals also gained powerful computational abilities. The development of PCs made it possible for individuals to perform many computations on their own devices.
But with the emergence of the internet in the 2000s, the importance of centralized servers grew again. The development of cloud technology enhanced the role of servers with data and computational power, leading to many computations being performed in data centers. This is how the concepts of ‘edge computing’ and ‘cloud computing’ emerged. Edge computing refers to computations performed close to the device, including on-premises servers owned by companies.
The Balance in the Cloud Era
In the 2010s, the widespread adoption of smartphones struck a good balance between edge devices and cloud computing. Billions of smartphones worldwide, supported by data centers, played their roles well. Cloud computing became a standardized resource like electricity, and related companies grew steadily.
Companies like Apple, Samsung, and Qualcomm fall under the device category, while Microsoft, AWS, and Google belong to the cloud category. Companies like NVIDIA and Intel are involved in both, as they produce GPUs for both PCs and data centers.
Generative AI: Only Possible in the Cloud!
However, the recent rise of deep learning and generative AI has disrupted the balance between terminal devices and cloud computing. Large language models like ChatGPT do not operate on smartphones or PCs. Such AI services require the support of the cloud and data centers.
When computation is performed in the cloud, someone has to pay the cloud provider. If consumers do not pay, the costs eventually flow to big tech companies. If the dependency on cloud-based AI is not reduced, it ultimately results in companies like NVIDIA benefiting financially.
Bringing AI to Edge Devices
Samsung is working to quantize and optimize AI models to fit them into edge devices. Microsoft is developing a small language model called Phi to run on PCs and smartphones. Qualcomm has introduced the Snapdragon X Elite, specialized for AI computation, and Apple is upgrading the AI computational capabilities of its self-designed semiconductors.
Apple is also expected to offer AI services in its data centers. There seem to be two ways to do this. One is to use AI models like Google’s Gemini or OpenAI’s GPT, and the other is for Apple to develop its own AI and operate it in Apple’s data centers.
Apple’s Data Center Investment
Apple is known as the big tech company with the fewest data centers. However, it is expected to increase its investments in AI data centers. This is evidenced by the hiring of semiconductor experts and the active efforts to build AI data centers.
What will Apple’s data centers look like? One can imagine a server computer system based on Macs. Efficiently designed data centers could provide the best AI and cloud services to Apple devices.
The Future of Data Centers
Elon Musk’s xAI plans to use substantial funds to build AI data centers. AI data centers are not just costs but are akin to factories producing products. The reason for building such expensive AI data centers is to support Tesla’s autonomous driving training and xAI’s learning processes.
Samsung once challenged the server computer market but failed. However, if it takes on the AI data center industry again, South Korea’s standing could change significantly.