As major semiconductor companies have been holding large-scale AI-related events since November and revealing key technologies and agendas to take the lead in the AI market, we have reached an era where we can no longer survive in the semiconductor market without discussing AI.

■ SK, building total AI from semiconductors to cloud

SK Group held the 'SK AI Summit 2024' on the 4th. At the event, SK Chairman Choi Tae-won personally gave a keynote speech and announced AI-related business plans and key technologies.

Chairman Choi Tae-won highlighted the challenges that need to be addressed for AI development, citing the shortage of AI accelerators and semiconductors. In particular, he introduced an anecdote where Jensen Huang, CEO of Nvidia, asked for the supply of HBM memory to be brought forward, and stated that the supply of semiconductors is not keeping up with demand.

To this end, they announced that they are accelerating technology development.

In relation to this, SK Hynix President Kwak No-jeong also said in his keynote speech that the company is currently developing a 48GB 16-layer HBM3E and plans to provide samples to customers early next year, predicting that the 16-layer market will open in earnest starting with HBM4.

In addition, they said that they plan to introduce logic processes to the base die starting with HBM4, and that Custom HBM is a product that optimizes performance by reflecting various customer needs such as capacity, bandwidth, and additional functions, and will become a new paradigm for future AI memory.

A lot of investment is being made not only in semiconductor development but also in infrastructure construction.

SKT CEO Yoo Young-sang also announced in his keynote speech that SKT will begin with a hyper-scale AI DC (data center) requiring over 100MW (megawatts) of power in its domestic base, and expand its scale to over GW (gigawatts) in the future to become an 'AI DC hub' in the Asia-Pacific region.

In addition, in order to quickly resolve the domestic GPU supply shortage, the company plans to launch GPUaaS (GPU as a Service) that provides GPUs in the cloud by converting the Gasan Data Center located in the metropolitan area into an AI DC. It mentioned that starting with GPUaaS based on H100 in December, it will introduce the latest GPU H200 for the first time in Korea in March 2025, and that it plans to gradually expand the quantity according to customer demand.

■ Samsung Electronics Emphasizes Sustainable AI Ecosystem

Samsung Electronics also held the 'Samsung AI Forum 2024' on the 4th and 5th. Samsung Electronics focused on traditional AI discourse and discussed the future of AI with global AI experts.

Samsung Electronics focused on building a more efficient and sustainable AI ecosystem.

Professor Yoshua Bengio of the University of Montreal in Canada, who attended the forum, emphasized building safer and more reliable AI systems in his keynote speech titled “Bayesian Oracles for AI Safety.”

Professor Bengio introduced research results showing that AI's performance has surpassed human levels in various fields, along with the future risks of AI, such as large-scale labor market impact, hacking using AI, and the emergence of superintelligence, and stated that in order to ensure AI safety, not only policymakers but also the public must have an accurate understanding of the current status and future of AI.

In the technology session, AMD CTO and Vice President Joseph Macri introduced AMD's AI solutions on the topic of AI everywhere, and explained the importance of collaboration with AI platforms and AMD's strengths.

■ Arm accelerates AI development through the ecosystem

Arm also held the 'Arm Tech Symposia 2024' on the 1st and announced that it would strengthen the ecosystem in AI data centers.

Arm Total Design is growing rapidly with the participation of about 30 companies, bringing together capabilities from design to foundry manufacturing.

Targeting cloud, HPC, AI/ML training and inference workloads, the platform builds ADTechnology’s Neoverse CSS V3-based compute into chiplets and combines Rebellion’s REBEL AI accelerator with those compute chiplets, implemented on Samsung Foundry’s 2nm Gate All Around (GAA) advanced process technology.

These collaborations accelerate ecosystem AI development by enabling faster access to hardware and software advancements.