Artificial intelligence isn’t merely going to amp up the entertainment and smart home industries. It will save your brain cells, mimic human intelligence, and man factories. Wherever AI emerges, semiconductors will be needed, so it’s no surprise that machine learning is driving the next big semiconductor insurrection. New needs and fresh manufacturing approaches will revolutionize chip design as manufacturers race to create a product that can handle massive data quantities.
Limitations and Innovations
It’s far more expensive to recreate a semiconductor chip than it is software, so deep-pocketed participants like IBM, MIT, and NVidia will drive the change. Venture capitalists are already excited about AI, having invested a little over $100 million in this niche. As it stands, graphics chips consume far too much energy, so chip manufacturers are beginning to produce more power-efficient processors. Bespoke processors can support some machine learning tasks, but new chips can now combine several processing tasks at once, making them the industry’s best hope.
Some brands and universities have already unveiled upgraded chips made specifically for AI, but the technology is still too new to have fallen victim to a universal approach. This creates a thrilling technological eco-system that will probably produce some fascinating innovations.
Non-volatile memory is in short supply in AI chips, and that creates fresh demand for 3D NAND technology. The tiny chips smartphones and computers use are insufficient. AI chips will be larger and data-centric. They’ll have to allow software to discover patterns and learn new things—two tasks that use immense power. Chips will need to suit hyperscale servers and autonomous technologies.
Apple and Google have already begun to use AI, relying on GPUs for machine learning. Chips will rescue the marketplace from that trend, creating an exciting future for several industries.