CHIPS Acts Around the World. The $280B CHIPS and Science Act signed in 2022 funds subsidies for the U.S. chips industry. Japan is spending $13B; Europe will spend $47B; India has a $15B effort. This follows massive Chinese subsidies for its chipmakers. It’s unclear if these investments will succeed in changing the chip supply chain but attempts to alter the status quo will likely continue.
More AI on the Edge. Most interactions with AI models are done via the cloud, but this requires internet access and some personal data is shared with the model maker. This is driving edge computing investments where interactions with AI models happen on a device (e.g. laptop/smartphone) and adding demand for faster edge chips that are smaller/cheaper/more energy efficient than data center chips. The Defense Department recently partnered with EnCharge AI to create a powerful edge chip that enhances privacy but operates on little power. It would support military applications, also AI-driven devices in homes and at work.
Big Tech Enters the Chipmaking Fray. Many companies are creating/training AI models of their own. Demand for cloud computing to train these models is rising fast. Amazon, Microsoft and Google provide much of that computing power, and want to make their data center chips in-house instead of buying from companies like Nvidia, a chip giant with the most advanced AI training chips. Amazon started its effort in 2015. Google moved in 2018. Microsoft launched its first AI chips in 2023. Meta introduced a new version of its AI training chips in April. Nvidia is attempting the reverse. Last year, it started a cloud service so customers can bypass tech companies and get computing time on Nvidia directly.
Nvidia Battles the Startups. Despite Nvidia’s dominance, money is flowing to startups to compete in the chip market. These startups promise faster AI training with different underlying technology, from quantum to photonics to reversible computation. Companies like SambaNova, Cerebras and Graphcore are trying to change chip architecture to deliver faster/cheaper AI training to customers, also working on software to facilitate a seamless transition to the new chip. Time will tell if these startups are successful and whether they can chip away at Nvidia’s market position.
MOSIMTEC can help your company go where you haven’t gone before with AI technology. We help companies add new technologies to improve performance and build competitive advantage every day. MOSIMTEC guides clients on their technology journey, with digital twin simulations to explore options and outcomes, so they get the most out of their business. Our data scientists/industrial engineers can help future-proof your business.
Read More: https://www.technologyreview.com/2024/05/13/1092319/whats-next-in-chips/