Nvidia (NVDA) has hit a new milestone with the unveiling of a first-of-its-kind AI supercomputer. Going for $3,000, Project Digits is the new supercomputer targeted at developers and data scientists looking for an affordable way of handling the lifecycle of generative AI. Digits have everything needed to create a complete generative AI solution, from inference to fine-tuning to agent development.
Project Digits
The new GB10 Grace Blackwell Superchip, at the core of the new supercomputer, is small enough to fit on a desk. While it can run from a standard power outlet, it has enough processing power to run complex AI models. This kind of processing power used to require much larger, more power-hungry systems.
This accessibility, which makes AI training more affordable for individuals and smaller businesses, is a significant step toward democratizing the field. Digits have a starting price of $3,000 and can handle AI models with up to 200 billion parameters. The product itself resembles a Mac Mini in appearance.
The GB10 chip provides UP to 1 petaflop of AI performance, which can execute one quadrillion AI calculations per second. To maximize the GB10’s performance and power efficiency, MediaTek, a company well-known for designing Arm-based chips, worked with the development team.
TSMC Sales Growth
Meanwhile, Taiwan chip manufacturer Taiwan Semiconductor Manufacturing Company Limited (TSM) saw a more than one-third increase in net sales last year as a result of the growing demand for AI technology. With a 33.9 percent increase, net revenue reached $87.8 billion. The sales growth surpassed the October projection of a nearly 30% increase in full-year revenue.
Consequently, it highlighted the expectation that companies like Microsoft and Amazon would continue to invest in AI infrastructure. Leading the way in the generative AI revolution, TSMC produces the most cutting-edge microchips in the world, which are essential for Silicon Valley products. The robust sales growth comes as the Taiwan chip giant owns more than half of the global chip production, which is used in everything from Nvidia’s state-of-the-art AI hardware to Apple’s iPhones.
Amid the better-than-expected results, TSMC is still grappling with geopolitical tensions between the US and China over technology import restrictions. Because of how susceptible the semiconductor supply chain is to shocks, governments have urged TSMC to shift more of its production outside of Taiwan.
The United States has announced new export regulations on advanced computing chips used for artificial intelligence to facilitate sales to allies and further restrict access to nations such as China.
The limitations, which also include guidelines for closed AI model weights, follow restrictions on exporting specific AI chips to China, which the US views as a strategic rival in the advanced semiconductor market, that were announced in 2023. In addition to requiring authorizations for exports, re-exports, and in-country transfers, the new regulations update controls to include a number of exceptions for nations deemed friendly to the US. AI information
The incoming president, Donald Trump, has yet to discuss his views on export restrictions. However, during his first term, he prohibited Huawei, a leading Chinese provider of telecom equipment, and other Chinese civilian tech firms from using American technology.
