Nvidia recently unveiled the world’s most powerful artificial intelligence (AI) chip, the B200 Blackwell, which the company says is up to 30 times faster than its previous market-leading chip.
With interest in AI white hot, the announcement made headlines around the world. Nvidia says the new chip will be twice as powerful at training large language AI models and will dramatically increase the speed at which ChatGPT and other AI models can respond to queries.
The B200 Blackwell is at the forefront of a semiconductor industry that McKinsey forecasts will grow by 6% to 8% annually until 2030, by which time annual revenues are predicted to reach $1 trillion.
This rapidly growing global appetite for chips is not only being driven by the explosion in the use of AI. Today chips are used in everything from weapons systems to washing machines. A new car rolling off the production line is likely to use between 1,000 and 3,000 semiconductor chips – and 5G phone networks, the Internet of Things and Blockchain technology all need huge amounts of chips.
The majority of the chips we use are central processing units (CPUs), which are generally comprised of between four and eight cores that process and perform tasks. Although CPUs are very fast, the growing demands of AI mean we need more and more powerful processing power.
The B200 Blackwell is a graphics processing unit (GPU). These chips were initially developed for gaming and video editing, which demand very fast, heavy-duty processing, and they can have hundreds of cores that work in parallel to simultaneously execute a series of tasks. This processing power means they are far quicker than CPUs.
Geopolitics
While globally the appetite for chips is growing exponentially, the exact level of demand at any given time can be very difficult to predict. This leads to shortages and periods of oversupply, which can result in sharp moves in the share prices of chip-making companies, as experienced by Nvidia in 2022. The key point though is that over the long term demand for chips is growing at a powerful pace.
As chips play an increasingly vital role in our lives, they have become an important geopolitical issue. In 1990 the US produced 37% of the world’s chips1. Today the US government believes that figure has fallen to only 12%1, which is roughly on a par with Europe2.
The world is hungry for chips and Joe Biden is committed to significantly increasing production in the US, through the CHIPS Act. It is estimated that the construction of between $223bn and $260bn 3 worth of US semiconductor production facilities is proposed or under way, supported by $52.7bn worth of grants4. As these figures highlight, semiconductor fabrication plants – widely known as fabs – are costly to build. Construction is also time consuming, with even the most efficient projects taking in excess of three years to complete. In addition, there is the need to recruit the skilled staff vital for manufacturing.
