The race to develop microchips capable of harnessing generative artificial intelligence technology is gaining pace as Nvidia has seen record sales nearly doubling as a result of their new AI chip.

Nvidia is reported to have already cornered about 95% of the AI chip market, and is mostly known as a chip-maker for data centres, and graphics cards for desktop computers and video games.

The company revealed it amassed $13.5 billion (£10.67bn) in revenue since this May, in its 2023 Q2 earnings. The earnings are largely credited to AI, as Nvidia designs the microchips used in data centres that run and train large language models like ChatGPT.

This is in line with predictions from Gartner which projected that the global AI chip market revenue would reach around $53bn (£42bn) this year.

Data centre revenue more than doubled in Q2, and Nvidia’s profit from this alone was up 843% year over year, amassing $6.188bn (£4.89bn). Even their gaming sales are back up, not quite at pre-pandemic levels, but increased 22% year over year.

Nvidia’s direct competitors are Intel and AMD which also design microchips for desktops and data centres, but have yet to release their own competing AI chip, meaning there is little current direct competition to corner the market.

Intel is set to release an AI chip potentially in time for the final quarter of 2023, but it may struggle to compete against a more established AI chip in the data centre market.

On the other hand, Nvidia is projecting another $16bn (£12.65bn) in revenue for the rest of 2023, and this is before their newer AI chip, GH200, is set to be released.

Nvidia is not the only chip designer making moves to profit off the AI-boom: designers of small microchips are also working on advanced AI microchips.

While Nvidia has cornered the data centre chip market, the race to dominate Edge AI – AI processors on devices without a cloud connection – is emerging, as Arm, Qualcomm, and Mediatek aim for centre stage.

Recommended reading

• Global AI Chips Revenue to Increase 21% in 2023, Gartner Forecasts
• UK Microchip Firm Arm Files IPO in US Nasdaq
• UK to Spend £100M on Chip Procurement Amid AI Computing Race

The drive to introduce AI-capable chips to smartphones could provide emerging technologies with an extra boost – one of the biggest obstacles for further advancements is the costly necessity of running models remotely on cloud servers. Natively operating AI on mobile devices would significantly reduce costs, and allow alternative developers to access the foundational tech.

Another major benefit to native AI operation is in sustainability – data centres already consume mass amounts of energy and water, with researchers from the University Colorado Riverside claiming ChatGPT training consumed around 700,000 litres of water, and a chatbot conversation consisting of 20-50 questions uses around 500 ml of water alone.

Meta has so far made two partnerships with Qualcomm and Mediatek to create these mobile AI microchips. MediaTek has said they will release a new chip to work with Meta’s Llama 2 AI model, after Qualcomm announced their partnership with the tech giant in July.

Arm, on the other hand, has already been working on this technology, but has focused on producing chips capable of native AI processing for varied Internet of Things (IoT) devices rather than just smartphones and tablets.