#Saudi #infrastructure – AI inference leader Groq sent a massive shipment of LPUs and GroqRacks to Saudi Arabia during the second half of December, destined for its new inferencing data centre in the Kingdom. Announced during the Global AI Summit (GAIN) in Riyadh in September, the new data centre is expected to be the world’s largest infrastructure cluster, offering at least 25 million tokens-per-second of compute by the end of Q1 2025. Former Aramco Digital CEO Tareq Amin confirmed receipt of the Groq shipment via a LinkedIn exchange with Groq CEO Jonathan Ross.
SO WHAT? – Groq and Aramco Digital made a joint announcement at GAIN in September about the development of a state-of-the-art inferencing data centre service in Saudi Arabia. However, the two organisations outlined a plan to provide an on-demand token-as-a-service for AI inference via a new local data centre by the end of 2024 Although it appears that the schedule to go live was moved back, the news of a big shipment of LPUs (language processing units) from Groq in December confirms that any delay was a minor one. So, it seems that we can expect a Groq inference services launch during the next few weeks.
Here are some key points in the timeline for the planned Groq data centre:
-
Groq and Aramco Digital first announced their partnership during LEAP24 in March 2024, together with a stated intention to build the world’s largest AI infrastructure-as-a-service data centre in Saudi Arabia.
-
In September, Groq and Aramco Digital reconfirmed their commitment to build an AI infrastructure-as-a-service data centre at the Global AI Summit (GAIN). Plans were announced to provide 20% to 40% of Groq’s on-demand token-as-a-service for AI inference via a new Saudi data centre by the close of 2024.
-
Groq CEO Jonathan Ross then articulated the company’s ultimate goal of deploying one billion tokens-per-second of capacity in Saudi Arabia.
-
The registration of Groq’s regional headquarters in Riyadh was also announced at GAIN in September.
-
At Fortune Brainstorm AI last month, Ross doubled-down on the company’s goal to deploy at least 25 million tokens-per-second of compute by the end of Q1 2025.
-
He also reconfirmed Groq’s commitment to work towards a target of 1 billion tokens per second together with Aramco Digital, which will provide the investment required to meet that goal.
-
Last week Ross shared a video via LinkedIn of cargo pallets of Groq hardware being loaded to fill a National Air Cargo Jumbo jet, apparently destined for Saudi Arabia. Former Aramco Digital CEO Tareq Amin confirmed receipt of a shipment destined for the new Saudi Groq data centre when reposting the news.
ZOOM OUT – The new Groq AI inference data centre has been planned as a global venture, to both impact AI usage far beyond Saudi Arabia’s borders and position the Kingdom at the centre of it all. Scaling to 25 million tokens-per-second of compute by the end of March, the data centre is expected to be the world’s largest infrastructure cluster dedicated to AI. Aramco’s backing will allow Groq to scale very fast, providing services to clients across EMEA and perhaps some other parts of Asia too. If Groq is able to meet its ultimate goal of deploying one billion tokens-per-second of capacity – which will take some doing – this could provide compute for 4 billion people globally.
Read more about Groq’s plans for Saudi Arabia: