Abu Dhabi's G42, Which Is Building an 'AI Supercomputer' in Dallas, Gets $1.5B from Microsoft

A month ago, technology holding group G42 announced it was building an “8-exaFLOP AI supercomputer” in Dallas, the third cluster of a constellation of AI supercomputers it’s building with California-based Cerebras Systems. Today, the Abu Dhabi, United Arab Emirates-based group had something even bigger to announce: a $1.5 billion investment from Microsoft.

The strategic investment by Redmond, Washington-based Microsoft—which earlier put big chips on the AI table by investing $13 billion in ChatGPT creator OpenAI—aims to strengthen G42 and Microsoft’s collaboration on “bringing the latest Microsoft AI technologies and skilling initiatives to the UAE and other countries around the world.”

As part of this expanded, “long-standing” partnership, Brad Smith, Microsoft’s vice chair and president, will join the G42 board of directors.

The expanded collaboration “will empower organizations of all sizes in new markets to harness the benefits of AI and the cloud while ensuring they are adopting AI that adheres to world-leading standards in safety and security,” G42 said.

As part of the collaboration, G42 will run its AI applications and services on Microsoft Azure and “partner to deliver advanced AI solutions to global public sector clients and large enterprises.” G42 and Microsoft will also work together “to bring advanced AI and digital infrastructure to countries in the Middle East, Central Asia, and Africa, providing these nations with equitable access to services to address important governmental and business concerns while ensuring the highest standards of security and privacy,” the companies said.

Aiming for ‘greater cooperation and synergy globally’

“Microsoft’s investment in G42 marks a pivotal moment in our company’s journey of growth and innovation, signifying a strategic alignment of vision and execution between the two organizations,” Sheikh Tahnoon bin Zayed Al Nahyan, Chairman of G42, said in a statement. “This partnership is a testament to the shared values and aspirations for progress, fostering greater cooperation and synergy globally.” 

The partnership also aims to support the development of “a skilled and diverse AI workforce and talent pool that will drive innovation and competitiveness for the UAE and broader region with the investment of $1B in a development fund for developers,” G42 said.

Bringing AI and digital infrastructure ‘to underserved nations’

Microsoft’s Smith said the two companies “will work together not only in the UAE, but to bring AI and digital infrastructure and services to underserved nations.”

“We will combine world-class technology with world-leading standards for safe, trusted, and responsible AI, in close coordination with the governments of both the UAE and the United States,” Smith added in a statement.

Peng Xiao, group CEO of G42, said his company will leverage Microsoft’s $1.5 billion strategic investment by “advancing our mission to deliver cutting-edge AI technologies at scale. This partnership significantly enhances our international market presence, combining G42’s unique AI capabilities with Microsoft’s robust global infrastructure. Together, we’re not only expanding our operational horizons but also setting new industry standards for innovation.”

Just the most recent of several collaborative moves in last year

Over the last year, Microsoft and G42 have made other collaborative moves. They include a joint plan announced in April 2023 to develop AI solutions tailored for the public sector and industry, leveraging Microsoft’s partner ecosystem and cloud capabilities. In September 2023, the companies entered into an agreement to introduce sovereign cloud offerings and collaborate on unlocking the potential of advanced AI capabilities on the Azure public cloud platform. And last November, Microsoft announced the availability of G42’s Jais Arabic Large Language Model on the new Azure AI Cloud Model-as-a-Service offering.

AI supercomputer in Dallas to ‘be available’ this quarter

The Condor Galaxy 3 AI Supercomputer is being built in Dallas by G42 and Cerebras Systems.[Image: Condor Galaxy]

Now, back to that game-changing supercomputer in Dallas. Last month, G42 and California-based Cerebras Systems, a pioneer in accelerating generative AI, announced the build of Condor Galaxy 3, the third cluster of their constellation of AI supercomputers called “the Condor Galaxy.” The one being built in Dallas features 64 of Cerebras’ newly announced CS-3 systems, all of which are powered by the industry’s “fastest AI chip,” the Wafer-Scale Engine 3.

The Dallas-based Condor Galaxy 3 will deliver 8 exaFLOPs of AI with 58 million AI-optimized cores. In case you’re wondering, that’s a lot, lot, more than what your laptop delivers, to put it mildly.

The Cerebras and G42 strategic partnership “already delivered 8 exaFLOPs of AI supercomputing performance via Condor Galaxy 1 and Condor Galaxy 2, each amongst the largest AI supercomputers in the world,” G42 said last month.The new one in Dallas brings the current total of the Condor Galaxy network to 16 exaFLOPs. It’s slated to “be available” in Q2 2024.

Training some of the leading open-source AI models in the industry

“With Condor Galaxy 3, we continue to achieve our joint vision of transforming the worldwide inventory of AI compute through the development of the world’s largest and fastest AI supercomputers,” Kiril Evtimov, group CTO of G42, said of the Dallas cluster. “The existing Condor Galaxy network has trained some of the leading open-source models in the industry, with tens of thousands of downloads. By doubling the capacity to 16 exaFLOPs, we look forward to seeing the next wave of innovation Condor Galaxy supercomputers can enable.”

So how fast is the AI supercomputer being built in Dallas? At its heart are those 64 Cerebras CS-3 Systems. Each is powered by the new 4 trillion transistor, 900,000 AI core WSE-3. Manufactured by the Taiwan Semiconductor Manufacturing Company at the 5-nanometer node, the WSE-3 “delivers twice the performance at the same power and for the same price as the previous generation part,” G42 said. Designed to train the industry’s largest AI models, WSE-3 delivers “an astounding 125 petaflops of peak AI performance per chip,” G42 said.

Andrew Feldman, co-founder and CEO of Cerebras, said the Dallas-based Condor Galaxy 3 and follow-on supercomputers “will together deliver tens of exaflops of AI compute.”

“This marks a significant milestone in AI computing, providing unparalleled processing power and efficiency,” Feldman added in a statement.

G42 says Condor Galaxy has already trained “state-of-the art industry leading generative AI models, including Jais-30B, Med42, Crystal-Coder-7B and BTLM-3B-8K.”

Training a leading large language model—in one weekend

Jais 13B and Jais30B are “the best bilingual Arabic models in the world,” G42 said, and are now available on Azure Cloud. BTLM-3B-8K is the number-one leading 3B model on HuggingFace, offering 7B parameter performance in a light 3B parameter model for inference. Med42, developed with M42 and Core42, is “a leading clinical LLM, trained on Condor Galaxy 1 in a weekend and surpassing MedPaLM on performance and accuracy.”