G42 and Liquid AI team up to deliver sovereign generative AI for finance, biotech, and telecom sectors

G42, an Abu Dhabi-headquartered artificial intelligence conglomerate, and Liquid AI, a U.S.-based foundation model developer spun out of MIT, have announced a commercial partnership aimed at building and deploying sovereign generative AI infrastructure across the Middle East and North Africa (MENA), the Global South, and other emerging enterprise markets. The collaboration leverages G42’s regional infrastructure and Liquid AI’s computationally efficient Liquid Foundation Models (LFMs) to deliver enterprise-grade, private AI solutions at scale.

This agreement reflects an accelerating push toward local AI capabilities, as organizations demand on-premises deployment, data privacy compliance, and lower latency across sectors such as finance, telecommunications, biotech, energy, and consumer electronics. The partnership also follows growing international interest in sovereign AI architecture, particularly following the U.S.–UAE AI acceleration framework announced earlier in 2025.

G42 subsidiaries Core42 and Inception will provide cloud infrastructure and collaborate on model training, co-design, and deployment, while Liquid AI will contribute its lightweight, high-performance multimodal models designed for regulated and resource-constrained environments.

What are the core elements of the G42–Liquid AI collaboration and how will it be implemented across different regions?

At the heart of the partnership is a shared ambition to roll out sovereign, enterprise-ready AI tailored to regional requirements. G42’s Core42 will provide the hybrid cloud infrastructure to support model training and deployment across G42-owned data centers. Inception, the innovation division within G42, will collaborate with Liquid AI to build multimodal foundation models capable of handling complex enterprise use cases involving text, structured data, and vision inputs.

Liquid AI’s models will be trained and fine-tuned on regionally relevant datasets while being deployed through Core42’s infrastructure footprint. This will allow customers—particularly in data-sensitive jurisdictions—to access enterprise-grade AI that complies with data sovereignty regulations and avoids routing through hyperscaler public cloud platforms.

Early-stage deployments are expected to roll out across the Gulf, South Asia, and parts of Africa, with future expansions targeting Latin America and Southeast Asia.

Why are Liquid Foundation Models seen as a differentiated alternative to transformer-based architectures in enterprise AI?

Liquid AI’s foundation models utilize a novel architecture based on “liquid neural networks,” which diverge from traditional transformer designs by integrating principles from physics, differential equations, and signal processing. This approach allows the models to dynamically adjust their internal representations in response to new inputs—resulting in faster learning, lower compute requirements, and enhanced generalization, even on smaller datasets.

The models are available in multiple configurations: a compact 1.3-billion parameter model, a 3.1B version, and a larger 40B Mixture-of-Experts (MoE) model. Despite their smaller size compared to massive transformer models, Liquid Foundation Models have demonstrated state-of-the-art performance on industry-standard benchmarks for enterprise tasks.

These attributes make them ideal for deployment in environments with hardware constraints, energy limitations, or compliance risks, including hospitals, banks, and sovereign data centers. Experts see these models as an enterprise-ready, explainable, and scalable alternative in contrast to the more opaque and compute-heavy offerings from larger U.S. and Chinese model providers.

How does this tie into broader analyst sentiment and investor interest in regionalized AI models and sovereign AI frameworks?

Institutional sentiment across the Gulf and emerging economies is increasingly aligned with the concept of sovereign AI—referring to models that are not only trained and deployed within national borders but also governed by local laws. The G42–Liquid AI partnership is being viewed as a template for this model, particularly in jurisdictions where privacy laws prohibit offshoring sensitive data to hyperscaler-operated infrastructure.

Liquid AI raised $250 million in a Series A round in December 2024, backed by AMD and venture firms aligned with enterprise AI expansion. AMD is closely involved with optimizing LFMs across its CPU, GPU, and accelerator ecosystems, enhancing Liquid AI’s compute efficiency and cost-to-performance ratio. Analysts say this backing has positioned Liquid AI as one of the few credible challengers to more public-facing AI companies like OpenAI, Anthropic, and Mistral.

Investor interest in localized deployment models has surged following new regulations in Europe, the UAE, and India that mandate stricter control over training data, model explainability, and deployment security. As these conditions spread globally, partnerships like G42–Liquid AI are being seen as the next frontier of compliant AI development.

What financial and operational details are known about the partnership and its timeline?

While the financial terms of the partnership were not disclosed, both G42 and Liquid AI have indicated a phased implementation approach. The first wave of deployments is expected by Q4 2025, targeting enterprise clients in regulated sectors. According to internal sources, the initial roadmap includes finance-specific large language models for compliance automation, multimodal models for diagnostic imaging and health records, and telecom-specific chat agents integrated with regional customer data systems.

Liquid AI’s latest funding round, which brought its post-money valuation above $2 billion, has enabled the development of on-premise developer kits, low-power inference tooling, and fast-fine-tune capabilities. These features align with G42’s enterprise focus and could allow models to be trained or adapted at customer premises in a confidential computing environment.

G42, meanwhile, continues to expand its AI partnerships globally. Recent collaborations include investments in Mistral AI, a regional partnership with Cisco, and a strategic link with OpenAI and Microsoft Azure under the U.S.–UAE AI framework.

How does this partnership position both companies within the global AI infrastructure and model landscape?

G42 is steadily building a vertically integrated AI stack that spans compute infrastructure, sovereign data hosting, model development, and enterprise deployment. With Core42 and Inception at the helm of its AI push, the firm is capitalizing on Abu Dhabi’s digital strategy to become a hub for AI innovation. By bringing Liquid AI’s efficient models into its fold, G42 expands its offerings beyond infrastructure into proprietary model IP—an area dominated by U.S.-based players.

Liquid AI, in contrast, benefits from access to a rapidly growing customer base across the Global South without building its own cloud or sales infrastructure. This partnership allows it to stay capital-efficient while deploying models at a meaningful global scale. Its positioning as a builder of efficient, low-compute models sets it apart from larger players who continue to scale up models at exponentially rising costs.

Together, the companies aim to deliver a value proposition that combines explainability, cost-efficiency, sovereign compliance, and performance—an intersection of requirements that is becoming increasingly central in enterprise AI decision-making.

What is the future outlook for the partnership and the wider enterprise AI market it targets?

Market analysts expect that demand for localized and sovereign AI will accelerate over the next two years as governments, regulators, and enterprises seek models that can be governed, audited, and trusted. Specific to this partnership, further announcements are expected later in 2025, possibly including pilot deployments in UAE government agencies, Gulf financial institutions, and African telecom operators.

Both companies have reiterated their long-term commitment to sovereign AI and plan to publish joint research on multimodal model development later this year. Additionally, the emergence of regulatory sandboxes in markets like the UAE and Saudi Arabia may further accelerate the use of Liquid Foundation Models in controlled, region-specific environments.

The AI race is no longer defined solely by model size—it is increasingly defined by who can deliver compliant, efficient, and locally trusted systems. In this evolving landscape, the G42–Liquid AI partnership appears positioned as a frontrunner for the new enterprise AI era.