#UAE #inference - DIFC-based AI infrastructure startup Enclave42 has launched a commercial AI inference service, offering businesses more control over how their AI workloads are executed, verified and deployed. The platform provides confidential inference using open-source and frontier AI models, together with TEE-backed execution for supported models. Enclave also offers customer-verifiable attestation rather than relying solely on trust in the infrastructure provider’s security claims. The company sources and tests GPU infrastructure for price and performance and is now deploying GPUs in the UAE for sovereign workloads. Following a successful pilot phase, self-service access to the platform is now available.
SO WHAT? - Most AI infrastructure providers ask customers to trust that their data stays protected. Enclave42’s approach lets customers check that for themselves. This is important for most for regulated UAE sectors weighing sovereign AI options and other businesses handling confidential data. Data residency sometimes gets treated as equivalent to confidentiality, whether the infrastructure operator itself can access that data. Enclave allows customers to submit their own verification request and confirm independently that the confidential computing environment was actually active, rather than simply claimed.
KEY POINTS:
Dubai-based AI infrastructure startup Enclave42 has launched a commercial AI inference service, offering SME and mid-market businesses lower-cost GPU access with assured data-confidentiality.
Enclave42 runs AI models inside Trusted Execution Environments (TEEs), hardware-isolated environments designed to prevent the underlying operating system, hypervisor or infrastructure administrator from accessing the data and code being processed, using Intel TDX enclaves on NVIDIA GPU infrastructure.
Customers can verify this protection themselves by generating a unique verification code, downloading attestation evidence and independently confirming the confidential environment was operating as claimed, rather than relying solely on the provider’s assurances.
The platform currently offers 13 open-source large language models through a single API covering language, reasoning, code, vision, embeddings, speech, reranking and multimodal workloads, with model access tied to subscription tier.
Pricing starts at $99 for subscription access to an eight-model service, with a $399 Business tier unlocking frontier-class models. Actual inference use is billed separately through a pay-as-you-go wallet with real-time usage monitoring. The platform also has a free ‘explore’ option for new customers.
The company currently has access to around 100 cloud GPUs sourced for price and performance, and is expanding beyond externally sourced capacity by deploying its own GPU infrastructure within the UAE.
Enclave42 is based at the DIFC Innovation Hub and is a partner of Ignyte, the DIFC-backed startup ecosystem platform, giving it access to UAE and GCC-focused onboarding, architecture and cost-optimisation support for customers.
The company’s next phase moves beyond shared model APIs toward private model deployments, dedicated GPU infrastructure, private inference endpoints and custom model hosting, letting customers eventually configure their own combination of model, GPU capacity, scaling and access controls.
ZOOM OUT - The technology behind Enclave42’s confidentiality claims comes from Intel TDX technology. Intel Trust Domain Extensions (TDX) is a hardware-based confidential computing feature built into Intel Xeon processors. It works by isolating entire virtual machines, called Trust Domains, from the hypervisor, operating system and even a cloud provider’s own administrators. Memory encryption happens at the hardware level, and remote attestation lets a customer verify that their workload really is running inside a genuine protected environment before they send it any data, the same mechanism Enclave42 exposes to its own customers. Intel positions TDX around four use cases: data sovereignty, IP protection, regulatory compliance and trusted AI deployment.
[Written and edited with the assistance of AI]
Source: Enclave42
Read more about UAE AI platforms:
e& first to embed AI in core connectivity (Middle East AI News)
OnDemand AI apps to run across Qualcomm platforms (Middle East AI News)
OpenAI now offers inference residency to UAE (Middle East AI News)
CNTXT AI raises $60M to deploy sovereign AI globally (Middle East AI News)