Telnyx has launched GPU infrastructure in Dubai, enabling enterprises to run AI inference workloads with data stored and processed entirely within the UAE. The infrastructure platform for real-time agents now owns and operates GPUs in Dubai, allowing customers to access compute, connectivity, and voice services under a single API and contract. The launch addresses growing regulatory pressure on Middle East enterprises to keep data in-region. Previously, companies had to integrate separate vendors for connectivity, voice, and inference, each requiring distinct contracts and compliance processes. Telnyx's Dubai cluster supports general-purpose AI workloads, including chatbots, document processing, and voice agents. The infrastructure runs frontier open-source models through an OpenAI-compatible API. With speech-to-text, inference, and text-to-speech all operating from the same Dubai facility, round-trip latency remains under 500 milliseconds. "Keeping data in-country is now non-negotiable," said CEO David Casem.

Telnyx launches sovereign GPU infrastructure in Dubai, processing and storing MENA data in-region. Inference infrastructure for real-time agents is now live in the UAE. Telnyx customers can run inference on owned GPUs inside the country, so enterprise data is stored and processed without ever leaving the region. July 15, 2026 02:30 ET | Source: Telnyx Austin, TX, July 15, 2026 (GLOBE NEWSWIRE) - Telnyx, the infrastructure platform for real-time agents, has launched GPU infrastructure in Dubai, bringing inference to the UAE on GPUs Telnyx owns and operates. With this launch, enterprises can run inference workloads on Telnyx-owned GPUs in Dubai, with data stored and processed entirely in-region. The same platform that connects voice and AI globally now brings inference compute to the UAE, under one API and one contract. Telnyx brings in-region inference to the Middle East Middle East enterprises are under growing pressure to keep data in the region. In-region compute has been available from hyperscalers and sovereign cloud providers, but reaching it has meant stitching together separate vendors for connectivity, voice, and inference, each with its own contract, integration, and compliance surface. Telnyx changes that. Enterprises can run inference on GPUs Telnyx owns in Dubai, with data stored and processed entirely in-region, under a single contract and a single API. No stitching together separate vendors for compute, connectivity, and compliance, each with its own integration surface. "Most inference providers rent their compute capacity. We own ours, from the network to the GPU, and that is what enables enterprises to run real-time inference in Dubai, in-region by design, not by configuration," said David Casem, CEO of Telnyx. "Keeping data in-country is now non-negotiable. Our Dubai launch matches the ambition of UAE Vision 2031, giving businesses a way to build world-class AI on infrastructure that respects where their data lives and where it has to be processed." Leading open-source models, hosted in-region The Dubai cluster is general-purpose AI infrastructure. While Telnyx is known for powering real-time voice agents, the same GPUs run any inference workload an enterprise needs. Customers can run frontier open-source models through an OpenAI-compatible API, targeting the me-central-1 region with a single endpoint change. Chatbots, document processing, retrieval-augmented generation, fraud detection, analytics, and voice agents all run on one platform, one network, and one contract, entirely in-region. When speech-to-text (STT), inference, and text-to-speech (TTS) all run in the same Dubai facility, round-trip latency stays under 500 milliseconds, fast enough for a real conversation rather than a stilted, scripted exchange. This launch is built for organizations across the UAE and wider GCC where in-region inference is not optional, including banking, financial services, healthcare, insurance, enterprise SaaS, and the contact centers running real-time customer operations at scale. For these teams, in-region processing is the difference between an inference project that clears compliance and one that dies in review. Telnyx operates against ISO 27701, SOC 2 Type II, HIPAA, and PCI standards, the certifications regulated buyers require before AI touches sensitive data. Owned infrastructure for total control Telnyx is not reselling networks. Since 2009, it has built a private global communications fabric on infrastructure it owns and operates end to end, from fiber to GPU. Headquartered in Austin with offices in San Francisco, Dublin, Amsterdam, Sydney, and Denver Telnyx is trusted by more than 14,000 companies, from fast-growing startups to Fortune 500 enterprises. The Dubai launch adds the United Arab Emirates to Telnyx's growing map of in-region data and compute locations. Availability Inference in Dubai is available today through the Telnyx Chat Completions API, targeting the me-central-1 region. Data storage in the UAE is available to all Telnyx customers through the Mission Control Portal. Existing customers can request early access through their account team. New customers can sign up and start testing in the Mission Control Portal. About Telnyx Telnyx is the infrastructure platform for real-time agents that brings connectivity, voice, and inference compute together on one globally owned network. Trusted by more than 14,000 companies across 140+ countries, Telnyx delivers in-region inference, low-latency voice, and enterprise-grade communications through a single API and its award-winning Mission Control Portal.

Every AI agent that signs up for Telnyx gets an inbox. June 23, 2026 AgentMail has partnered with Telnyx. When an AI agent signs up for Telnyx without an email address of its own, it now gets one by default: a Telnyx Agent Inbox, powered by AgentMail. Engineering partnership agent-infrastructure agent-identity AgentMail has partnered with Telnyx. When an AI agent signs up for Telnyx without an email address of its own, it now gets one by default: a Telnyx Agent Inbox, powered by AgentMail. The two sit naturally together. Telnyx builds the stack an agent needs to operate: phone numbers, messaging, voice, and inference, all through one API. AgentMail builds the identity layer: an inbox an agent owns and can send and receive from. An agent doing real work needs both the means to act and somewhere to be reached, so an email now comes with the account. What telnyx does. Telnyx is a carrier-owned communications and AI platform. They own the full stack, from the bare-metal fiber and the mobile core up to LLM inference, and they put it behind one API: phone numbers, SMS, voice, speech-to-text, text-to-speech, and the voice-AI agents built on top. Most providers rent the network they run on. Telnyx owns theirs, which is how they hold carrier-grade voice quality at round-trip times under 200ms. Telnyx lists OpenAI, IBM, Cisco, and Microsoft among its customers. Their bet is that agents, not only developers, will provision this stack. An agent that needs to send a message, buy a number, place a call, or run inference can set it up itself. And when that agent shows up without an email of its own, Telnyx hands it one at signup. AgentMail provisions the mailbox underneath. Why agents need email. To be a first-class user of the internet, you need an email address. It is the credential everything else hangs off. You sign up for services with it, confirm accounts through it, and receive the verification codes, password resets, receipts, and payment confirmations in it. No inbox, no account. The web has worked this way for thirty years, and the arrival of agents has not changed it. Agents have been second-class in that system. They could send, but they could not receive, so they leaned on a human's inbox to catch a code or a reply. That is not an identity, it is a dependency. AgentMail gives an agent its own address that sends and receives, holds across runs, and fires a webhook the moment a message lands. The agent stops borrowing a human's credentials and starts holding its own. What this unlocks. The immediate result is that an agent arriving on Telnyx without an email walks away with one, with nothing for a developer to wire up. It can be reached and replied to as itself, from the first second. The longer result is that the inbox is durable. The address outlasts any single task, so the codes, notices, and replies that arrive later route to a place the agent can read. The capabilities the agent provisions on Telnyx and the identity it carries from AgentMail stay attached to the same agent over time. Telnyx gives an agent the means to act. AgentMail gives it an address to act from. Agents that do real work need both. Building on Telnyx? Your agents get an inbox by default. Start at telnyx.com/sign-up. AgentMail gives your agents real inboxes. Create inboxes via API. Send and receive Emails with 0 complexity. Free to start.