Anam's interactive avatars now run on CoreWeave Cloud

CoreWeave Cloud will power the next phase of Anam's interactive avatar platform.

Under the agreement, Anam will run training and inference workloads on NVIDIA RTX PRO 6000 GPU nodes across CoreWeave infrastructure in the United States and Europe. The goal is simple: give our avatar platform the compute headroom, regional footprint, and production reliability needed for real-time face-to-face AI conversations at scale.

That matters because interactive avatars are not ordinary video assets. Every conversation is generated live: the user speaks, the agent thinks, the voice is produced, and the face renders in sync with that audio. When latency drifts, the conversation stops feeling natural.

Why does interactive avatar infrastructure matter?

Model quality is only half the question. Infrastructure determines whether that model can run fast enough for live conversation.

A pre-rendered avatar video can take seconds or minutes to generate because the user is not waiting inside a live conversation. Anam's real-time interactive avatars work differently. They need to respond while the user is still emotionally present in the exchange.

That changes the whole stack.

The avatar has to receive audio, predict motion, render pixels, stream video, handle interruptions, and stay synchronized with the agent's voice. This all happens while the rest of the agent pipeline is also running: speech recognition, LLM inference, tools, memory, and text-to-speech.

For a single demo, that is hard enough. For production deployments across customer support, coaching, education, sales, and simulation, it becomes a compute problem very quickly.

This is why infrastructure becomes part of the user experience. A faster, more reliable cloud does not sit behind the product as an invisible line item. It shows up as tighter turn-taking, steadier video quality, and fewer moments where the user feels the system waiting on itself.

What does CoreWeave Cloud add for Anam?

CoreWeave is built for AI workloads that need dense GPU compute and predictable production performance. Its GPU compute platform is designed around NVIDIA-accelerated infrastructure, high-throughput networking, and AI workloads that move from research into production.

For Anam, the agreement supports three things.

1. More headroom for training. Our models keep getting more expressive, more responsive, and more controllable. That requires training infrastructure that can handle bigger experiments and faster iteration cycles.

2. Lower-latency inference at scale. Avatar conversations are live inference workloads. They need predictable response times while many sessions are running at once.

3. Regional production capacity. Running across CoreWeave infrastructure in the United States and Europe gives us more options for serving customers closer to their users.

CoreWeave has also moved early on the specific GPU class we will use. The company announced that it became the first AI cloud provider to offer NVIDIA RTX PRO 6000 Blackwell GPU instances at scale, which is the kind of infrastructure profile that matters when the workload is both media-heavy and latency-sensitive.

What did CoreWeave say about the agreement?

"Real-time AI interactions leave no room for latency or reliability gaps," said Jon Jones, chief revenue officer of CoreWeave. "CoreWeave's AI cloud platform gives Anam the compute performance and global footprint to deploy emotionally responsive AI avatars at scale, where production performance is what users actually experience."

That last part is the reason this partnership matters to us.

Production performance is what users actually experience. No one on the other end of an avatar conversation cares what the benchmark looked like in isolation. They care whether the agent responds naturally, whether the face stays in sync, whether interruptions work, and whether the session holds up when traffic grows.

CoreWeave's public performance record is part of the draw here too. The company has pointed to record MLPerf results, Platinum rankings in SemiAnalysis ClusterMAX, and a recent #1 ranking for inference speed and price-performance for Moonshot AI's Kimi K2.6 in independent benchmarking by Artificial Analysis.

What does this mean for Anam customers?

Our customers use Anam when the interface matters as much as the model behind it.

For customer-facing AI agents, that might mean an avatar that can answer onboarding questions, explain a product, or hand off to a human with the full context intact. The post on AI avatars for customer success covers why this matters in support and onboarding workflows.

For training and simulation, it might mean a realistic roleplay partner that employees can practice with repeatedly. The same foundation applies to interactive avatars in learning and development and AI sales coaching, where the goal is sustained attention, trust, and repeat practice.

In those products, small delays compound. A half-second pause feels different when it happens once. It feels much worse when it happens every turn for a 20-minute session.

CoreWeave gives us more capacity to keep those experiences fast as demand grows.

Why are real-time avatars so compute-heavy?

Anam's platform does more than animate a mouth over audio.

The avatar has to generate facial motion, lip sync, eye behavior, head movement, and expression in real time. The system also has to recover cleanly when the user interrupts, shifts context, or changes emotional tone.

That is why our work on Cara 3 interactive avatars, adaptive bitrate streaming for avatars, and building AI voice agents with a face keeps coming back to the same point: realism only works when it is fast enough for conversation.

The infrastructure choice affects that directly.

Training needs enough compute to improve the model. Inference needs enough compute to run many live sessions without delay. Networking needs to keep media moving smoothly. Operations need to keep all of that reliable for customer deployments.

CoreWeave Cloud gives Anam a stronger foundation for that full loop.

What does this unlock next?

Ben Carr, CTO of Anam, put it this way:

"Our interactive avatars create more natural, emotionally intelligent interactions in real time. That's why our customers see users stay longer, adopt faster, and convert more: a real-time avatar holds attention in a way text and voice never have. But it's demanding: each avatar conversation takes a lot of compute. CoreWeave Cloud gives us the ability to run our avatars at large scale without compromising on uptime SLAs or response latency."

That is the whole story.

The next wave of AI agents will do more than answer questions. They will teach, coach, sell, support, onboard, and roleplay in real time. The interface for that work needs to feel present enough for people to stay with it.

For Anam, that means building the best interactive avatars we can, then running them on infrastructure that can keep up with the demand.

CoreWeave is now part of that foundation.

For more on where this category is going, read our posts on conversational video AI and what interactive avatars mean for businesses.

Frequently asked questions

What did Anam and CoreWeave announce?

Anam and CoreWeave announced an agreement for Anam to run training and inference workloads on CoreWeave Cloud. The work will support Anam's real-time avatar platform across infrastructure in the United States and Europe.

Why does cloud infrastructure matter for interactive avatars?

Interactive avatars are generated live during conversation, so latency and reliability directly shape the user experience. Cloud infrastructure affects how quickly the avatar can respond, how stable the video stream feels, and how well the system handles production traffic.

What GPU infrastructure will Anam use on CoreWeave?

Anam will run workloads on NVIDIA RTX PRO 6000 GPU nodes across CoreWeave infrastructure. These nodes support the training and inference demands of real-time, media-heavy AI workloads.

How does this affect Anam customers?

The agreement gives Anam more capacity to serve real-time avatar sessions at scale. Customers should benefit from Anam's continued focus on low latency, production reliability, and regional deployment options.

Does this change how developers build with Anam?

No. Developers can continue using Anam's existing APIs, SDKs, and integrations. The CoreWeave agreement supports the infrastructure behind the platform rather than changing the developer surface.

What is an interactive avatar?

An interactive avatar is a real-time AI face that can listen, respond, and stay visually synchronized with speech during a live conversation. Unlike pre-rendered video avatars, interactive avatars generate the conversation and the face in the moment.

What is CoreWeave Cloud?

CoreWeave Cloud is an AI cloud platform built around high-performance GPU infrastructure for training, inference, and other compute-heavy workloads. It is used by AI teams that need dense compute, strong networking, and production-scale performance.

Where can teams try Anam?

Teams can try Anam through Anam Lab or build directly with the API and SDKs. Anam supports real-time avatars for customer agents, coaching, training, onboarding, and other live conversational applications.