Cloudflare today announced several new products and partnerships aimed at making AI more accessible and optimized for developers.
First, the company unveiled Workers AI, a serverless AI platform powered by NVIDIA GPUs on Cloudflare's global network. This allows developers to leverage pre-trained AI models with a pay-as-you-go pricing model, reducing infrastructure costs and management. Models can be deployed at the edge, close to end users, for low-latency performance.
Next, to support building customized AI applications, Cloudflare also launched Vectorize, a vector database for indexing and storing AI model embeddings. This provides fast, affordable storage for powering AI use cases beyond basic inference.
Additionally, Cloudflare announced AI Gateway, giving organizations visibility into their AI workloads and tools to cache responses, rate limit requests, and monitor deployments across clouds. This assists with optimizing latency, managing costs, and maintaining reliability.
Cloudflare's vision for AI emphasizes three pillars:
- Compliance: Recognizing the increasing governmental regulations around AI and data residency, Cloudflare offers a network that stretches over 300 cities. Its Data Localization Suite ensures that AI deployments remain local, catering to in-country data storage mandates.
- Budget: AI can be expensive. Cloudflare says it is determined to provide an affordable AI platform, ensuring developers are charged only for what they use.
- Privacy: With AI becoming integral to customer experiences, Cloudflare reassures its users of a privacy-first approach. They have committed that customer data utilized for inference will not be used to train large language models.
"Our announcements mark the beginning of Cloudflare's journey into AI, much like Workers did for serverless computing," said Matthew Prince, CEO. "We believe the edge is the optimal location for inference, and our platform makes it seamless for any developer to build AI-powered apps."
Bolstering its AI platform, Cloudflare revealed partnerships with NVIDIA, Microsoft, Hugging Face, and Meta. The collaboration with Meta will bring the Llama 2 open source AI model to Cloudflare Workers, granting all developers access to a robust large language model.
Overall, Cloudflare is aiming to democratize AI and eliminate infrastructure barriers through its edge network and serverless approach. By abstracting away complexity, it hopes to enable a broader range of innovators to experiment with AI and build the next generation of intelligent applications.