Ceramic.ai emerges from stealth with $12M to make AI training faster and more cost-effective

Ceramic.ai, an AI startup founded by former Google VP of Engineering and Gradient Ventures founder Anna Patterson, has emerged from stealth with $12 million in funding to reshape how enterprises train and fine-tune AI models.

Backed by investors including NEA, IBM, Samsung Next, Earthshot Ventures, and Alumni Ventures, Ceramic’s infrastructure is designed to improve efficiency, making AI training faster and less expensive for partners like AWS and Lambda.

This funding will accelerate product development and help meet the demand of enterprise customers looking for better AI training solutions.

While traditional AI infrastructure can scale by a factor of 10, the true exponential growth—100x or more—demands a fundamental overhaul. That’s the problem Ceramic.ai is solving.

Its technology enables AI models to train with long contexts and across any cluster size, allowing companies to develop and scale their AI models more effectively. In early trials, Ceramic.ai has shown speeds up to 2.5 times faster than state-of-the-art platforms when running on NVIDIA H100 GPUs. For large-scale, long-context models, it stands alone as the best option.

“In the midst of a surge in AI adoption, too many companies are still hindered by barriers to scale – from prohibitive costs to limited infrastructure,” said Anna Patterson, founder and CEO of Ceramic.ai. “We’re democratizing access to high-performance AI infrastructure so companies can navigate the complexity of AI training without spending hundreds of millions in research and engineering resources. But the shift to enterprise AI isn’t just about better tools – it’s about changing how businesses work. If AI adoption were a baseball game, we’d still be singing the national anthem.”

Breaking the Barriers to Enterprise AI

AI investment has surged, climbing from $16 billion in 2023 to an expected $143 billion by 2027. Yet, scaling AI remains a struggle for many companies. Building AI infrastructure is costly and resource-intensive, leaving smaller enterprises at a disadvantage while tech giants pour billions into proprietary systems.

Ceramic.ai is tackling this challenge head-on. Its software delivers an enterprise-ready solution that improves scalability, reduces costs, and simplifies the AI training process. The company’s platform is uniquely capable of training models with long-context data, outperforming existing benchmarks and delivering high efficiency even for models with 70 billion parameters or more.

Anna Patterson, founder and CEO of Ceramic.ai, put it bluntly: “Too many companies are stuck dealing with sky-high costs and infrastructure roadblocks. We’re making high-performance AI infrastructure accessible, so companies can scale AI training without burning through hundreds of millions in research and engineering. If AI adoption were a baseball game, we’d still be singing the national anthem.”

A Different Approach to AI Training

Ceramic.ai isn’t just another AI infrastructure company—it’s rethinking the entire training process. Speed and efficiency are at the core of its approach. Compared to open-source alternatives, its platform delivers training speeds up to 2.5 times faster while cutting costs. That means businesses can develop AI models more efficiently without overspending on compute resources.

But speed alone isn’t enough. Ceramic.ai stands out with its ability to handle long-context training, something other solutions struggle with. For enterprises working with massive datasets, this translates to more accurate models and better overall performance. In fact, the company has outperformed all reported benchmarks for long-context training, proving its ability to maintain efficiency even with models exceeding 70 billion parameters.

The platform also improves reasoning models. In recent testing, Ceramic trained an AI that pushed its Pass@1 score on GSM8K from 78% to 92%, surpassing Meta’s Llama 70B 3.3 base model and outperforming DeepSeek’s R1 score of 84%.

Data processing is another area where Ceramic.ai is making an impact. Instead of the usual approach that either masks irrelevant documents or forces models to pay attention to unnecessary data, Ceramic reorders training data so that each micro-batch is aligned by topic. This optimization ensures models learn more efficiently, significantly improving training outcomes.

Early enterprise trials have already shown that this approach reduces costs and enhances model performance. With partners like AWS and Lambda on board, Ceramic.ai is set to bring these benefits to even more businesses.

Fueling Growth with Strategic Investment

Ceramic.ai’s $12 million seed round will support its rapid expansion, product development, and enterprise adoption. Investors see its potential to redefine AI training efficiency.

“AI has been like a rocket tied to a horse-drawn carriage—until now,” said Lila Tretikov, Partner and Head of AI Strategy at NEA. “Ceramic.ai has shattered a major bottleneck in model training, making it faster, more efficient, and truly scalable.”

IBM is also backing the company’s vision. “We’re excited to partner with Ceramic to bring down AI compute costs and make training more efficient,” said Emily Fontaine, Vice President at IBM Global Head of Venture Capital.

As AI adoption accelerates, enterprises need infrastructure that can keep up. Ceramic.ai is positioning itself as the go-to solution for companies looking to train and scale AI without breaking the bank.