Machine Learning & AI

Looking Ahead to 2023: How CoreWeave Is Using NVIDIA GPUs to Advance the New Era of AI and Machine Learning

Looking Ahead to 2023: How CoreWeave Is Using NVIDIA GPUs to Advance the New Era of AI and Machine Learning

CoreWeave raised $100M in strategic funding in December  2022 to help build the infrastructure companies need for the future of AI.

It’s exciting to see how moments this year helped lay the groundwork for what’s promising to be an eventful, fast-moving future for AI. This month, we announced that Magnetar has backed CoreWeave with a $100 million investment. This investment is going directly into scaling our infrastructure, enabling us to deliver a massive capacity of NVIDIA GPU-accelerated compute to AI and machine learning companies, VFX studios, and other enterprise and start-up businesses.

Earlier this year, GPT-NeoX-20B became the largest publicly available language model. Trained by EleutherAI on CoreWeave’s NVIDIA A100 Tensor Core GPU training cluster, the release offered a glimpse into the next generation of open-source AI systems. Since then, CoreWeave’s serverless Kubernetes infrastructure has grown in demand among businesses and researchers looking to build innovative products powered by LLMs.

Generative AI is already becoming more ubiquitous in our everyday lives.  In the past few weeks, ChatGPT, a large language model (LLM)  from OpenAI, gained  1 million new users testing out the platform in the first week of its launch. Friends and family flooded our social feeds with AI-generated self-portraits, notably from Lensa AI, which used the Stable Diffusion model. While these models are not without their own kinks to smooth out, together they show where the industry is heading: forward, and fast.

If it feels hard to keep up with the pace at which AI is moving mainstream, you’re not alone. But it’s only going to move faster. That raises the question: Is this growth sustainable? Does the industry have the infrastructure in place to support the speed at which AI and machine learning are growing—and the immense amount of data that comes with it?

As AI and machine learning continue to rapidly evolve, demand for infrastructure rises with it. Start-ups and enterprises alike will need reliable, flexible, and highly available compute resources at affordable prices to fuel their growth.

NVIDIA H100: Take an order-of-magnitude leap in accelerated computing

In Q1  2023, clients will have access to NVIDIA H100 Tensor Core GPUs on CoreWeave Cloud, making us one of the first providers to offer cloud instances of NVIDIA H100-based supercomputers.  

The NVIDIA H100 Tensor Core GPU delivers unprecedented performance, scalability, and security for every workload. NVIDIA H100 GPUs feature fourth-generation Tensor Cores and the Transformer Engine with FP8 precision, further extending NVIDIA’s market-leading AI leadership with up to 9x faster training and an incredible 30x inference speedup on large language models. For high-performance computing (HPC) applications, the H100 delivers up to 7x higher performance.

As a CoreWeave client, you can  reserve capacity of NVIDIA H100 GPUs today, available at scale in Q1 2023 and starting at $2.23/hr. 

New capabilities, integrations, and features on CoreWeave Cloud  

This year, we expanded our offerings and partnerships to bring more value to clients and our growing community. Some highlights from 2022:

  • Object storage: a simple and powerful storage solution for high volumes of unstructured data, starting at $0.03 / GB per month, with no access or transfer fees
  • New integrations with innovative partners: including Determined.AI for large-scale AI training, Zeet for managing Kubernetes infrastructure, and EleutherAI to help make open-source LLMs more accessible.
  • New AI examples: added library of examples to fine-tune and serve LLMs and generative AI models, like BLOOM and Stable Diffusion.
  • Goose AI: a fully managed API to serve pre-trained, open-source models

We also launched the CoreWeave Accelerator Program, an initiative to provide innovative startups with the NVIDIA GPU-accelerated compute resources they need—on demand and at scale—to help expedite their business growth.  In addition to resource support, this program creates opportunities for startups to increase visibility for their business and cultivate new connections and partnership opportunities.

All these changes reflect CoreWeave’s values of fostering creativity and innovation. Looking to 2023, we have an exciting product roadmap and announcements that enable us to deliver on our commitments to our clients and extend value to their businesses.

Keep an eye on what we’re building here:

Client spotlight: Incredible creations that blew us away 

Speaking of building, our clients continue to amaze us with their ingenuity and products, from innovations in AI that fundamentally change how people engage with technology to more efficient, cloud-based workflows empowering VFX and Animation studios to create better content faster than ever before.

VFX studio finds compute solution for remote teams

Spire Animation Studios took to the virtual stage with their presentation at NVIDIA GTC 2022, in partnership with CoreWeave. Spire’s Engineering leadership shared how their team leveraged a cloud-native workflow to develop an end-to-end animated feature film using the Unreal Engine

This allowed Spire to flip the traditional, linear animation pipeline structure on its head. Now, Spire’s team can iterate in real-time across the country.

“By having faster iterations, we are able to get more of them, and the refinement process, whether we want to give reviews, change the lighting or move characters, can be done interactively, rather than waiting 2 or 3 hours for a render to come back in a traditional setting. We can make our decisions quickly because everything is interactive and everything is at full resolution.” 
Rajesh Sharma, VP of Engineering at Spire Animation Studios

NVIDIA GPUs for AI enable natural language processing (NLP) for Japanese

Yasu Seno and Bit192, Inc. have been busy: they trained a 20-billion parameter Japanese model from scratch, using a slightly modified GPT-NeoX codebase.

The team also released AI Novelist, a chatbot engine turned writing assistant that Yasu and the team used to develop the Starbirth plot. Today, professional writers, editors, and casual users across Japan use AI Novelist to create stories, haikus, and more. Even Yasu’s family is enjoying the tool:

“I demoed the service for my grandmom—who is in her 90’s—and when I showed her that AI Novelist can generate a haiku, she was excited, which made me feel incredibly fulfilled.” 
Yasu Seno, CEO Bit192, Inc.

New faces, new headquarters, and new data centers 

This year, we opened a new data center in the New York City region (LGA1). This new facility provides ultra-low latency to more than 20 million inhabitants across the NYC metropolitan area, including sub 1ms latency to Manhattan, as well as CPU-only instances and high-performance Block and Shared File System storage.

We also put our new headquarters on the map. Home base is now Roseland, New Jersey, with additional offices in Brooklyn, New York, and Philadelphia, Pennsylvania.

All the while, we’ve been growing our team at CoreWeave. It’s incredible to see what was once three guys sitting at a pool table become a team of 70+ employees across the U.S.—all excited about what we’re building and helping others to build. Our engineering team has seen incredible growth as we strive to meet the demands of our growing user base and to help them find and create unique products and solutions. 

Our hope for the next year is that more engineers and creators take advantage of all that CoreWeave has to offer. The future of generative AI requires infrastructure that can keep up with its rapid growth and developers who will continue to innovate. 

Our flexible and fast infrastructure. Your incredible ideas. Together, we can compute, create, and innovate.

Connect with us

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.