Infrreport: Nvidia close to acquiring AI cloud provider Lepton AI in a nine-figure deal

When we talk about technological advancement in artificial intelligence (AI), we’re witnessing what I’d call a remarkable change in the marketplace. Nvidia is making strategic moves on the AI board with its latest pursuit.

According to recent reports, Nvidia Corp. is in talks to acquire cloud startup Lepton AI Inc. My sources say this deal could be worth several hundred million dollars. This would be a big win for Lepton AI’s investors, CRV and Fusion Fund. They led an $11 million seed round for the company just two years ago.

Lepton AI emerged in 2023, founded by Yangqing Jia and Junjie Bai, former researchers at Meta’s AI lab. These guys weren’t just regular researchers; they helped develop PyTorch, which is crucial in the industry. Lepton AI has a team of about 20 employees and works with several venture-backed startups.

Their cloud platform specializes in AI workloads. It allows customers to both train models and perform inference, which means running a neural network after training. The platform provides an easy-to-use visual interface for setting up training clusters with various Nvidia graphics card options. For teams dealing with multiple models, they offer centralized management of hardware allocation.

During training, AI models sometimes encounter errors that can lower output quality. Lepton AI’s platform includes features to detect such errors and identifies technical issues, like excessive memory usage during training. After development, customers can deploy models on inference-optimized instances in Lepton’s cloud. The platform boasts processing speeds exceeding 600 tokens per second and keeps latency under 10 milliseconds. An auto-scaling feature automatically allocates more hardware resources when demand spikes.

The company speeds up inference using open-source tools like vLLM. It consolidates similar LLM prompts into single requests, reducing processing demands. Developers also use vLLM to reduce their models’ memory footprint.

Nvidia’s plans with Lepton aren’t fully clear yet. They might be after talent or planning to keep the startup’s cloud platform running. This platform competes with products from major Nvidia customers. This includes CoreWeave Inc., which is going public soon.

This news comes right after Nvidia’s reported acquisition of Gretel Labs Inc. This AI startup deal is valued at over $320 million. Gretel creates synthetic data for AI model training, now important for building strong AI systems.

What is the significance of Nvidia’s acquisition of Lepton AI?

Nvidia’s purchase of Lepton AI is significant for the AI and cloud computing industries. First, it’s a strategic move to expand Nvidia’s capabilities in AI cloud services. Lepton AI’s platform fills a niche in handling AI workloads, which Nvidia can leverage. This step shows Nvidia’s commitment to growing its dominance in AI technology.

Moreover, this acquisition could lead to faster advancements in AI services. Developers and end-users may benefit from improved AI model training and inference capabilities. Cloud computing tasks could become more efficient, saving time and resources for businesses and individuals.

Who are Lepton AI’s competitors?

Lepton AI faces competition from various companies in the AI cloud space. CoreWeave Inc. is notable among them. They offer similar cloud-based AI model training and inference options. However, differentiating factors might lie in service offerings, such as scalability and speed. These differences could influence market preference toward Lepton AI or CoreWeave Inc. based on customer needs.

We’re watching the AI landscape reshape itself in real time. The stakes couldn’t be higher, and the competition couldn’t be more intense. These aren’t just business deals; they’re part of transforming computing in the future. Next up, we’ll see how Nvidia integrates Lepton AI into its broader strategy. Stay tuned to see if this decision ignites further market shifts.