From Social Media to AI Compute: The Piotr Tomasik Story

 

Piotr Tomasik, co-founder and COO of TensorWave, was interviewed on Value Drivers. Tomasik discusses his journey into AI, his entrepreneurial background, and the founding of TensorWave. He grew up in the Bay Area, and moved to Las Vegas to attend UNLV for computer science. He was drawn into the startup world after being recruited by a professor into a green energy company. He then worked for a prepaid debit card processor before co-founding a social media startup called Activeside. He also co-founded a company called Influential, a social media influencer marketing platform, which was sold to Publicis for $500 million. Tomasik's entry into AI came when he received a cold email from someone at IBM Watson.

In 2021, Tomasik started a nonprofit called Startup Vegas to connect investors with entrepreneurs. At one of the events, he met Jeff who became one of his co-founders at TensorWave. They began discussing the AI ecosystem, including Nvidia's supply constraints. His co-founders had prior experience deploying FPGAs in the cloud. Together, they recognized a gap in the market for AI compute, especially for startups that were having difficulty sourcing the necessary resources. They decided to partner with AMD, becoming a launch partner on the MI300X in December 2023.

TensorWave aims to make AI compute more accessible by providing an alternative to Nvidia . They chose to work with AMD because of differentiators in their product line, particularly the amount of memory per chip . The AMD MI300X has two times the amount of onboard memory as the Nvidia H100, which allows for better unit economics . For example, a full Llama model can fit on one MI300X, whereas it would take two and a half H100s . The company seeks to provide better total cost of ownership and more options for workloads, particularly those with large memory components such as video, long documents, and long-context workloads . Tomasik notes that while Nvidia has a dominant position and strong software ecosystem, AMD is also a strong competitor and has a history of consistent execution .

TensorWave raised $43 million in safe funding, one of the largest in Las Vegas history . This funding was necessary because the servers they use cost as much as a house, approximately $250,000 or more, each containing 8 data center GPUs . The company is using the funds primarily for hardware .  Tomasik also discussed the importance of the US leading the AI revolution .

 

Key Takeaways

Identify and Capitalize on Market Gaps

A key lesson from Tomasik's story is the importance of identifying gaps in the market and developing solutions to fill those gaps. Tomasik and his co-founders at TensorWave recognized that there were supply constraints and a lack of accessibility in the AI compute space. Rather than becoming "the 30th Nvidia cloud," they chose to partner with AMD to offer an alternative. They saw that AMD's technology, particularly the increased memory capacity of the MI 300X chips, could provide a competitive edge . This approach demonstrates the importance of strategic thinking and not just following the crowd.

Strategic Partnerships and Long-Term Vision Are Essential

Tomasik emphasizes the significance of strategic partnerships, and the importance of a long-term vision, particularly in capital intensive industries. TensorWave's partnership with AMD is central to its strategy, providing a strong differentiation in the market from those using Nvidia GPUs. He also seeks customers who are willing to commit to longer term contracts, which provides the company with more financeable revenue streams. Tomasik's focus on building a team of trusted colleagues also reveals