Earlier this week, The New York Times highlighted the challenges that founders face in securing essential computing power for their emerging artificial intelligence startups. With major corporations and wealthy nations in a fierce competition to acquire these resources, one founder likened graphics processing units (GPUs) to "rare earth metals." According to the article, many founders are resorting to various tactics to obtain these vital chips, including seeking assistance from contacts at large equipment vendors and navigating a little-known U.S. government initiative called Access.
To combat this significant shortage, Index Ventures, a global investment firm, has forged a partnership with Oracle. This collaboration aims to provide GPU access—specifically the highly sought-after Nvidia H100 and A100 chips—to Index's portfolio companies.
To gain further insights into this innovative arrangement—one that other venture firms may likely attempt to duplicate—we spoke with Erin Price-Wright, a Bay Area partner at Index Ventures who specializes in enterprise software and AI. Before joining Index in 2019, she led product development for Palantir's data analytics and machine learning platform. Here are edited excerpts from our conversation, which you can also hear in full here:
Can you explain your partnership with Oracle?
Erin Price-Wright: Access to computing power is one of the most significant hurdles AI startups encounter, especially for those at the early stages. The issue isn't so much the cost but rather that over 95% of GPU capacity is already reserved for larger players, which makes substantial pre-commitments with cloud providers. For startups trying to initiate their training or model fine-tuning, waiting for GPU availability can take anywhere from three months to over a year, complicating their launch efforts.
Early-stage companies often don't even know how many GPUs they may require as they determine their product direction. This unpredictability adds to the challenge. Our partnership with Oracle aims to provide GPU access to our earliest-stage portfolio companies, eliminating this barrier so they can concentrate on their core missions from day one. Ultimately, we want to help these companies scale up to independent clusters. We're not in the business of providing extensive GPU resources long-term; our goal is to give them a leg up for accelerated development.
How did this agreement come about?
We wanted to ensure that innovators targeting tangible business solutions didn’t have to modify their models or fundraising strategies just to obtain GPU access. This idea emerged after repeatedly witnessing early-stage companies struggling with this issue. Index has significant clout in the market, and we can leverage our position and relationships to aggregate demand and deliver valuable services to our founders.
Did Index Ventures make a down payment or purchase chips directly from Oracle? Are you giving Oracle equity in these startups?
We are not purchasing chips outright. Instead, Index makes the precommitment for our startups and covers the cloud expenses. Oracle manages the GPU cluster and has been an exceptional partner, allowing our companies to use this resource at no cost.
Did you discuss this advance payment with your investors? That’s not a conventional approach for a venture firm.
For the specifics of the agreement, I’ll refrain from revealing too many details.
Is this an exclusive partnership? Can other venture firms replicate this model?
Absolutely, other firms can follow suit; there’s no exclusivity involved. Oracle benefits by connecting with emerging successful companies as early as possible through this initiative. Our goal is to help our companies eventually secure their dedicated deals with Oracle, AWS, and other major cloud providers rather than relying on our cluster indefinitely.
One of your portfolio companies, Cohere, has Oracle and Nvidia as backers—two crucial allies for your startups right now.
Connecting our portfolio companies with the right people at the right time is key to ensuring they have the resources they need.
Index has at least 20 AI/ML-focused portfolio companies, including Cohere, which has raised $445 million, and Mistral AI in France, which recently completed a substantial seed round. Is the influx of investment in generative AI excessive, or is it still in the early stages?
We’re still in the early stages. While I anticipate a cooling period concerning large investment rounds from traditional VCs, a significant infrastructure gap remains between the potential and actual utility of core technologies. It will require time to bridge this gap.
Over the next year, while I am optimistic about the transformative potential of AI technology, I expect some firms to reassess their approaches as they determine ROI and prioritize use cases, moving beyond mere prototypes to generate tangible solutions. This will pave the way for the infrastructure necessary to support scaling these use cases.
How do you ensure your AI companies don’t compete directly? Is it more challenging than in traditional startups?
The approach isn’t notably different from how we analyze competition in other sectors. While AI is often treated as a standalone category, I foresee that every piece of software will embed AI as a foundational element in the near future. The core competencies of software—similar to how databases function—will incorporate AI, allowing for unique applications. Although we are still in the initial market phase, there will undoubtedly be shifts as companies identify specific problems to address using these tools.
In summary, while AI is becoming ubiquitous, our strategies for investment and oversight remain grounded in traditional methodologies.
Ranking
