IBM Ventures brings generative AI to the enterprise with $500M fund

Roger Premo, IBM Ventures general manager of corporate strategy and ventures (Joey Schaffer/PitchBook News)
Corporate investors are betting big on generative AI. IBM Ventures is no exception: It launched a $500 million enterprise AI fund last month.

IBM Ventures is led by Roger Premo, who ran Boston Consulting Group's global software practice before joining IBM in 2020. Premo has the hard task of walking the line between fostering strategic partnerships and generating strong financial returns. The secret to success, he says, is walking directly into challenges facing the enterprise space. IBM Ventures plans to invest between $2 million and $5 million in mostly early-stage deals.

IBM has evolved in recent years from a focus on hybrid cloud computing services to AI. The company bought enterprise financial software developer Apptio for $4.6 billion in June.

Premo sat down with PitchBook to discuss enterprise AI investing, what makes IBM Ventures' investing model unique, and how it plans to compete for deals in the crowded generative AI space. This interview has been edited for length and clarity.

PitchBook: What is IBM Ventures' strategy?

Premo: We talk to IBM clients a lot about what they want to get done with AI. They want it to be well governed. How do you govern AI? There are so many unsolved problems along the way, and we're leaning into them. How do you work with the open-source community and make their [large language] models accessible across the enterprise space? There are all sorts of security threats as well.

The venture fund is about seeing those challenges and working with the startup community that's helping build those solutions. We think of the fund as a way to better connect IBM to the early-stage innovation community. We've piloted our venture champion model that we've premised the enterprise AI fund on. We've seen early signs of success with it. Much more to be done to see that through, but we do think we have a best-in-class corporate venture model to bring this fund to life with.

What is the "venture champion model"?

We pair leaders from IBM product teams with our venture investing team to make sure that we have the best ideas in front of the early-stage communities that deeply understand those spaces.

If you want to live in the "Goldilocks zone" that we're looking to live in—investing in companies that are good investments but also ones that are good fits with IBM so that we get commercial success—that requires staying engaged with the early-stage community, but it also requires staying aligned with IBM product strategy.

A ditch that many corporate venture arms find themselves in is some are very financially oriented and make investments that don't relate to the core business. There are other ones that are almost overconnected to the core business. They don't have the flexibility to make good financial decisions. We're trying to solve for that.

How do you think this model has helped you source deals?

One of the challenges of corporate venture capital has been [that CVCs] may not get the chance to invest in the companies they're interested in. We've had a really high batting average on that from the model. I think a lot of it is driven by an early-stage company seeing somebody that really can be an expert and adviser to them as an investor sitting across the table.

The other side of the model is that it has teed up so many commercial collaborations—because from the start we know where that technology could integrate with our portfolio. We don't make it a prerequisite (asking a company for a commercial arrangement before investing) because that's a failure mode we've seen in the research. When the investment is placed, our startups know exactly where we could and should collaborate.

What trends in enterprise AI are you watching?

We see a few patterns. One is using multiple [large language] models. We see clients using open-source models, commercial models and a variety of models they use to bring generative AI to life. We think it's a multi-model world, and that's going to be incredibly important going forward—orchestration of models and model enablement. You want to harvest all that innovation.

We're looking at everything that helps clients get a better return out of generative AI with use-case specificity. The outcome is better model optimization. The cost of running these models is enormously expensive; we're looking to take costs out of that, so that running the technology doesn't cost more than the benefit it creates.

How much importance are you now placing on startup governance, given everything that has happened with OpenAI?

I don't think our standard has changed. We've always placed a heavy emphasis on governance. I do think it was a pretty unique governance structure at OpenAI, and I think it's fair to say that would have probably raised some questions for us.

We're going to stick to our traditional approach of founder track records and trust in governance structures we know and understand and believe in going forward.

IBM Ventures is seeking both strategic and financial returns. How do you plan to balance those priorities?

We've set up strong portfolio management functions that create linkages after the investment. [In] 80% of our portfolio, we've created commercial partnerships and linkages. I can point to some startups that have had a double-digit percentage of revenue growth that comes from working with IBM. We want to use those capabilities. If we can partner together to pair our platform with one of our portfolio companies to better solve what an enterprise needs, that's real commercial success.

This article originally appeared on PitchBook News