Laura Herman (Potentiary), Richard Hong (TNE.AI, ex-Microsoft) and Mark Terrell (unDavos) debate whether trillion-dollar data-centre buildouts are the battleships of our era. Richard reveals how his company shrunk a one-trillion-parameter cloud model to 1.7 billion parameters — running on a $1,000 laptop at 25 watts instead of 20 kilowatts — while delivering the same results for enterprise KYC and anti-money-laundering. Mark argues the real unlock is secondary markets and pre-exit liquidity that could free up hundreds of billions in early-stage capital, enabling four-person startups in Nairobi to build what used to require Silicon Valley budgets. The conversation spans AI sovereignty, critical infrastructure fragility (Berlin lost power for two days from a single substation attack), quantum computing's looming "Q-day" that could expose every encrypted record, and why every country now needs to think about education, energy independence and technology sovereignty as a single stack.
SPEAKERS
• Laura Herman — Founder, Potentiary | 20 years in nuclear science & tech transfer
• Richard Hong — CEO, TNE.AI | Ex-Microsoft (12 years), former VC (15 years)
• Mark Terrell — Founder, unDavos | PhD in Collective Intelligence (Intel)
unDavos — unconventional conversations at Davos. 200+ sessions across 12 tracks, bringing together founders, investors, policymakers and scientists for the discussions that matter.🔗 https://undavos.com📺 https://www.youtube.com/@undavos
TRANSCRIPT
And the stage is yours. All right. Thank you very much. Richard, if you want to take the microphone and you want to sit, I'm going to moderate. So I'll stand. Sure. We've got two chairs. That makes sense. My name is Laura Herman. I'm with Potentiary. This company was founded as an investment advisor after I had 20 years working in nuclear science and technology across the entire fuel cycle, doing tech transfer from our national laboratories in the United States out into the startup world. So in the last decade or so, I've had the opportunity to work with a lot of startup companies who are trying to break into large industrial systems-oriented types of industries. And so that brings me here today and to Davos this week. And I am excited to be having a conversation with Richard Hong, who is former Microsoft. He's going to tell us a little bit about his company and Mark Terrell, who has been organizing these spotlights here on Davos. The one and only. Yeah. So Richard, please, do you want to take a few minutes and tell our audience about what you're working on and how you got started in the AI infrastructure? Well, thank you, everybody. I think that, you know, we have an AI house here, we've got all these crazy things happening, and I guess I'm here to tell you that actually making these things work is way harder than you think. And I often ask for a show of hands, because if you look at the press, it seems like it's just happening everywhere. And I guess I spend a lot of time talking with the very biggest companies, and the answer is it really is very hard. So by way of background, I worked at Microsoft for 12 years. You can blame me for all the issues with Windows Office and our server products, I guess that's one thing. And then I spent 15 years as a venture capitalist, and in the last five years I've been doing AI companies. And we are a company called Total Neural Enterprises, TNE.AI, and we're really trying to help the biggest companies adopt this technology in an efficient way. I love to hear you say that transferring artificial intelligence into these enterprise systems is not easy, because too often I hear it kind of presented as a panacea. So we'll talk a little bit more about how it can be difficult. It's called the San Francisco AI House Party, by the way, if you're wondering what the actual techni
