Dan Williams discusses AI security risks, Delta-v's strategic exit focus, and the enterprise adoption gap.
Summary
Dan Williams, Partner at Delta-v Capital, explains that traditional cybersecurity fails against AI due to its non-deterministic nature, requiring bespoke runtime security solutions. He details Delta-v's investment strategy, which targets Series B/C+ infrastructure software companies with a focus on strategic exits to large tech buyers. Williams highlights a widening gap between AI capability growth and enterprise adoption, driven by trust issues and context window limitations. He also describes Delta-v's operational discipline, including maintaining smaller check sizes and utilizing secondary markets for liquidity within their 10-year fund lifecycle.
Key takeaways
- 1 AI right now is like giving a five-year-old five year old the keys to a car and and letting them drive.
Dan Williams · 0:23
- 2 we've in addition to Correlate, we've done a number of tender offers across our our firm's history. We think that they're they're really important um motivational uh steps for for a company. They they create retention within the employee base. You help these employees who many of whom have been at these companies for seven 10 years, have all of their, you know, net worth on paper sitting in private company shares they can't get liquidity for. It helps them understand that the shares that they own and what they've been building at this company have real value.
Dan Williams · 5:43
Episode sections
- 0:00 – 1:45
AI Security Risks and Non-Determinism
Dan Williams explains that traditional cybersecurity fails against AI because models are non-deterministic and unpredictable. He argues that defenders must adopt AI rapidly to counter attackers who are leveraging the same technology.
- 1:45 – 3:24
Career Background and Product-Centric Investing
Williams discusses his transition from engineering at Cisco to private equity, emphasizing how his technical background allows him to assess product differentiation beyond marketing buzzwords. He believes great, unique products are the primary driver of value in tech.
- 3:24 – 5:39
Strategic Exit Strategy and Corp Dev Alignment
Williams details how Delta-v designs investments with strategic exits in mind, often targeting large tech buyers like Cisco, Palo Alto, and CrowdStrike. They engage with corporate development teams early to understand strategic fit and hierarchy.
- 5:39 – 8:27
Secondary Tender Offers vs. Direct Secondaries
The speaker contrasts employee tender offers, which aid retention and provide liquidity to long-tenured staff but involve complex disclosure processes, with direct secondaries, which are faster and more bespoke but lack broad impact.
- 8:27 – 10:22
Primary vs. Secondary Investment Agnosticism
Delta-v treats primary and secondary investments interchangeably based on what the company needs. They may lead a primary round while buying secondary shares, or start with a secondary check to enter a company that doesn't need immediate growth capital.
- 10:22 – 20:15
Hidden Layer and AI Attack Surfaces
Williams highlights Hidden Layer's focus on securing autonomous coding agents. He notes that traditional EDRs fail because AI agents access system resources differently than humans, requiring bespoke runtime security solutions.
- 20:15 – 24:07
AI Adoption Gap and Trust Issues
Williams identifies a widening gap between AI capability growth and enterprise adoption, driven by trust issues. He cites context window limitations, the need for proper instrumentation, and user education as key barriers to adoption.
- 24:07 – 30:29
Developer vs. Security Friction and Board Roles
The friction between developer velocity and security compliance remains constant but intensifies with AI-generated code. Williams explains that his role as a board observer or member focuses on influence and support rather than voting, aligned with a 'flight plan' for each portfolio company.
- 30:29 – 32:40
Fund Lifecycle and Secondary Liquidity
Dan Williams explains that Delta-v Capital operates on a 10-year fund lifecycle, which may necessitate selling positions before a full exit to provide liquidity to other shareholders. He notes that the firm acts as both a buyer and seller in the secondary market to manage this lifecycle constraint.
- 32:40 – 34:59
Public Market Signals and AI Capex
Williams states that Delta-v tracks public markets to gauge AI Capex trends from hyperscalers like Google and Amazon, which influence private market pricing and deal flow. He highlights a disconnect between public company growth rates (under 25%) and Delta-v's portfolio average entry growth (145-150%).
- 34:59 – 36:23
Portfolio Marking and Volatility
The firm marks investments quarterly using public comps as a key input, though they focus primarily on underlying business KPIs rather than daily public volatility. Williams contrasts Delta-v's portfolio stability with the higher volatility experienced by hedge fund peers.
- 36:23 – 38:49
Investment Discipline and Check Sizes
Despite managing over $1.6 billion, Delta-v maintains a disciplined check size range of $5-50 million to avoid strategy drift and maintain a broad investable universe. They utilize strategic LPs for larger co-investments when opportunities arise, citing Arctic Wolf and Vero as examples.
- 38:49 – 42:35
Talent Development and Culture
Williams describes a long-term retention culture at Delta-v, contrasting it with the 'up-or-out' model of other firms. Associates are immersed in sector-specific pods (e.g., cybersecurity) and participate in democratic IC voting processes to develop deal underwriting skills early in their careers.
- 42:35 – 47:24
AI Security and Governance Beliefs
Williams rejects doomsday narratives about AI taking over humanity, arguing instead that the immediate risk is a lack of security and governance. He uses the analogy of giving a five-year-old the keys to a car to describe the current state of AI implementation without constraints.
Claims worth verifying
Delta-v Capital has over $1.6 billion in assets under management.
Unidentified speaker (speaker_0) ·0:00
How to verify: Company website or press release for Delta-v Capital; SEC Form ADV for assets under management. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Primary sourceDan Williams began his career in engineering and product roles at Cisco.
Unidentified speaker (speaker_0) ·1:45
How to verify: Dan Williams' LinkedIn profile; Cisco press releases or employee announcements from his tenure. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Strategic exits account for about half of Delta-v's exits.
Dan Williams ·3:41
How to verify: Delta-v Capital portfolio page or press releases detailing exit history; comparison of strategic vs. financial exits. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v led a $42 million all-secondary employee tender offer for Core Lite.
Unidentified speaker (speaker_2) ·5:39
How to verify: Press release from Delta-v Capital or Core Lite announcing the transaction; SEC Form D if applicable. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Primary sourceDelta-v led a $100 million Series B for Hidden Layer.
Unidentified speaker (speaker_2) ·10:22
How to verify: Press release from Delta-v Capital or Hidden Layer announcing the Series B round. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
AI models are non-deterministic.
Dan Williams ·0:06
How to verify: Technical documentation from AI model providers (e.g., OpenAI, Anthropic) explaining model behavior and temperature settings. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Open source components make up 80-85% of all code written.
Unidentified speaker (UNKNOWN) ·19:14
How to verify: Industry reports from Synopsys, Sonatype, or GitHub on open-source usage in software development. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Dan Williams serves as a board member for Log Rocket and Hidden Layer.
Unidentified speaker (UNKNOWN) ·24:07
How to verify: Company press releases announcing board appointments; LinkedIn profiles of Dan Williams and the companies. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Dan Williams holds board observer roles at Clarity and RapidFort.
Unidentified speaker (UNKNOWN) ·24:07
How to verify: Company press releases announcing board observer appointments; LinkedIn profiles. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
PHP had a massive vulnerability last week.
Dan Williams ·19:13
How to verify: CVE database or security advisory from the PHP Foundation or relevant security researchers for the week prior to the podcast release. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v Capital manages over $1.6 billion in assets under management.
Unidentified speaker (speaker_2) ·36:21
How to verify: Check Delta-v Capital's official website or press releases for AUM figures; verify against SEC Form ADV if publicly filed. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Primary sourceDelta-v operates a 10-year fund lifecycle.
Unidentified speaker (UNKNOWN) ·31:13
How to verify: Review Delta-v Capital's fund terms or investor presentations for fund duration specifications. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Public companies in the AI sector are growing higher than 25% (or under 25% as stated in context of comparison difficulty).
Dan Williams ·33:06
How to verify: Analyze earnings reports and growth metrics of major public AI-related companies (e.g., from S&P 500 or Nasdaq) for the relevant period. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v's average entry point in their current fund is 145-150% growth.
Unidentified speaker (UNKNOWN) ·33:14
How to verify: Request internal fund metrics from Delta-v Capital or verify against portfolio company growth rates if disclosed. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v invested $6 million initially in Arctic Wolf and over $80 million total.
Dan Williams ·37:12
How to verify: Check Crunchbase or PitchBook for Arctic Wolf funding rounds; verify initial check size and total investment from Delta-v. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v completed a multi-hundred million dollar investment in Vero.
Dan Williams ·38:04
How to verify: Search for press releases or news articles regarding Delta-v Capital's investment in Vero and the deal size. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v has board seats in approximately 80% of its deals.
Dan Williams ·46:14
How to verify: Review Delta-v Capital's portfolio page and cross-reference with company press releases or SEC filings (if public) for board composition. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Primary sourceOpenAI agent was given a task and executed it creatively without nefarious plans.
Unidentified speaker (UNKNOWN) ·43:09
How to verify: Review public reports or statements from OpenAI regarding the specific incident referenced (likely the 'hugging face' or similar agent jailbreak incident) to confirm the nature of the event. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v invests in Series B or Series C and beyond.
Unidentified speaker (UNKNOWN) ·45:32
How to verify: Review Delta-v Capital's investment thesis or portfolio company stages to confirm typical entry rounds. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Delta-v invests broadly across the United States, with opportunistic investments in Europe and Israel.
Dan Williams ·45:27
How to verify: Analyze Delta-v Capital's portfolio geography and press releases for investments in European and Israeli companies. For recent events, verify 2026-08-17 through 2026-11-17. Use another period only when the speaker explicitly names it.
Do your own research
- Verify Delta-v Capital's fund size and vintage to assess their capacity for Series B/C+ investments.
- Review public market growth rates for hyperscalers (Google, Amazon) and chipmakers (Nvidia) to validate claims about AI Capex trends.
- Analyze Hidden Layer's technology stack to determine if their runtime security solutions address non-deterministic AI agent behaviors.
- Check Delta-v's portfolio company exit history to confirm the percentage of strategic exits versus IPOs or other M&A.
- Assess the context window limitations of current LLMs by reviewing technical documentation from major AI providers.
- Evaluate the effectiveness of employee tender offers as retention tools by reviewing industry benchmarks for turnover rates in tech startups.
- Compare Delta-v's portfolio entry growth rates (145-150%) against public comps to understand the valuation disconnect.
- Review Delta-v's investment thesis for infrastructure software to identify potential conflicts of interest with strategic LPs like Cisco or Palo Alto.
Full transcriptexpand
- 0:00
- Unidentified speaker (speaker_0): Where do traditional cybersecurity firewalls completely fail when it comes to defending enterprise AI models?
- 0:06
- Dan Williams: Because these models are non-deterministic. No one knows what they're gonna do, you know, until they do it.
- 0:11
- Unidentified speaker (speaker_0): The more we trust it, I think the more uh the risk increases.
- 0:15
- Dan Williams: Defenders need AI and they need it fast because the attackers are going to be leveraging it.
- 0:21
- Unidentified speaker (speaker_0): What is one non-consensus non consensus belief you hold?
- 0:23
- Dan Williams: AI right now is like giving a five-year-old five year old the keys to a car and and letting them drive.
- 0:28
- Unidentified speaker (UNKNOWN): Hey
- 0:28
- Unidentified speaker (speaker_0): everyone, this is Prashantra.
- 0:45
- Unidentified speaker (speaker_0): Secondary investments. Dan leads the firm's infrastructure software practice. In this episode, we talk about why. AI is the hardest attack surface yet to secure,
- 0:54
- Unidentified speaker (UNKNOWN): how
- 0:54
- Unidentified speaker (speaker_0): corporate buyers shape his exit strategy, and why he treats primary and secondary deals the same. So without wasting any time, let's dive straight in. This episode is brought to you by Podcast Enix, the team behind the show you're listening to.
- 1:08
- Unidentified speaker (UNKNOWN): As
- 1:08
- Unidentified speaker (speaker_0): an investor, you're already talking with founders, allocators, and operators. We just turned turn those conversations in
- 1:14
- Unidentified speaker (UNKNOWN): the world.
- 1:15
- Unidentified speaker (speaker_0): into a branded podcast. And each recording becomes a YouTube video, an audio episode, short clips, LinkedIn posts, and a newsletter, activating all your marketing channels at once. This keeps you top of mind with LPs and founders without taking away much time from your week. You can book a free strategy call with me at podcasttenx.com. I'll add the link
- 1:35
- Unidentified speaker (speaker_0): in the show notes below. Now enjoy the episode.
- 1:38
- Unidentified speaker (UNKNOWN): Hey
- 1:38
- Unidentified speaker (speaker_0): Ryan Dan, Sokar so good to have you on the VC TenX podcast. How are you doing?
- 1:42
- Dan Williams: Doing well. Thanks, Prasak. Great to be here. Thank you for having me.
- 1:45
- Unidentified speaker (speaker_0): Absolutely. It's a pleasure having you on. Uh and to start things off, let's talk
- 1:48
- Unidentified speaker (speaker_2): about how you started your career in engineering and product roles at Cisco before transitioning to MA and eventually private project equity. How does the engineer engineering in you alter how you assess technical risk in a startup architecture compared to a traditional financially trained investor?
- 2:05
- Unidentified speaker (UNKNOWN): Yeah,
- 2:05
- Dan Williams: it definitely frames how I think about uh investing and engaging with companies. I you know, I'd like to think that it helps me to ask some of the right questions uh as we're you know engaging with with these companies.
- 2:19
- Unidentified speaker (UNKNOWN): All
- 2:19
- Dan Williams: of what we do is technology, and so it allows me to go a level deeper with with these entrepreneurs and to really try to understand how the product works and why that creates some kind of differentiation between others in the market. I think, especially in cyber, which is a big area that I spend time, um, you can go to these black hat and RSA conferences
- 2:39
- Dan Williams: every year. And every company sounds the same because marketing, you know, has has found all attached itself to all the same buzzwords and and really promoted those. But you got gotta to get under the hover uh under the cover. covers to really understand what's going on and and so we try to do that for sure. Um I'm I'm uh I I think what it has done is it's
- 2:59
- Dan Williams: driven me to be a product guy at heart. I I have this innate belief that if you build a great product that is unique and and it's built to create real value for your customers, that But hopefully you can solve the rest. In some cases that's that's naive, like go to market uh scaling a business is has as 0.8 challenges as well that we've learned along
- 3:19
- Dan Williams: the way. But um, but you know, product I think is is first and foremost here in tech.
- 3:24
- Unidentified speaker (speaker_2): Yeah, absolutely. And given your time at Cisco's Cooperate Development Group, uh, where the core strategy was heavily powered by acquisitions, how does your deep understanding of how tech giant giants spy buy companies influence the exit strategies you map out for Delta V? portfolio companies today?
- 3:40
- Unidentified speaker (UNKNOWN): Yeah,
- 3:41
- Dan Williams: it's a good question. Your background definitely informs how you think about the world.
- 3:47
- Dan Williams: Strategic exits or exits to large uh tech buyers have been about half of our firm's exits over time. And I think Um, part of that is by design. We intentionally, as we go through our underwriting process of thinking about investing in a company, um, one of the key factors that we we take into account is is this
- 4:07
- Dan Williams: product and is this category strategic to the the folks downstream from us uh from a strategic perspective?
- 4:16
- Unidentified speaker (UNKNOWN): Uh
- 4:16
- Dan Williams: sometimes it winds up being a PE buyer, and that's fantastic as well. They, you know, they have have um gotten more active, they will pay prices on par with strategics, but having a strategic
- 4:27
- Dan Williams: exit path just sets the company up in a different, different manner than I think others do. And so as we go through our underwriting, one of our key steps is to call on those corp dev teams from each of the big, if it's cyber, we're calling Cisco and Palo Alto and CrowdStrike and Central One. We want to understand do they know the company?
- 4:46
- Unidentified speaker (UNKNOWN): You
- 4:46
- Dan Williams: know, what did they think of it? Where does this category fit on their strategic hierarchy? of importance of where they want to go add product in the future? And, you know, it's almost putting that corp dev hat back on as you think about investing in the same way that that the corp dev teams think about buying. Like what is the what are the most here at
- 5:06
- Dan Williams: Delta V within cyber, we have
- 5:08
- Unidentified speaker (UNKNOWN): a
- 5:09
- Dan Williams: view of right you now on and what are the next three categories that we want to make investments and bets in um because of the strategic. importance downstream and so huge input on our underwriting.
- 5:19
- Unidentified speaker (speaker_2): Absolutely that's amazing. And Delta V also recently led a 42 million dollar all secondary employee tender offer for Core Lite after the CEO explicitly called you for liquidity solutions. What are the unique operational and psychological complexities of leading a tender offer for tenured employees? versus
- 5:39
- Unidentified speaker (speaker_2): doing a traditional secondary purchase from early VCs?
- 5:43
- Unidentified speaker (UNKNOWN): Yeah,
- 5:43
- Dan Williams: we've in addition to Correlate, we've done a number of tender offers across our our firm's history. We think that they're they're really important um motivational uh steps for for a company. They they create retention within the employee base. You help these employees who many of whom have been at these companies for seven 10
- 6:03
- Dan Williams: years, have all of their,
- 6:05
- Unidentified speaker (UNKNOWN): you
- 6:05
- Dan Williams: know, net worth on paper sitting in private company shares they can't get liquidity for. It helps them understand that the shares that they own and what they've been building at this company have real value. Uh, even if that means they're only taking 10 or 20% home, they get to take some amount of cash home and put the kids through college or or put the addition on the house or
- 6:25
- Dan Williams: whatever is important there. Um, and so we view it as a a very strategic process for a company to go through. We like being helpful in that way. It is to your question, it is a little more challenging than a direct secondary purchase. The company does have to go through a process to prepare disclosure statements to all
- 6:45
- Dan Williams: the potential sellers and make sure that they are just as informed uh about the sale process as we. are on the buy side. There's a period of time. It's usually done on some kind of a technology platform where there's an opt-in period, call it 45 days, where they have a chance to decide how much they'd like to sell. And as the buyer,
- 7:05
- Dan Williams: we don't know until the end of that process exactly how much uh supply there'll they'll be of shares. So there's some uncertainty. Uh, it's a little more arm's length of a turn. transaction, but but I think is
- 7:17
- Unidentified speaker (UNKNOWN): is
- 7:17
- Dan Williams: worthwhile to to back to the motivating point to to take a step for the company to help them to uh to
- 7:24
- Unidentified speaker (UNKNOWN): you know
- 7:25
- Dan Williams: provide this value to their employee base. On the one-off direct secondary purchase side,
- 7:30
- Unidentified speaker (UNKNOWN): you know,
- 7:30
- Dan Williams: things are easier. They're faster. Those tend to be one-off One off negotiated purchases from a uh typically a larger shareholder, shoulder, an early angel um, maybe a departed CEO or a a large exec. And in those cases, structure can be bespoke, price or structure. You can do things like create
- 7:50
- Dan Williams: earnouts or um or revenue sharing down the line or royalty sharing down the line. Um, there's no formal process to run through. So it it can be a little more custom um and and quicker, but but again.
- 8:03
- Unidentified speaker (UNKNOWN): you
- 8:03
- Dan Williams: you don't get the same breadth of uh impact that you get with an employee tender offer.
- 8:08
- Unidentified speaker (speaker_2): Absolutely. That's interesting. And Delta V is known for utilizing secondary transactions to back rapidly scaling tech companies. In a market where high growth companies are staying private longer, how do you balance using secondary for cap-table cap table cleanup versus injecting primary growth capital into a company's balance sheet?
- 8:27
- Unidentified speaker (UNKNOWN): Yeah,
- 8:28
- Dan Williams: we at the end of the day, we are looking to back great entrepreneurs and companies in the sectors that we focus on. We're heavily sector focused. I lead our infrastructure software practice. We have a focus around cloud services businesses, and then a a third in and around more vertical AI applications. And so
- 8:48
- Dan Williams: we've kind of narrowed the universe to
- 8:50
- Unidentified speaker (UNKNOWN): do that.
- 8:51
- Dan Williams: companies in the sectors that we know well. On the flip side, the fact that we can we can purchase shares on a secondary basis or invest on a primary basis allows us to access some of these companies at different points in their journey. Not every company is out there looking to raise around today. They may have plenty of cash in the balance Bow sheet. she They might be EBITDA
- 9:11
- Dan Williams: positive and never need to raise capital again. And so in those cases, secondary is the the only way to access the captive and become an an investor. Um, and and in other cases, there are no secondary sellers, but the company really needs cash to go take the next steps around growth.
- 9:27
- Unidentified speaker (UNKNOWN): And
- 9:28
- Dan Williams: so, really, I think it's more about what does the company need at what stage of their growth and and journey, and how can we be most helpful to that? Um, we today are are you know fairly agnostic about whether the investment that we make is on a primary or on a secondary basis. Many of our
- 9:47
- Dan Williams: transactions have a combination of the two. We might lead a primary round and purchase some secondary shares at the same time. We might start small with a small secondary check and then subsequently lead or participate in the company's next primary financing round.
- 10:03
- Unidentified speaker (UNKNOWN): Um,
- 10:03
- Dan Williams: at the end of the day, it's the same company, same entrepreneur that we're looking to back. Uh, and and the method by which we wind up owning shares is a little less important to us. Um, it's really about the back to the company, the entrepreneur, the opportunity in front of them, and and the sector that that we know well.
- 10:22
- Unidentified speaker (speaker_2): Absolutely. Yeah, I love that. I uh the way you treat uh secondaries and primary, not in almost the same way because you're getting
- 10:28
- Unidentified speaker (UNKNOWN): the
- 10:29
- Unidentified speaker (speaker_2): same equity for both those deals. And I think uh a lot of VCs don't think of them the same way. And uh Delta V recently led a massive hundred million dollar series B for hidden layer, focusing heavily on securing autonomous coding agents and LLM applications. From an
- 10:48
- Unidentified speaker (speaker_2): infrastructure standpoint, where do traditional cybersecurity firewalls completely fail when it comes to defending enterprise AI models from adverse serial prompts or data poisoning?
- 11:00
- Dan Williams: Yeah, we super excited to have uh partnered with Hidden Lair and Led Their Series B. Um, Tito and the team that are building something special.
- 11:09
- Dan Williams: Um I'd say that the the behavior and risk around AI are completely different. than traditional cyber.
- 11:17
- Unidentified speaker (UNKNOWN): Um,
- 11:18
- Dan Williams: and and this is not necessarily new to AI. Anytime a new technology emerges, um, whether that was cloud or SaaS or the Edge or endpoints, very large security companies are born to go and protect
- 11:32
- Unidentified speaker (UNKNOWN): them.
- 11:33
- Dan Williams: There is a need for new custom built solutions
- 11:37
- Unidentified speaker (UNKNOWN): in
- 11:37
- Dan Williams: order to protect those. um those new attack surfaces and the nuances of them because the the behavior, the way that attackers uh think about about thwarting them changed with every new technology. And so you had Wiz with Cloud, you had, you know, Cloudflare and Zscaler in in SaaS and Edge, you
- 11:57
- Dan Williams: had CrowdStrike and Endpoint. And so,
- 12:00
- Unidentified speaker (UNKNOWN): you know,
- 12:00
- Dan Williams: we think that hidden layer could be that next. very large company built to secure AI. Um, AI in particular is unique. We think it's the largest and potentially one of the most difficult attack surfaces that we have seen to try to protect because these models are non-deterministic. No one knows what they're going to do, you know,
- 12:20
- Dan Williams: until they do it. Um, the actions that they're taking are at a scale and a complexity that we have never seen before.
- 12:29
- Unidentified speaker (UNKNOWN): And
- 12:29
- Dan Williams: they're connected into data and tools,
- 12:33
- Unidentified speaker (UNKNOWN): you
- 12:33
- Dan Williams: know, that that we need to give them access to in order to to be as impact impactful as they possibly can. Um, and so so we've had a thesis for a long time that we wanted to make a bet in AI. Um, back to your question about how it's different. You know, what one good example of this. that that Tito would explain to you is that if you
- 12:53
- Dan Williams: had a, let's say you had a coding agent on your your endpoint, you've got an endpoint uh product from CrowdStrike or whoever running there, watching what's going on. Um, the behavior that that that endpoint agent would experience is very different uh when they're monitoring a coding agent than they would have just monitoring a human. And
- 13:13
- Dan Williams: by that I mean that. agent's gonna start to access system resources and um all sorts of of components of of the desktop that a normal normal human user would not use day to day. And um that's going to throw alerts at alarm bells within the EDR. Um and so what what is good and what
- 13:33
- Dan Williams: is bad behavior can be very different based on the context. Um and so I think it really takes
- 13:39
- Dan Williams: something that's bespoke and custom built for these AI models and agents to be looking for things like prompt injection attacks. You know, the the system prompt that you give to to an agent or a model, that that agent treats the same as it does the text that's on a website that it's scanning to try to,
- 13:59
- Dan Williams: you know, take in more context to provide an answer back. And uh and it can be thwarted by things like prompt injection attacks. And so um new attacks, new surfaces require new techniques to protect them. And and um you know, we think Hidden Layer is out ahead of that of that curve.
- 14:16
- Unidentified speaker (speaker_2): Absolutely. That's quite interesting. And I do believe personally, what I see is cybersecurity Security might be facing its its biggest moment yet, or the biggest challenge that is it is facing.
- 14:27
- Unidentified speaker (speaker_2): in in the history of the internet, I guess. Uh, because what's happening is we are g giving it all the data, and like you said, we are also giving it access. And what happens because of human behavior is that when you see that an agent is good at something, you sometimes uh lose that oversight or you see, okay, it's doing its job, let me do something else,
- 14:47
- Unidentified speaker (speaker_2): right? So you're not gonna be looking at it uh constantly or what are uh what is it accessing and so on. And I think that's where things can go wrong. Uh the more we trust it, I think the more uh the risk increases uh these cyber attacks happening and cyber risks. Uh uh if you look at cybersecurity holistically, what are the different kinds of
- 15:07
- Unidentified speaker (speaker_2): solutions are you uh looking at or want to look at and that you think can be valuable in in current times? Do you think the big players are already obviously they're already uh trying to bring up new solutions to uh counter that? What are the startups uh that you're looking at trying to do to get ahead of
- 15:27
- Unidentified speaker (speaker_2): them?
- 15:27
- Dan Williams: Yeah, we even in my Cisco days, I I think very large companies move a little slower than than the startups do, and the innovation in large companies is not not quite as strong. So I
- 15:38
- Unidentified speaker (UNKNOWN): we
- 15:38
- Dan Williams: continue to be excited about new um again, bespoke solutions to to protect against the new attack surface from some of the startups. Um I think within AI,
- 15:50
- Dan Williams: there there are likely to be a number of subsectors around securing AI. We think that um the runtime security, which is where hidden layer sits, is is kind of at the core of it because the model is central to every type of perm permutation of that, whether it's an agent or whatever next year brings. Um, and so that's why we
- 16:09
- Dan Williams: started there.
- 16:10
- Unidentified speaker (UNKNOWN): Um,
- 16:11
- Dan Williams: but I think identity has become even more important in the world of AI, not not just human identity, but what is the agent allowed to do? How does that identity inherit from the human? What do they then have access to? Which kind of dovetails into data security. So maybe. three different prongs, you know, around AI, which is more
- 16:31
- Dan Williams: runtime detection response, um, actual identity in and around the agent, and then and then data, um, probably third there. Um, but completely agree with you. I
- 16:41
- Unidentified speaker (UNKNOWN): think, you know,
- 16:42
- Dan Williams: there was a moment in time where where mythos Those came out and everyone, you know, the public cyber socks tanked and people thought that cyber was gonna going to be I guess, bycoated. I'm I'm not quite sure what the logic was, but um
- 16:55
- Unidentified speaker (UNKNOWN): we
- 16:55
- Dan Williams: we were screaming from the rooftops that cyber was going to only become more important. Uh, whether that was to to triage all these these mythos class models can go find vulnerabilities at rapid speed and find ways to exploit them. Uh, we have an investment, a business called Nucleus, that helps to triage those vulnerabilities and and make sure that. that they're
- 17:15
- Dan Williams: getting patched. We have another investment, a business called RapidFort, that helps to harden open source software, which also has a number of vulnerabilities that continue to be found at at uh breakneck speed. So I think cyber broadly is now, you know, only more important than it was given what's gone on with AI. Um, and and you're not gonna going to vibe
- 17:35
- Dan Williams: code a solution to that. Uh it takes super technical, often machine learning-based techniques to go and and solve those problems.
- 17:44
- Unidentified speaker (speaker_2): Absolutely. And and the instances like some of them have already come out.
- 17:49
- Unidentified speaker (speaker_2): I think recently there was some vulnerability in JavaScript that came out. I think that was because of AI. And people have been using uh JavaScript and you know WordPress uh for websites for for years now. And they're they in didn't know that uh these vulnerabilities uh exist, but now they are coming out. So I'm just trying to
- 18:09
- Unidentified speaker (speaker_2): imagine the kind of scale. And when you know the technologies like quantum are coming up next to when you m multiply that AI power with quantum, I know that of course they can be deployed on both sides, on the defense side as well as the attack side. Uh but that will be a crazy world to live in. And I think that might be the biggest crisis
- 18:29
- Unidentified speaker (speaker_2): uh facing the internet economy. Uh these attacks can really rapidly scale uh without people being able to comprehend what's happening at that scale or try to build those defenses if defense does not stay ahead of the curve, right?
- 18:44
- Dan Williams: Yeah, we defenders need AI and they need it fast because the attackers are going to be leveraging it uh you know, very quickly, even if
- 18:53
- Unidentified speaker (UNKNOWN): you know
- 18:53
- Dan Williams: these open source models or open weight models are now just as powerful as many of the foundational lab models and completely open and available to the attackers. Um but couldn't agree with you more. I th I think the PHP was the other one that I think we saw a massive vulnerability in, which underpins most of the the internet uh last
- 19:13
- Dan Williams: week. It's I
- 19:14
- Unidentified speaker (UNKNOWN): think
- 19:14
- Dan Williams: just the tip of the iceberg of what we're going to continue to hear about. Open source components are something like 80, 85% of all of the code that's written out there. So it's not just that the code that Persat
- 19:27
- Unidentified speaker (UNKNOWN): you know
- 19:27
- Dan Williams: auto-generated from a cogen tool that has a vulnerability in it that now needs to be protected. It's that that PHP component or whatever the other open source packages happens to be in every application, you know, under the sun. And so the this again, the scale and breadth, as you pointed out, and and up our need to go protect
- 19:47
- Dan Williams: them means we need AI to go find the vulnerabilities as well and patch them uh ahead of the the attackers.
- 19:55
- Unidentified speaker (speaker_2): Absolutely. Yeah. Crazy times. And right now the market is heavily focused on the application layer of AI and high-level cyber cybersecurity. protocols. However, you've spent your entire career uh looking at the actual plumbing of technology. From your vantage point at Delta V, what is the most critical architectural, engineering, or
- 20:15
- Unidentified speaker (speaker_2): computing bottleneck in today's modern enterprise stack that developers are hitting every day, but public markets and late stage investors are completely underappreciating or mispricing?
- 20:25
- Unidentified speaker (UNKNOWN): Yeah,
- 20:25
- Dan Williams: there's a lot there uh in that question. In fact,
- 20:28
- Unidentified speaker (UNKNOWN): so
- 20:28
- Dan Williams: I I
- 20:29
- Unidentified speaker (UNKNOWN): think
- 20:29
- Dan Williams: um we just had our annual meeting two weeks ago and and um talked about kind of our views on what we call the adoption gap. So um I would actually go maybe one step up for you and then I'll double click down, which is um
- 20:44
- Unidentified speaker (UNKNOWN): I
- 20:44
- Dan Williams: think the rate at which AI is is growing and and expanding in terms of its capabilities is just exponential. It's off the curve off the chart. Um the adoption within enterprises is not keeping pace. You know, I'd say it's here if if adoption's here. And so that there's this wedge between the two that is I think growing
- 21:04
- Dan Williams: every day. Um that that gap I think is created by uh concern around trust. And trust breaks down in a couple of ways, but I think effectively what we need to do is is to secure AI. We talked about hidden layer, we've we've hammered on that. Two, you have to instrument it or equip it. You got to provide all
- 21:24
- Dan Williams: the the rails and the tools and components that it needs to be able to do its job accurately at speed with the right context um to be able to get to the correct answer. Because going down this path and not trusting the output from it is just gonna you know lead to a lack of adoption as well. And then finally, many people today just don't even
- 21:44
- Dan Williams: know how to use AI. And so you've got to walk it into the building. You got to, you know, kind of change the way that work is done, either via services, which is, you know, now a
- 21:54
- Unidentified speaker (UNKNOWN): it's
- 21:54
- Dan Williams: an area we've s we've invested in from our very early days. Um, but today is now, you know, much more uh of a flashy word
- 22:03
- Dan Williams: for deployed engineers and what have you to help help people rethink work. Um and so you know I I think that is the biggest high-level gap within infrastructure, which maybe is the instrument part of that view. Um
- 22:17
- Unidentified speaker (UNKNOWN): I
- 22:18
- Dan Williams: think one of the biggest areas that we're seeing constraints around is is context and memory. Um
- 22:24
- Unidentified speaker (UNKNOWN): one
- 22:25
- Dan Williams: of the exciting things about agents is that they can synthesize decisions from large swaths of data. More than you and you or I could keep in our heads. And
- 22:36
- Unidentified speaker (UNKNOWN): you
- 22:36
- Dan Williams: know, the problem though still is that that context window for an LLM or sitting in Arnest is much smaller than I think people would like it to be. Um, Encoding in coding today, for example, you can't fit an entire code base into the context. window for an LLM. And so it doesn't understand that the component that you're building might be connected
- 22:56
- Dan Williams: to 40 other components downstream and what they require. And and um and so
- 23:01
- Unidentified speaker (UNKNOWN): I
- 23:01
- Dan Williams: think that limitation is both hardware driven based on what actually sits inside the GPU is hardware uh memory that can actually be accessed, but then uh much of it is software-based. And so I think context engineering. it's a whole category that's emerged, um, is going to get super important. And how you ensure that the model has the right
- 23:21
- Dan Williams: data to make the right decisions at the right time. Um, you know, right now it's probably one of the most challenging components, and and that leads maybe downstream to things like evals, uh, ensuring that the output you're getting from it is correct. But
- 23:35
- Unidentified speaker (UNKNOWN): but the
- 23:35
- Dan Williams: input's got to be good and it has to be wide enough and knowledge. knowledgeable enough to be able to provide the right answers back. So um
- 23:43
- Unidentified speaker (UNKNOWN): yeah,
- 23:43
- Dan Williams: that's that's probably what I point to.
- 23:45
- Unidentified speaker (speaker_2): Absolutely quite interesting. And with your boats and investments spanning companies like Log Rocket, Hidden Layer, Cloud Bees, and Rapid Fort, you sit at the epicenter of developer tools and security operations. What is the biggest friction point right now between giving developers the velocity to build with AI and the security team's requirement to maintain zero
- 24:05
- Unidentified speaker (speaker_2): trust compliance?
- 24:07
- Unidentified speaker (UNKNOWN): Yeah,
- 24:07
- Dan Williams: I think you you touched on this with your prior points around vulnerability, but um the friction is candidly, it's the same that it's always been. Developers want to create code, they're not worried about
- 24:18
- Unidentified speaker (UNKNOWN): or
- 24:19
- Dan Williams: that the most important thing is shipping a feature that works the way that they want it to, and whether that has a security vulnerability and it in has always been a separate team that was almost like a
- 24:29
- Unidentified speaker (UNKNOWN): you know
- 24:30
- Dan Williams: big brother telling you you had to do something that you didn't have time or or the bandwidth to do. The same thing applies here, but the magnitude is just different. The amount of code that's being developed and shipped is higher than it's ever been. And it it it's only going to get um, you know, even more ballooned from here. The speed at which those mythos class models
- 24:50
- Dan Williams: can go find. these vulnerabilities in code is growing. And so you just wind up with um a whole different level problem. Um
- 24:59
- Dan Williams: maybe the the other friction point though is is is the adoption of these coding agents. And I this goes back to the adoption gap. We are seeing many many enterprises, typically the very large Fortune 500s that have been quite apprehensive to adopt uh AI coding agents and and maybe start with Copilot and
- 25:19
- Dan Williams: then eventually adopt and and green light cloud code and and so um have kind of lagged behind the adoption of the the best new coding agents that's out there. Um and so again, it's not trusting that the coding agent is secure enough to actually put your code base into
- 25:39
- Dan Williams: to point your point it at your crown jewels. Um and so you need solutions like hidden layer to secure those coding agents, uh, close that gap, you know, effectively. But
- 25:49
- Unidentified speaker (UNKNOWN): um
- 25:49
- Dan Williams: same friction, bigger issues, uh cyber tools needed as always.
- 25:55
- Unidentified speaker (speaker_2): Absolutely. And uh you serve as a board member for companies like Lock Rocket and Hidden Layer, Lear, but take on board observations. observer roles at others like uh Clarity and Traffit Fort. How do you adjust your influence and relationship with the CEO when you shift from a voting board member to a board observer?
- 26:13
- Unidentified speaker (UNKNOWN): Yeah,
- 26:13
- Dan Williams: that's a good question. One one I uh thought a lot about earlier in my career at I think um where I'd come out is that the two roles are really not that much different. Um ultimately my job is to, and and any board member's job is to help the company and CEO to find ways to
- 26:33
- Dan Williams: build the company into the most successful business that it can be and to build shareholder value for all those are involved. Um, whether you're a board member or a board observer, you were invited into that room, you know, for a reason. You weren't just invited in to listen and be a fly on the wall. Uh Um, and so,
- 26:52
- Unidentified speaker (UNKNOWN): you know,
- 26:52
- Dan Williams: I think the input that you provide, showing up with the highest possible valuable ideas, being prepared, you know, reading board materials, staying apprised of what's going on in the industry, um, seeing other portfolio companies that help inform your views on on where things are headed and and what the right next steps for the
- 27:12
- Dan Williams: business are, um, you know, are important. There are most decisions in a in a company's um or in a board don't come down to a vote. They come down to influence and um a uh a debate amongst a bunch of smart folks around the room as to where things should be headed. And ultimately the CEO takes that
- 27:32
- Dan Williams: input and makes a decision on on where to uh take the next step. And so um, you
- 27:37
- Unidentified speaker (UNKNOWN): know, we're
- 27:38
- Dan Williams: there for support really. We back fantastic entrepreneurs. I've had the pleasure of being um, you
- 27:44
- Unidentified speaker (UNKNOWN): know,
- 27:44
- Dan Williams: getting to work shoulder to shoulder with with some fantastic CEOs and entrepreneurs and just being able to do what I can to support them is really the focus. Um we do at Delta V,
- 27:55
- Unidentified speaker (UNKNOWN): uh
- 27:56
- Dan Williams: we build what we call a Delta V flight plan for every company that we invest in. Uh that that goes back to our Delta V kind of name change of trajectory for a for a rocket.
- 28:07
- Dan Williams: And and those lay out areas that we want to be helpful and add value through our whole period in an investment. Um it it usually we align with the management team on that going in to ensure that it matches Which is what where they think thing think they'd like to go grow, whether that's you know MA or uh expanding
- 28:27
- Dan Williams: in product or go-to-market um success. Um, and so uh that Delta B flight plan is in place in both situations. And and again, we're working behind the scenes tirelessly to help them. I think the only real difference is
- 28:43
- Unidentified speaker (UNKNOWN): it
- 28:43
- Dan Williams: tip typically comes outside of boardroom uh endeavors, which is sitting on comp committees and audit committees and there are certain responsibilities of a board member, obviously, that extend beyond an observer that take more time and and uh and focus, but they're a little more nuanced than I think
- 29:01
- Unidentified speaker (UNKNOWN): you
- 29:01
- Dan Williams: know, the most important input that uh a member in and around the board can provide.
- 29:07
- Unidentified speaker (speaker_2): Absolutely. And you've invested Investing in companies like Cloud CloudBee Bees is for a significant tenure and Arctic Wolf for over six years in the group. growth stage, how does your alignment framework change over the management team as in as investment transitions from a standard four-year horizon into a long-term partnership?
- 29:24
- Dan Williams: Yeah, I don't I don't know that I'd say four years is standard.
- 29:27
- Unidentified speaker (speaker_3): Uh every company is different. And uh and there's quite a range of old periods. We've been And
- 29:33
- Unidentified speaker (UNKNOWN): uh
- 29:34
- Unidentified speaker (speaker_3): you
- 29:34
- Unidentified speaker (UNKNOWN): know,
- 29:34
- Unidentified speaker (speaker_3): we've had a company exit like three months after we invested unintentionally. Uh, and we've had others, as you pointed out, that have that we've held for much longer. And um the last four years here, the liquidity window has been quite tight. Uh everyone's going through the same,
- 29:51
- Unidentified speaker (UNKNOWN): the
- 29:51
- Unidentified speaker (speaker_3): same dynamics um with that constrained liquidity window, whether
- 29:55
- Dan Williams: it was IPOs, it still is to a large extent, but also PE and strategic purchases are down quite a
- 30:01
- Unidentified speaker (UNKNOWN): bit. Um but
- 30:02
- Dan Williams: at again, at the highest level as a board member,
- 30:06
- Unidentified speaker (UNKNOWN): our
- 30:06
- Dan Williams: focus remains unchanged, regardless of whether we're in month six or
- 30:11
- Unidentified speaker (UNKNOWN): or
- 30:11
- Dan Williams: year six of our investment in that company.
- 30:13
- Unidentified speaker (UNKNOWN): Uh,
- 30:14
- Dan Williams: our goal is to maximize shareholder value, find a way to help build that business and support the CEO along that journey. And that that changes along the way, you know, what what the journey looks like in the early days. it may be rapid growth phase,
- 30:29
- Unidentified speaker (UNKNOWN): you
- 30:29
- Dan Williams: know, late, late in the company's uh journey, it might be shifting to profitability and managing rule of 40. Um, it probably means navigating tech cycles and you know this this current AI wave is created. Um, you
- 30:44
- Unidentified speaker (UNKNOWN): know,
- 30:44
- Dan Williams: many areas that we're digging in with our portfolio open and them think about how to reinvent themselves and transform themselves in this new AI world,
- 30:52
- Unidentified speaker (UNKNOWN): uh,
- 30:52
- Dan Williams: ensuring that that they're
- 30:54
- Unidentified speaker (UNKNOWN): in
- 30:54
- Dan Williams: in the flow and that and AI is a tailwind as opposed to something that's going to disrupt their business. Um, but um,
- 31:02
- Unidentified speaker (UNKNOWN): you
- 31:03
- Dan Williams: know, ultimately it it's no different. We're depending on where we are. I'd say the one um area that maybe you're you're pushing on is that private equity funds do have a
- 31:13
- Unidentified speaker (UNKNOWN): a
- 31:13
- Dan Williams: fixed life as well, right? We we have a 10-year fund. And um so at times there are points in a company's journey where um as a you know with just my Delta V hat on, where we are in a fun life may dictate uh a need to sell a position before ultimately the company has a full exit.
- 31:32
- Unidentified speaker (UNKNOWN): Um
- 31:33
- Dan Williams: but luckily that's where this secondary market has emerged to help provide liquidity to even uh other GP shareholders that that have a need for liquidity. Um and we've clearly been a buyer in that regard. We've purchased shares from early investors and companies where they were kind of long in the tooth in their investment uh with them. And um
- 31:53
- Dan Williams: we have been a seller at times where that that made sense for our fund. Um, but two two very different hats, board member versus
- 32:01
- Unidentified speaker (UNKNOWN): you
- 32:01
- Dan Williams: know, birt per versus fund manager. um from an alignment perspective.
- 32:05
- Unidentified speaker (speaker_2): Absolutely. And as someone who's investing into the AI theme, which is kind of hot right now and everyone is excited about
- 32:13
- Unidentified speaker (UNKNOWN): it
- 32:14
- Unidentified speaker (speaker_2): or not so excited about it,
- 32:16
- Unidentified speaker (UNKNOWN): uh
- 32:16
- Unidentified speaker (speaker_2): do you keep track of uh how the public markets are behaving at all or as a signal to what can happen in the private markets? Because private markets obviously earlier stage, uh but obviously the Capex of the big tech companies, uh do they give you a sense of where things are moving? Is it spending accelerating or is it slowing down? Do keep track of that
- 32:36
- Unidentified speaker (speaker_2): and what do you infer from that information if you track that?
- 32:40
- Dan Williams: Yeah, we do keep an eye on the public markets. Um
- 32:44
- Dan Williams: right now
- 32:46
- Unidentified speaker (UNKNOWN): w
- 32:46
- Dan Williams: we keep track of them for a number of reasons. One, you know, the cops were where they're trading uh often has a trickle-down effect on where uh rounds get priced and where secondary track transactions occur. Um right now it's that that comp comparison is really challenging because most of the public companies are not
- 33:06
- Dan Williams: growing higher than 25%. And our average entry point uh in our current fund is 145% growth. So
- 33:14
- Unidentified speaker (UNKNOWN): very
- 33:15
- Dan Williams: you know, not apples to apples when you think about a comp comparison set. We do pay attention to where the public uh competitors are are acquiring businesses back to your earlier point on MA. Um, and and so we track them from that perspective. Uh and then, you know, things like spend on CapEx as
- 33:35
- Dan Williams: it relates to AI is clearly an input to the whole market. Um where how Nvidia's chips are being sold, how uh Google and and Amazon compute is being purchased, all about um
- 33:47
- Unidentified speaker (UNKNOWN): it
- 33:48
- Dan Williams: I wouldn't say it changes where and how we're investing, but it is a barometer for us to understand, you know, it if we have a company that's selling into that market, uh, for instance, uh we have an investment in a business called Bureau. that is a digital infrastructure asset, sells or builds fiber networks that help
- 34:08
- Dan Williams: to connect data centers for the hyperscalers. Obviously what the hyperscalers are building and how much demand there is there trickles down to how many new fiber routes should we either acquire or build. So it depends on the company specifically as to what the direct uh implications are. But
- 34:25
- Unidentified speaker (UNKNOWN): yeah,
- 34:25
- Dan Williams: hard not if we can get data we want the data in the public markets to clearly help provide a good portion of
- 34:31
- Unidentified speaker (UNKNOWN): that.
- 34:31
- Unidentified speaker (speaker_2): Absolutely. And when the obviously the public companies are uh very volatile, especially if they're in the AI thesis, you know, one day AI is the next big thing, the next day it's not such a big thing. So while you're in the private markets and you have a portfolio in the companies you have back, you're obviously holding them at certain marks. At
- 34:51
- Unidentified speaker (speaker_2): what point do you start reevaluating the marks you've put uh for the companies in your portfolio? How often does that exercise happen?
- 34:59
- Dan Williams: Yeah, we um we are an uh institutionally backed fund. And so we we have a very we're SEC registered, uh, we go through audits once a year, et cetera. And um the way that we mark our investments is quarterly and it is to the public. comps and and in some cases a mix of private comps as well.
- 35:19
- Unidentified speaker (UNKNOWN): Um
- 35:19
- Dan Williams: but the publics are a key input to to that. And so it it uh it does move our marks up or down, depending on what what's in vogue or not, um day in and day out. But um
- 35:32
- Dan Williams: our focus really at the board level and at the fund level is growing these businesses and ensuring that the underlying
- 35:39
- Dan Williams: metrics and KPIs of those businesses continue to improve. Um and so
- 35:45
- Unidentified speaker (UNKNOWN): you know,
- 35:45
- Dan Williams: I'd I'd say it is an input, but it is not the focus of how we feel about the success of the portfolio at any point in time.
- 35:53
- Unidentified speaker (UNKNOWN): Uh
- 35:53
- Dan Williams: it moves a little less. I've got hedge fund buddies.
- 35:56
- Unidentified speaker (UNKNOWN): Our
- 35:56
- Dan Williams: luckily our portfolio moves a little a little less volatility than than theirs too.
- 36:01
- Unidentified speaker (speaker_2): Absolutely. And Delta V now manages over one point six billion dollars in assets across Denver and Dallas. Uh as a firm scales its AUM, the temptation is often to write larger checks into later stage companies. How do you maintain the discipline of your core infrastructure software vertical without
- 36:21
- Unidentified speaker (speaker_2): experiencing strategy drift?
- 36:23
- Unidentified speaker (UNKNOWN): Yeah,
- 36:23
- Dan Williams: we we we've grown them, but grown them in a very healthy manner that has not created a a dynamic where we felt like we were pressured to significantly raise our our check sizes. If anything,
- 36:34
- Unidentified speaker (UNKNOWN): you
- 36:34
- Dan Williams: know, we probably a decade ago we were doing um maybe a few less deals a year than we are doing now. And so um pace of for the probably the the number of companies we back is going up. The check size uh has has not believed ballooned in any way that
- 36:52
- Unidentified speaker (UNKNOWN): I
- 36:52
- Dan Williams: think was unhealthy. Um and and we think that's important competitively um to be able to find, as I mentioned, these unique ways to invest in the companies we back, we may find a five million dollar secondary uh check opportunity and we don't want to throw that away or ignore it. We want wanted it to move the needle for the fund. Um
- 37:12
- Dan Williams: in in many cases Arctical Arctic for Wolf, example, our first investment into Arctic Wolf was a six million dollar check. And we wound up investing over $80 million in that company over our our period there.
- 37:25
- Unidentified speaker (UNKNOWN): And
- 37:25
- Dan Williams: so a a small check of the onset is not a bad thing from our perspective. And we wanted our funds to to allow for that. There are fund sizes out there that mean that you have to show up and you have to write a hundred million dollar primary check out of the gate. And to us that just shrinks your investable universe, you know, pretty dramatically. We do have the ability to to
- 37:45
- Dan Williams: to accordion our investment size and to grow it a bit larger. We have a number of um great partners, uh, strategic uh limited partners in our fund that like to invest in alongside us into deals as those present the opportunity for a a larger check. And so we
- 38:03
- Unidentified speaker (UNKNOWN): uh
- 38:04
- Dan Williams: for example, we we completed a multi-hundred million dollar investment uh this year into a digital infrastructure business called Vero that I'd mentioned. Um and and so that that is more of an anomaly than the norm. Uh
- 38:17
- Unidentified speaker (UNKNOWN): out
- 38:17
- Dan Williams: of our fund,
- 38:18
- Unidentified speaker (UNKNOWN): you know,
- 38:18
- Dan Williams: we we look to write five to fifty million dollar checks, um, but have kept that fairly constant over our probably over the last decade of investment.
- 38:27
- Unidentified speaker (speaker_2): Absolutely. And uh Delta V recently elevated Garrett Marsilio to partner, making him the youngest partner in the firm's history. As a longstanding partner yourself, what cultural frameworks have you built at Delta V to ensure young investment talent can successfully transition from sourcing deals to leading complex
- 38:47
- Unidentified speaker (speaker_2): technical investment thesis? pieces.
- 38:49
- Unidentified speaker (UNKNOWN): Yes, Yeah,
- 38:49
- Dan Williams: it's really important and uh Garrett,
- 38:52
- Unidentified speaker (UNKNOWN): you
- 38:52
- Dan Williams: know, super, super proud to have promoted him to partner and to have seen him grow up through the firm. Um, we I I was with a bio firm prior to Delta V. This was over 12 years ago now. Um, and and many, many of the large firms go through a rotation process where you hire an associate and then their expectation is they're there two
- 39:12
- Dan Williams: years, maybe four years, and then they rotate out. Um, and that has never been our our view or our belief. We have we think that finding great people is hard. And if you can find great people that also fit the culture uh and are are hungry and are working hard, uh, that that we want to keep them in the family, continue to
- 39:32
- Dan Williams: grow them throughout their Other careers. Um, and so we go in the other other direction, you know, in that entirely. Um
- 39:38
- Unidentified speaker (UNKNOWN): from
- 39:39
- Dan Williams: uh you talked about maybe the
- 39:42
- Dan Williams: the tactical day-to-day uh of different levels in the career in in a firm like ours. And uh we actually start our associates hit the ground really focused on deal underwriting and going deep in around a thesis around surround our sectors, um, so more focused on the company and learning, learning
- 40:02
- Dan Williams: uh how to underwrite a deal, how to do diligence, you know, things like that. That is a different skill set, as you point out, than the partner that needs to be able to source a hundred deals a year and build relationships with CEOs and negotiate a transaction and ultimately sit on a board. And so it is a it's a learning
- 40:22
- Dan Williams: process along each step of the journey and requires new skills to be learned as you go from associate to VP to principal, et cetera. But um we have a pod or pods that focus on each of our folk or of our core sectors. And so an associate comes in and uh Nikki, who joined us a year ago, for example, and my team, hit the ground,
- 40:42
- Dan Williams: focused on cyber from day one. She went to RSA and Black Hat and the big sector conferences last year. She just got back from one with Data Tribe and um she's getting as knowledgeable as she can about the industry. And we're getting her um some reps around sourcing that'll ultimately be an important part of her role as she continues to
- 41:02
- Dan Williams: grow. So it's um you know, they used to say VC is a um an apprenticeship model. I think that's still true, even in this day. Really seeing reps, seeing more deals, learning how to engage with and and uh build relationships with entrepreneurs is important. Um
- 41:19
- Dan Williams: maybe one other aspect are when we vote on a deal at the IC level at our investment committee level, every member of the firm votes on that deal.
- 41:29
- Unidentified speaker (UNKNOWN): Uh
- 41:30
- Dan Williams: whether it passes or or fails depends just on certain IC voting members that are partners, but we want to know what the
- 41:37
- Unidentified speaker (UNKNOWN): the
- 41:37
- Dan Williams: the associate, associate the joint. that joined a week ago thinks about the deal. We think it's important to record their vote to get their input on it. We're a very democratic organization. So from a culture perspective, it is responsibility and expectations from day one. But at the same time, we we work hard and we play hard. We have regular offsites quarterly
- 41:57
- Unidentified speaker (UNKNOWN): in
- 41:57
- Dan Williams: places like Carmel and St. George. And um we went to uh Ben, Bend, Utah or Ben, Bend, Oregon one year.
- 42:03
- Dan Williams: Um very active culture. We get out together, ride bites, you know, get out in the nature and and so good supportive culture, but in important that folks learn and and we're excited when folks can uh scale their careers here and stay here long term.
- 42:19
- Unidentified speaker (speaker_2): Absolutely. It's amazing. And if you look forward, what is one non non-consensus consensus belief you hold about the future of infrastructure software? or decentralized data security that the rest of the venture cover in c industry or community is currently overlooking or enterprising?
- 42:35
- Dan Williams: Yeah, I would have said that um this, you know, when Mythos came out there was the fear around cyber and and I would have said that it's the importance of cyber and securing AI. Um now probably there's a new hot take, which is that AI is gonna going to take over the world and and uh humankind will face its its demise because of it. And
- 42:55
- Dan Williams: there's a bit of a doomsday uh view after the OpenAI attack and hugging face. So I think there's a tendency to overreact right now, anytime some of this new news comes out. Um and I think in that case in particular,
- 43:09
- Unidentified speaker (UNKNOWN): the
- 43:09
- Dan Williams: OpenAI OpenAI agent uh was given a task and it went and did what it was asked to do. It was creative and how it accomplished that, but I don't think it had its own nefarious plans.
- 43:19
- Unidentified speaker (UNKNOWN): Um,
- 43:19
- Dan Williams: I don't subscribe to the worries about AI taking over humanity today. I think we simply simply need good security and governance around it. Um Tia, the CEO of Himlayer, would tell you
- 43:31
- Unidentified speaker (UNKNOWN): that uh
- 43:32
- Dan Williams: he makes an analogy that
- 43:35
- Dan Williams: um AI right now is like giving a five-year-old the keys to a car and and letting them. drive. Like we we need to be smart about
- 43:41
- Unidentified speaker (UNKNOWN): we
- 43:41
- Dan Williams: we have a new technology. It has great capability. I think it can be a transformative enabler to to the world. But we've got to be very smart about how we implement it, what we allow it to do, how we put constraints around it. And if we can do that smartly, um
- 43:59
- Unidentified speaker (UNKNOWN): I
- 43:59
- Dan Williams: I I believe in humanity's humanities uh persistence over AI.
- 44:03
- Dan Williams: Um but I
- 44:06
- Unidentified speaker (UNKNOWN): you
- 44:06
- Dan Williams: know, so I
- 44:07
- Unidentified speaker (UNKNOWN): uh I
- 44:07
- Dan Williams: think that's probably the biggest myth right now that I I I would uh
- 44:11
- Unidentified speaker (UNKNOWN): I
- 44:11
- Dan Williams: would shun.
- 44:13
- Unidentified speaker (speaker_2): Absolutely. Uh quite interesting. But yeah, I completely agree with your view. I think uh and and a lot of people and companies are maybe making that mistake of uh giving too much reins in the hands of AI uh without getting the security around. it in place. And I think that's the important thing that needs to be there before uh we give it that
- 44:33
- Unidentified speaker (speaker_2): amount of freedom and trust. So yeah, let's uh see how that plays out. And now let's do a rapid fire round and uh wherein I'll ask you six quick questions about the investing you're doing, and you have to give six quick answers. Sounds good?
- 44:47
- Dan Williams: Happy to do. it. Yep.
- 44:49
- Unidentified speaker (speaker_2): Awesome. So the first one is what are the sectors and regions you invest in?
- 44:53
- Unidentified speaker (UNKNOWN): So
- 44:53
- Dan Williams: we invest in uh infrastructure software, which is a combination of cyber and and AI infrastructure today, um, cloud and AI services, uh, so services around large large ecosystems like anthropic, uh,
- 45:07
- Unidentified speaker (UNKNOWN): and
- 45:07
- Dan Williams: then vertical AI is another big area that we we like to spend time. And then from a regional perspective, um, we have investments broadly across the United States. That's really our main focus. But we opportunistically will invest in Europe and and in particular Israel has been a real hotbed for cybersecurity companies. We have a few investments over
- 45:27
- Dan Williams: there.
- 45:28
- Unidentified speaker (speaker_2): Awesome. Uh what's the typical stage of investment where you come in?
- 45:32
- Unidentified speaker (UNKNOWN): We
- 45:32
- Dan Williams: are looking to back companies that are past venture risk. They're really in that scaling phase of a company. uh, which for us tends to be kind of series B or series C and beyond. There's no upper bound. Uh we don't invest in the public markets, but in terms of privates, uh along that scaling phase is really the focus.
- 45:52
- Unidentified speaker (speaker_2): Great. Uh do you lead rounds?
- 45:54
- Dan Williams: We do. Absolutely. We lead, we we are flexible. We will lead a primary round, we'll participate in a primary round, we'll buy secondary on a direct basis as we mentioned. We'll fund MA activities on a one off basis as well. And uh yeah, I think our current fund, we have board seats in something like 80% of our deals. So uh the leading part
- 46:14
- Dan Williams: has become a a bigger and bigger part of our business over time.
- 46:18
- Unidentified speaker (speaker_2): Absolutely. And what's the typical check size you put in initially?
- 46:22
- Dan Williams: Five to fifty million dollars is our check size range out of the fund as I mentioned, we can scale that up a bit with uh some some investment from our strategic LPUs, but uh that's kind of down the fairway.
- 46:35
- Unidentified speaker (speaker_2): Great. Uh where can found us get in touch in case there's a direct
- 46:38
- Unidentified speaker (UNKNOWN): way.
- 46:39
- Dan Williams: Absolutely. I'm probably most active on LinkedIn, uh Dan Williams, and uh or you can find the Delta V capital page on LinkedIn as well. Uh, Feel free to reach
- 46:49
- Unidentified speaker (UNKNOWN): out.
- 46:49
- Unidentified speaker (speaker_2): Awesome. Where you can listen listeners follow you, it'll be LinkedIn again, I guess.
- 46:53
- Unidentified speaker (UNKNOWN): Yeah,
- 46:53
- Dan Williams: I tend to be fairly active on LinkedIn and a little bit on X as well. Um, we also have a a views uh tab on our website that you can check out where we announce new investments or views on them and and other important uh sale points.
- 47:10
- Unidentified speaker (speaker_2): Right. Awesome. I'll make sure to put those links in the show notes below. Uh thanks so much for
- 47:14
- Unidentified speaker (UNKNOWN): the next video.
- 47:15
- Unidentified speaker (speaker_2): doing this for the stand. I wish you happy investing.
- 47:17
- Dan Williams: Thank you for having me. I really appreciate it. It's a great conversation.
- 47:21
- Unidentified speaker (speaker_2): Love it.