In the latest episode of What the Tech from Boast, we sat down with Tony Kim, a corporate investor focused on AI infrastructure, deep tech, and enterprise SaaS. Tony works at a Fortune 100 company’s corporate development and venture arm, where he’s engaged in everything from M&A to venture capital and LP investing.

What makes this conversation particularly timely is the perspective Tony brings: He’s operating at the intersection of traditional corporate strategy and the breakneck pace of AI innovation. As someone who started his career as an investment banker covering infrastructure and telecom in New York City, he’s watched the foundational layers of technology evolve. Now, he’s helping shape how a major corporation responds to the AI revolution.

His insights on corporate venture capital (CVC), AI investment strategy, and what separates genuine innovation from hype are exactly what founders need to hear as they navigate 2025’s rapidly changing landscape.

The CVC Difference: Strategy Over Pure Returns

One of the most valuable parts of our conversation was Tony’s breakdown of how corporate venture capital differs from traditional VC; distinctions that many founders don’t fully appreciate.

Traditional VCs are primarily financially driven, operating with fund structures backed by limited partners (LPs). The goal is straightforward: Maximize returns for investors.

CVCs, on the other hand, often operate differently. Some have their own fund structures, but many (like Tony’s) invest directly from the company’s balance sheet. This fundamentally changes the mandate.

“Instead of financial returns, there’s also often cases where CVCs are seeking strategic goals. It could be forming a closer partnership or it could be forming a closer go-to-market channel relationship. The mandates from a CVC perspective is often a little broader or also differentiated.”

What This Means for Founders

The approval process is more complex. While traditional VCs might approve deals after partners review an investment memo, CVCs typically require buy-in from business units as well. Someone from the business side has to champion the deal, which means more extensive diligence but also potentially deeper strategic alignment.

Understanding this distinction is crucial when you’re fundraising. If you’re talking to a CVC, you need to articulate not just why your company will generate returns, but how it aligns with the parent company’s strategic objectives. That might mean demonstrating potential partnerships, go-to-market synergies, or technology that solves problems the corporate parent is facing.

AI Has Changed Everything: Traction Matters More Than Ever

If there’s one theme that dominated our conversation, it’s how dramatically AI has shifted the venture landscape in just the past 18-24 months.

The barriers to launching have dropped. We’re seeing companies reach annual recurring revenue (ARR) milestones in 12 months that traditionally took three to five years. AI-native companies are fetching higher valuations and attracting intense competition from investors trying to participate in rounds.

But here’s the flip side: AI isn’t an excuse for low traction anymore.

“As much as they’re fetching a high valuation, [AI companies] also have really a strong growth rate from an ARR perspective. It’s been interesting to see a lot of companies applying, even though they’re pre-seed and seed stages, already having traction that I don’t normally typically see a couple years ago.”

Translation: If you’re building an AI company, investors expect you to move faster and show more traction earlier than they would have for a traditional SaaS company five years ago. The tools are available; the question is whether you’re using them effectively.

VCs Are Changing Their Approach

In response to this competitive landscape, VCs are innovating to get closer to the talent ecosystem. We’re seeing:

This represents a shift: Instead of founders primarily chasing brand-name VCs, VCs are actively working to attract AI-native talent before they even formally launch companies.

Separating Real Innovation from AI Wrappers

With so much capital flowing into AI, one of the most important questions for investors (and for government R&D programs like SR&ED) is: What’s actually innovative versus what’s just riding the hype wave?

Tony’s answer gets technical, which is exactly the point.

“The fundamental design as well as the architectural structure matters a lot more. The goal here is that you’re diving into these companies to understand they’re not just a wrapper in some ways. They’re not just an OpenAI wrapper, they’re not just an Anthropic wrapper.”

What Investors Look For

When evaluating AI companies, Tony digs into:

  • AI memory architecture – How is data being stored and accessed?
  • Infrastructure choices – What GPUs are you using? Why those specifically?
  • Network and storage architecture – How have you customized your infrastructure provisioning?
  • Training vs. inference focus – Are you optimizing for the right workload?

Companies that can’t answer these questions in detail are likely just API wrappers using a single external API without any meaningful differentiation in their technical stack.

Real innovation involves customizing the infrastructure, making thoughtful architectural decisions, and building technology that’s fundamentally different from what a competitor could replicate in an afternoon with access to the same APIs.

The Boast Connection

This distinction matters enormously for R&D tax credits. Whether you’re pursuing Canadian SR&ED credits or US R&D tax credits, the fundamental requirement is the same: You must be doing something new and technologically uncertain.

As Tony and I discussed, we’ve seen an explosion of R&D tax credit providers that are essentially AI wrappers themselves, ie. automated tools that promise to handle your claim with minimal human involvement. The problem is these tools often miss the deep technical understanding and expertise needed to properly document genuine innovation.

“You’re getting what you pay for with a lot of those off-the-shelf solutions. They’re not best-in-breed. You absolutely need somebody with that expertise and you need a human on your team.”

The companies that successfully claim R&D incentives are those that can clearly articulate:

  1. What technological uncertainty they’re tackling
  2. Why existing solutions don’t address the problem
  3. What systematic approach they’re taking to solve it
  4. How their architecture or methodology is genuinely different

If you’re just wrapping an existing API, you don’t meet these criteria, and you shouldn’t expect government funding programs to reward that work.

The Full-Stack AI Investment Thesis

Tony’s investment focus spans the entire AI stack, from bottom to top:

Hardware Layer:

  • Semiconductors and AI accelerators
  • Optical interconnects
  • Next-generation memory architectures

Infrastructure Layer:

  • Data centers (both greenfield and brownfield projects)
  • Storage and networking solutions optimized for AI workloads
  • Energy infrastructure to support massive compute requirements

Software Layer:

  • Large language models (LLMs)
  • AI agents and agentic frameworks
  • Application layer innovations

This full-stack view is important because supply constraints remain the biggest bottleneck. Demand for AI compute, memory bandwidth, and energy far exceeds current supply. That creates enormous opportunities for companies that can address these constraints at any layer of the stack.

The AI Agent Revolution: Still Early Innings

Despite all the hype around AI agents, Tony’s perspective is that we’re still in the very early innings, particularly on the agentic AI side.

The vision is compelling: AI systems that can book your travel, coordinate complex workflows, and operate semi-autonomously across multiple tasks with minimal human intervention. Companies like Lovable (formerly Lovable Labs) show what’s possible, building impressive ARR in record time.

But fundamental questions remain:

  • How efficient can AI agents actually become?
  • Will they truly replace human labor in meaningful ways, or augment it?
  • Can they operate reliably across complex, multi-step processes?

“The craziest thing is that I don’t think we are far from there. I think we are getting there. It could potentially solve all the fundamental problems for you.”

The benchmark data over the next six months will be critical in answering these questions. We’ll see whether the promise of agentic AI translates into measurably better outcomes across industries like healthcare, legal, and beyond.

What’s Next: Six-Month Predictions

When I asked Tony for his crystal ball on the next six months, several themes emerged:

Hyperscaler Infrastructure Buildout

The major cloud providers (hyperscalers) are committing $500 billion+ to AI infrastructure over the next few years. This includes:

  • New data center construction (both greenfield and brownfield projects)
  • Custom chip development (reducing reliance on third-party vendors)
  • End-to-end stack integration (owning as much of the value chain as possible)

Global AI Race Intensifies

We’re seeing an arms race dynamic play out:

  • US focus: Massive private investment in closed-source models (OpenAI, Anthropic, etc.)
  • China response: Significant catch-up investment, with interesting innovation in open-source models

This isn’t just a commercial competition—it’s becoming a matter of national interest, with geopolitical implications that will shape policy and investment for years to come.

Continued Full-Stack Investment

Funding will continue flowing across the entire AI stack, but AI agent innovations will capture outsized attention. The question of how these systems can fundamentally change workflows and replace (or augment) human labor will dominate conversations.

The Infrastructure Imperative

Beyond the algorithms and agents themselves, the infrastructure required to support AI at scale needs dramatic expansion. Tony’s background in infrastructure and data center financing gives him a unique lens here—he’s seen this story before, and knows that foundational infrastructure investments often create the most durable value.

Key Lessons for Founders

Tony’s parting advice for founders navigating this landscape:

Embrace Change

“Because of AI, if you’re an AI-native company, maybe it made fundraising slightly easier. But it could also be in many regards for other traditional SaaS companies, it could be pretty daunting for them as they maybe have to pivot into AI. Try to embrace changes in any way as possible. Don’t be scared.”

Traction Still Matters

Even if you’re not AI-native, strong thesis and strong traction will still attract investment. Don’t feel pressured to rebrand as an AI company if it’s not authentic to your core value proposition, but do think seriously about how AI might enhance what you’re building.

Understand Your Audience

If you’re talking to CVCs, understand that the evaluation criteria are different from traditional VCs. Strategic fit, business unit buy-in, and long-term partnership potential matter as much as pure financial returns.

Build Real Differentiation

Don’t be an API wrapper. Invest in understanding your infrastructure, making thoughtful architectural decisions, and building something that’s genuinely difficult to replicate. This matters for fundraising, but it also matters for capturing R&D incentives that can extend your runway.

The Bottom Line

We’re living through a fundamental shift in how technology companies are built, funded, and scaled. AI has lowered barriers to entry while simultaneously raising the bar for what constitutes impressive traction. Corporate venture capital is expanding its mandates and competing more aggressively for deals. And the full-stack infrastructure required to support AI at scale represents a multi-hundred-billion-dollar investment opportunity.

For founders, the message is clear: embrace the change, build real technical differentiation, and understand that the old playbooks don’t fully apply anymore. Whether you’re pursuing venture capital, corporate strategic investment, or non-dilutive funding through government R&D programs, the fundamentals haven’t changed—you need to solve real problems with genuine innovation.

But the pace has accelerated, the competition has intensified, and the opportunities have never been bigger.

Listen to the Full Episode

Want to hear more about corporate venture capital, AI investment strategy, and what separates genuine innovation from hype? Tony shares insights on everything from semiconductor alternatives to the global AI race.

The views expressed in this interview by Tony S. Kim are his own and do not reflect those of his employer.