AI Investor Panel: Where Smart Money Is Actually Going in AI | EP 219

David Blundin stressed that the amount of capital needed by AI companies far exceeds the $200B/year U.S. venture industry, creating a vacuum that can be filled by anyone who shows up early. He used the Anthropic seed round as the ultimate cautionary tale: 21 established VCs said no, yet the few who invested gained pro-rata rights that would have allowed them to pour in billions of follow-on capital and capture a massive share of the wealth creation. The Mercor example—valuation rocketing from $30M to $10B in two years—showed that even non-foundation-model companies can deliver historic returns. David's core message: ‘It's just not that hard. You just need to get into the loops, get into the places that are making these investments and get in the game.’ The failure to do so isn't about lack of capital, but about lack of presence in the right networks.

David drew a direct parallel to the internet era: the technology was real and eventually delivered, but overinvestment in non-core, capital-intensive ideas caused a crash that left the whole sector frozen for years. Today, AI voice automation for sales alone represents a $500 billion global payroll market where AI already outperforms humans—deploying that is a near-certain winner. On the other hand, he sees money pouring into fusion energy and robotics under the guise of 'AI-related,' which are far more speculative and could consume capital without returns. Bonnie added that the opaque price discovery of AI valuations means public investors entering late could be the bag-holders when the bubble deflates. The protocol is to stay disciplined: overweight vertical AI applications with clear ROI, underweight capital-intensive moonshots unless you have the risk appetite and horizon to tolerate a bust.

Anjan framed the AI transition as a societal time bomb: productivity gains will vaporize large portions of sectors like IT services (e.g., double-digit percentages of India's GDP), yet the gains are concentrated in a handful of private companies and their early backers. He noted that tech leaders already receive death threats, and the Sam Altman $1M retention bonus backlash signaled public anger. To avoid a ‘where's my piece?’ crisis, sovereign and pension funds must become active on cap tables. He contrasted this with current reality—most of his outreach to such funds was met with passivity, despite the Anthropic case demonstrating the opportunity. He sees these institutions as the proper vehicle to democratize AI wealth, far better than directly listing for retail investors who may buy at inflated prices. His call: ‘it's our job to educate them and make them more aggressively take a position.’

Bonnie Chan highlighted that Hong Kong has become the world's top IPO market by deal volume, with a pipeline of 300 deals, half tied to AI. Chinese companies are embedding AI across sectors—from manufacturing to drug discovery—and need public capital to scale. The exchange's investor base has evolved from simple retail to ‘protel’ investors using algorithmic strategies, creating a sophisticated demand pool. However, she acknowledged the tension between democratizing AI wealth and protecting retail investors from inflated valuations. Her solution is to use the exchange as a platform to match capital with opportunity, bringing in diverse pockets of demand worldwide, while ensuring pricing transparency. For AI founders, a HKEX listing provides an alternative to US exchanges, with the benefit of tapping into the immense Asian savings pool.

Anjan explained that a16z was founded as a verticalized firm, but those labels have collapsed. Whether teams train foundation models or build applications, every investment is now an AI investment. This mirrors how the internet ate all other software; investors who don't see that all sectors will be AI-reliant will miss the wave. The implication is that capital is being funneled almost exclusively into AI, to the detriment of non-AI startups. David reinforced that the volume of deals coming out of MIT and Harvard has quadrupled, with success rates near 100%, suggesting that the opportunity set has expanded so much that almost any talented team tackling a real use case will win. The shift is structural and permanent, not a hype cycle.

Anjan described the 'pref stack' of compute: raw cash converts to GPUs, GPUs convert to tokens, and tokens become an input for application developers. The emergence of reasoning models that generate 10x more tokens than earlier gen AI models has intensified demand, creating a Jevons paradox where algorithmic efficiency gains only increase the appetite for compute. Even with massive infrastructure buildout, we 'somehow just need more compute, more infrastructure.' This means application developers aren't just competing for funding; they're competing for token allocations from foundation model providers, which may force them to vertically integrate or pay premiums. The scarcity shifts bargaining power toward model providers like OpenAI, Anthropic, and Mistral.

Bonnie noted that mainland Chinese companies that aren't doing it at the core of AI development 'probably quite unable to compete.' Her pipeline includes not only infrastructure plays but also data-intensive sectors like drug discovery, where AI is dramatically shortening discovery cycles. The retail investor base in Asia has evolved; they now have algorithmic trading strategies and a deep appetite for tech. However, she cautioned that opening public markets to retail at sky-high private valuations creates the risk of the public being the last to participate before a correction. Her solution is to find ways to match capital with opportunities through multiple avenues: private, public, credit, equity—anything that broadens participation while managing price discovery transparency.

Anjan personally made 22 introductions to VCs up and down Sand Hill Road; 21 declined. The round was pieced together from angels. He remains 'shocked at how often today traditional venture sovereign funds, traditional pension funds are not being aggressive enough' despite this clear lesson. David Blundin hammered home the opportunity: anyone who invested at the seed got pro-rata rights, potentially allowing them to pour in 'what four five ten billion dollars of follow on.' The asymmetry was enormous, yet many institutional investors stayed on the sidelines. This episode is a case study in the gap between AI wealth creation and institutional participation—a gap that, if not closed, will leave the public angry and excluded.

David argued that companies applying AI to voice-driven sales and customer support represent a half-trillion-dollar payroll market where AI already outperforms humans—investment here 'you cannot go wrong.' In contrast, pitches for fusion energy or quantum computing are being sold as AI plays because energy and compute support AI, but they are much more speculative and capital-intensive. If those bets fail, they could scare off the entire investment community, mirroring the 2001 aftermath. Bonnie added that valuation discovery is currently opaque and driven by a small group, so retail investors entering late at inflated prices could be the greatest losers. The solution: pour capital into proven vertical AI wins while being cautious about hype-adjacent peripherals.