Walk into any investment committee today and the conversation has shifted. Allocators are no longer just asking 'how do we get access?'—they’re asking 'is this 1999 all over again?
It is the right question to ask. And for me, it’s personal.
I’ve navigated the extremes of market cycles before—founding a startup during the dot-com mania and managing distressed debt during the 2008 crisis. Today, as a YC founder and GP with over 230 early stage investments at Eight Capital, I view the AI wave through the dual lens of builder and allocator.
I know what irrational exuberance feels like, and I know exactly what capital destruction looks like.
We are seeing some of those same patterns today: enormous capital spend, a handful of dominant public names carrying the index, and narratives swinging violently between “this changes everything” and “this ends badly.”
In a recent memo, Howard Marks lays out that history—railroads, radio, the internet—and introduces a distinction that perfectly frames our thesis at Eight Capital:
- Mean-Reversion Bubbles: Financial fads (like subprime mortgages) that inflate asset prices without fundamentally improving the world. These end in pure value destruction.
- Inflection Bubbles: Periods where a genuinely transformative technology (railroads, electricity, the internet) gets massively overbuilt. These bubbles destroy investor capital in the short term, but they permanently raise the world’s productive capacity.
My take: AI is almost certainly an Inflection Bubble. It will change the world, but it may still destroy capital along the way.
So, how do we invest in an inflection bubble without getting burned? By applying the lessons of 2000 and 2008.
Two Bubbles, Not One
Marks’ memo makes a critical distinction often missed in the headlines: there is a difference between a "company behavior" bubble (hyperscalers, GPU build-outs, debt) and an "investor behavior" bubble (pricing, lottery-ticket thinking).
We don't fund trillion-dollar Capex or $2B seed rounds. Our strategy at Eight Capital sits in a completely different part of the stack:
- Invest in the Application Layer: We back early-stage, capital-light software companies—specifically YC startups—that consume AI infrastructure rather than build it.
- Write Small, Diversified Checks: We enter at the pre-Demo Day and seed stage, often into companies that already have users and revenue, avoiding the valuation distortion of late-stage rounds.
- Avoid Leverage: We use zero debt at the company or fund level.
In short: if the AI infrastructure build-out turns out to be overbuilt and debt-heavy, the pain will sit with those financing $5T of data centers—not with a YC company using that cheap compute to sell workflow automation to banks or biotech firms.
I am not trying to be the next NVIDIA. I am backing the founders building profitable, compounding businesses on top of what Nvidia and OpenAI provide.
The "Installation" vs. "Deployment" Phase
The memo cites Carlota Perez’s framework: technological revolutions begin with an "Installation Phase"—a mania of over-investment that lays the rails and wires future winners will run on. This phase is chaotic, expensive, and prone to crashes.
But it is followed by the "Deployment Phase": the less glamorous, highly profitable period where the new technology is embedded into the fabric of the economy.
Eight Capital is deliberately biased toward the Deployment Phase.
We back companies that don't need the world to be perfect for AI infra valuations. They simply need:
- Customers with real problems.
- Willingness to pay for time saved.
- Foundational models that become cheaper and more capable over time.
As the "Installation Bubble" overbuilds capacity, these companies benefit. They get to buy better compute at lower effective prices, translating the macro excess into better micro unit economics.
What We Intentionally Avoid
Having navigated the GFC, I have a deep aversion to the behaviors that historically end in tears. I use these as an anti-portfolio filter:
- "Lottery-Ticket Thinking": We avoid betting on massive outcomes with near-zero probability. Across our 230+ investments, we don't need a single company to return the fund; We aim for a high hit rate of solid, growing businesses.
- Pre-Product Mega-Rounds: We do not participate in $1B+ "seed" rounds for companies with no shipped product.
- Circular Financing: We avoid businesses relying on vendor financing or debt to fuel growth.
YC as a Filter Against "Pure Narrative"
In the dot-com era, capital flooded into everything—including weak teams with good stories. While no filter is perfect, Y Combinator provides structural advantages in an overheated market:
- Selection: YC screens thousands of applicants down to the top ~1.5%.
- Discipline: The 3-month batch structure forces founders to ship product and talk to users, rather than just raising money on slides.
- Data: We have seen comparable data having invested across 11+ batches, allowing us to spot real traction (revenue, usage) versus noise.
As a current YC founder, I see this discipline firsthand. Within that top tier, we narrow further—selecting roughly the top 10-15% based on traction, capital efficiency, and founder quality. We are not buying the "AI hype"; We are investing in specific teams solving specific problems.
Conclusion: Holding Two Truths
My experience has taught me to hold two competing thoughts at once:
- "AI enthusiasm will almost certainly overshoot."
- "AI is one of the most important technology shifts of our lifetimes."
Our job isn't to predict when the bubble pops. My job is to avoid being the marginal dollar funding the excess, and instead ensure we own a diversified basket of the companies that will survive the correction and define the next decade.
I'm not betting on the bubble. I'm betting on the builders.
Ravi Chachra is the Managing Partner of Eight Capital Management, a seed-stage fund focused on Y Combinator companies
