Second Brain

This week I co-wrote Redpoint's 2026 Market Update for our Limited Partners with my colleagues

and

. We do this every year and it's a great time to step back and reflect on the state of everything going on.

It's an unusual moment- the most dynamic markets I've seen, so we tried to be rigorous about what's actually happening. The whole deck is linked at the bottom, but a few things that standout.

Not the Dotcom Bubble 2.0.

Every time there's a pullback, someone pulls out the 1999 comparison. I understand the instinct but I think it's wrong, and the data is pretty clear.

In 2000, fiber utilization was below 3% at the peak. Revenue was essentially zero. Today, OpenAI and Anthropic are each generating $20B+ ARR. More than 90% of new data center capacity is pre-committed before breaking ground. Physical constraints including power, land, and interconnect make speculative overbuilding structurally harder than it was in the dotcom era, when laying extra fiber strands was cheap. ChatGPT reached 1B monthly active users in roughly 4 years. The internet took the same amount of time to reach 70M.

The infrastructure isn't running ahead of demand. Demand is pulling the infrastructure forward.

The Agent Maturity curve is still early, but the TAM implications are enormous.

Until last year, we were largely in “Copilot” territory: AI making people more productive, mostly competing for existing software budgets. Useful, but fundamentally a fight over the same dollars.

We're now entering Task Agent territory. Agents that can run discretely for minutes, execute end-to-end tasks with some human oversight. This is where things get interesting, because it starts to access labor spend rather than just software spend. The US software market is roughly $0.5T. By our calculation, Task Agents expand the market to $1.2T.

Workflow Agents, AI orchestrating multi-step business processes across systems for hours at a time, are here in part and coming in full. Autonomous Agents running for days with minimal oversight are further out but no longer science fiction. If you're really able to run autonomously, you start to target the whole of knowledge worker pay ($6.2T in the US). The maturity curve is real, and we are early on it.

The software selloff is real, but it hasn't hit every layer equally, and that's not random.

Software is down 20% YTD, the worst-performing sector in the S&P 500. Public SaaS multiples are at 4.1x NTM revenue, the lowest in over a decade. The market is pricing in near-zero long-term growth for a lot of these companies. But the selloff has landed very differently across vertical SaaS, infrastructure, and horizontal SaaS, and the reasons matter.

Vertical SaaS is largely holding (+3% LTM). These companies often own an irreplaceable moat: proprietary industry data accumulated over years, workflow systems embedded in regulated processes, switching costs that are existential rather than cosmetic. AI can augment these systems. It cannot displace them without re-acquiring all that data. That's not a feature request. It's a decade of work.

Infrastructure is roughly flat (+2% LTM). We believe AI is a tailwind here, not a threat. More AI deployment means more compute, more data, more observability spend. Snowflake, Datadog, MongoDB, Crowdstrike, Cloudflare: new AI workloads run on or through all of these. It feels like the mild compression reflects rate normalization and some misguided sentiment, not an existential question about the business.

Horizontal SaaS is down 35% LTM. This is where the structural concern is real. Horizontal apps were built to serve every industry equally, which in practice meant integrating deeply with none. They don't hold industry-specific data. They often don't own a bunch of core workflows. And critically, many of them were built to answer a simple question: who is doing what, and when? That's a coordination problem. AI solves coordination problems natively. When the core value proposition of your product is something AI does by default, that's a different kind of threat.

The data from our CIO survey tells a consistent story.

45% of AI budgets are coming out of existing software budgets, not net new spend. 54% of CIOs are actively pursuing vendor consolidation. Only 3% expect AI to lead to more vendors. Existing revenue is less sticky than it looks on the surface. The categories most open to AI displacement, salesforce automation at 83%, customer service management at 56%, ITSM at 55%, aren't random. They're the categories where AI can do the work, not just assist it.

That said, a lot of these incumbents are going to get every shot at building the AI workflow systems. 54% of CIOs still prefer incumbent vendors adding AI over AI-native alternatives. The question is whether they can do it.

Re-founding is genuinely hard.

The challenge for incumbents isn't shipping an AI feature. Building an AI-native company is different across almost every organizational vector.

Two that stand out: AI-native companies don't need executives who have "been there, done that." They need first-principles thinkers, because genuinely nothing has been done before at this layer. And product development is inverted. Instead of listening to customers and building to spec, AI-native product teams have to deeply understand what customers' day-to-day looks like and then reason forward from what the models can actually do. That requires a completely different orientation.

This isn't a mobile app launch. It's a re-founding. The companies that treat it as an incremental feature effort will lose to the ones that treat it as an architectural reset.

On private market valuations: the spread looks insane until you growth-adjust.

Our comp set of tier 1 high growth Series B and C software companies are getting down at a median of 61x ARR in 2026 YTD. Public high-growth software is trading at 9.7x. A 528% premium sounds disconnected from reality.

But the median ARR growth rate for those private companies is 640%. Public high-growth software is growing at roughly 29%. When you divide multiples by growth rates, the picture flips: private companies are trading at an 86% discount to public comps on a growth-adjusted basis. You are paying significantly less per unit of growth in the private market right now.

This analysis is imperfect. It's not a clean apples-to-apples comparison and growth rates at that level are hard to sustain. But given how compressed public market growth is outside a handful of names, the simple multiple comparison misses what's actually going on.

If history is any guide, we're in the window.

Across the internet, cloud, and mobile eras, the companies that became durable winners were predominantly founded in years 4 and 5 of each platform shift. Google and Salesforce after Netscape. Snowflake and Datadog after AWS launch. Robinhood and Coinbase after the App Store. ChatGPT launched in November 2022. We are in year 4. It certainly feels possible that OpenAI and Anthropic capture the lion’s share of value this go round, but the last 3 transitions were largely consistent in this.

This is also the most crowded, fastest-moving, highest-bar environment I've seen, with rounds closing in days and valuations that require exceptional execution to justify from day one. It’s hard not to be excited and uncertain every day.

Full deck: