Linda Park

VC Money Is Flooding Into AI and Someone Has to Pay for It - Turns Out It's the Public Markets

March 8, 2026

I want to tell you what's actually happening here, because most of the coverage I've read treats this like two separate stories. It isn't. It's one story, and the two halves are directly connected in a way that should concern anyone running a business that depends on software - which at this point is basically everyone.

The first half of the story is the money. Venture capital poured $211 billion into AI companies in 2025 - up 85% year over year from $114 billion in 2024, according to Crunchbase data. That's not a typo. AI captured close to 50% of all global venture funding last year, up from 34% the year before. To put that in context: one category, artificial intelligence, absorbed roughly half of every dollar invested in every startup on earth. The biggest single deal was OpenAI's $40 billion SoftBank-led round in March, which by itself represented more than a third of all global VC investment in Q1. One company. One quarter.

The second half of the story is what happened to public software companies at almost exactly the same time. The SaaS index - the one that tracks publicly traded subscription software businesses - fell 6.5% in 2025 while the S&P 500 rose 17.6%. Then in early 2026, things got considerably worse. The IGV ETF, which tracks North American software, fell almost 20% year to date by early February and was sitting close to 30% off its September 2025 peak. Software price-to-sales ratios compressed from 9x to 6x - levels not seen since the mid-2010s.

These two things are not a coincidence. They are cause and effect, and I think the business press has been annoyingly slow to say so plainly.

What the Numbers Actually Mean

Here's what I keep coming back to. Late-stage AI companies are commanding a 100% premium over non-AI peers at Series C, according to SVB data. A one-year-old startup building AI agents is walking into a room and getting twice the valuation of a seven-year-old SaaS company with real revenue and real customers. That premium has to come from somewhere, and it's coming from investor appetite that used to flow toward established public software names.

Stephanie asked me last week if this was just a "rotation" thing - like investors moving from one bucket to another temporarily. Stephanie grew up with a house that had an actual name, so her instinct is always that these things sort themselves out. I don't think this is that. A rotation implies the money comes back. What's happening here feels more structural than that.

The funding concentration is genuinely staggering. In 2025, 58% of all AI funding went into megarounds of $500 million or more. Foundation model companies alone - OpenAI, Anthropic, xAI, Mistral - raised $80 billion, representing 40% of all global AI funding. OpenAI is now the most valuable private company in history at a $500 billion valuation. Anthropic sits at $183 billion. Together, those two companies represent close to 10% of the value on the entire Crunchbase unicorn board.

And while all of this was happening in private markets, the companies on the other side of the equation - the public software businesses that actually have to answer to shareholders every 90 days - were watching their installed bases get questioned, their seat-based pricing models get challenged, and their forward multiples get slashed.

The Mechanism Is Brutally Simple

Here's what I think people are missing about how this actually works. The VC flood into private AI companies isn't just competing for investor dollars. It's competing for enterprise budgets. The companies getting all that private funding are building products that enterprise software buyers are seriously evaluating as replacements for things they currently pay for.

IT budgets are not growing fast enough to absorb both. A January 2026 CIO survey found that IT budget growth is expected to decelerate to 3.4% in 2026 - flat-to-down from prior years. And critically, funds are being diverted away from application software to pay for the massive compute costs tied to the AI infrastructure buildout. The money that used to go to SaaS subscriptions is going to AI. That's it. That's the whole story.

The catalyst that brought this into sharp focus was Anthropic's release of Claude Cowork and Claude Code in early 2026. Suddenly you had AI agents that could independently create reports, generate tables from screenshots, extract information from documents, and complete programming tasks in minutes that used to take hours. Within 30 days of that release, approximately $2 trillion in market capitalization evaporated from the software sector. SAP lost around $130 billion in market value. Oracle's stock plummeted 13% in a single day in December 2025 when AI costs exceeded expectations and revenue fell short. Salesforce dropped 14% over five days.

Gerald and I were watching the news the night the Palantir story broke - the one where CEO Alex Karp said AI had become so powerful at writing and managing enterprise software that many SaaS companies were in danger of becoming irrelevant. That single earnings call wiped out $300 billion in market cap from Microsoft, Salesforce, ServiceNow, and others. Gerald just looked at me and said something about the weather. He's not much for tech drama. But I kept thinking about it all week.

The Part Where I Push Back on the Panic

Okay. I've been in this industry long enough to remember when SaaS itself was supposed to destroy enterprise software, and then cloud was supposed to destroy SaaS, and on and on. The pattern is always the same: a new capability arrives, the press declares the old model dead, some of the old model does die, but a lot of it just adapts and survives.

"Sticky" software - ERP systems, CRM platforms, things organizations have spent years and tens of millions of dollars integrating - will not be discarded overnight because an AI agent can write a project ticket. I've spent enough time watching companies try to migrate off platforms they've used for eight years to know that the actual cost of switching is almost always higher than what you'd save in subscription fees. The switching friction is real. It doesn't show up in the analyst reports but it absolutely shows up in the timeline.

I'll also say this: the Klarna story is instructive. CEO Sebastian Siemiatkowski announced in 2024 and 2025 that Klarna would halve its workforce, eliminate 1,200 SaaS tools including Salesforce and Workday, and replace everything with internal AI solutions. An OpenAI-based chatbot reportedly handled the work of 700 customer service reps and saved $40 million a year. Sounds great. Except by early 2025, Siemiatkowski had publicly admitted the approach had gone too far and that quality was suffering. Klarna began rehiring customer service people because the AI systems had created quality problems.

That's not a story about AI failing. It's a story about the gap between what AI can theoretically do in a demo and what it actually delivers at scale under operational pressure. That gap is real and it's going to matter for a lot of the startups currently sitting on $500 million funding rounds with sky-high private valuations.

What This Means if You're Running a Business

The most important thing I want to say here is about timing. The companies getting punished in public markets right now are being punished based on what investors think is going to happen to their business models - not necessarily what is currently happening. Public companies have to live in the future all the time; that's the nature of forward multiples. Investors are essentially saying: we believe AI agents will erode your seat count and compress your net revenue retention, and we're pricing that in now.

That means if you're a business that buys software - which is most of us - you're in an interesting position. The software vendors you currently use are under serious pressure to demonstrate value in an environment where alternatives are proliferating. That's actually good for buyers in the short term. Renewal conversations are going to get more favorable. Vendors who were comfortable with their pricing two years ago are going to be a lot more flexible.

I've noticed this already in how some of the outreach tools work. The cold email and sales automation space specifically is being disrupted in real time - I've watched pricing shift meaningfully in the last year as AI-native competitors entered the market and the incumbents scrambled to add features they probably should have built two years earlier. The sales automation software market right now looks a lot like what's now happening at scale across the whole software industry - a noisy, confusing middle period where everything is being repriced and nobody really knows which tools are going to be around in three years.

And if you're buying software for a platform or business you've built on - which is a different and scarier situation - you should be paying attention to what happens when a platform gets acquired or destabilized. The current environment is going to produce a lot of M&A. Last year, M&A activity for AI startups surged to a record 177 deals, double the five-year average of 89. Companies under pressure in public markets become acquisition targets. That's going to affect product roadmaps, pricing, and support in ways that aren't always telegraphed in advance.

Derek spent an entire lunch last Tuesday explaining why the original Star Wars trilogy had an unfair advantage over the sequels because audiences weren't yet "AI-fatigued." I'm not entirely sure what that means, but it did make me think about how expectations shape reception. The software companies getting hammered right now aren't failing - most of them are still growing. They're just failing to meet the expectations that their previous valuations implied. There's a difference, and it matters when you're deciding which vendor to bet on for the next few years.

The Part the Hype Is Getting Wrong

I think the narrative that "SaaS is dead" is doing real damage to how people are thinking about this. It's not dead. It's being restructured. The per-seat model is under pressure - genuinely, structurally under pressure - but that doesn't mean software stops being valuable. It means the pricing model has to change. Outcome-based pricing, usage-based pricing, value-based contracts - these are conversations that vendors and buyers are going to be having for the next several years, and the businesses that navigate that transition well will come out fine.

What I'm actually more worried about is the private side of this equation. When VC money is flooding into private AI companies at historic rates, with 58% of it in rounds of $500 million or more, you're creating a class of well-capitalized companies that can afford to acquire customers at a loss for years. That is deeply disruptive to competitors, but it also means a lot of those companies are going to need exits that may not materialize if the public markets for software stay cold. The executives who committed to AI strategies in front of investors are now in a difficult position: they can't walk it back, but the ROI timeline keeps getting pushed.

And the IPO market for this wave of private AI companies is already showing stress. Valuation volatility has effectively frozen the IPO window. High-profile debuts have been pulled from the calendar, leaving private equity firms and venture capitalists with few exit strategies. The money went in. It is not easily coming out. That's going to create pressure on some of these companies to actually generate revenue - real revenue, not "AI-accelerated growth" that is mostly sales headcount - sooner than they'd like.

I made Gerald his favorite chicken casserole last Thursday and he came home and just ate it quietly and didn't say much. I could tell he liked it. I thought about the VC situation the whole time I was making it - specifically about how the people putting $40 billion into a single company in one quarter are making a very loud argument that they know something the public markets don't. Maybe they do. They've been right before. But I've also been in enough software selection cycles to know that the gap between a compelling demo, a funded company, and a product that actually holds up in a real operational environment is very, very wide.

The companies getting punished in public markets aren't necessarily going to die. Some of them have the resources and the existing relationships to survive this transition and come out the other side as AI-native platforms themselves. The ones that are going to get hurt are the ones in the middle - mid-market SaaS companies without the cash reserves of a Salesforce or ServiceNow, and without the VC backstop of a well-funded AI startup. That's the part of the market I'd be watching most carefully right now.

The money is real. The pressure on incumbents is real. The disruption is real. But the idea that this resolves cleanly and quickly into a world where AI agents replace all existing software - that part is not real. Not yet. And by the time it is, a lot of the companies currently burning through their $500 million rounds at a loss will have needed to figure out something else entirely.