If you’re reading this, you already know we’re living through the most consequential technological shift in a generation, and the real question isn’t what AI will do — it’s who gets to keep the upside
The Misread on Where Value Goes
In 2011, Marc Andreessen published “Why Software Is Eating the World” in the Wall Street Journal. He was right. Amazon ate retail. Netflix ate video. Spotify ate music. For fifteen years, the playbook was clear: go asset-light, move fast, build in software. Physical assets were liabilities. Code was the moat.
Then, in February 2026, more than $1 trillion in market capitalization was wiped from the SaaS sector in what investors started calling the “SaaSpocalypse.” The reason: AI was now eating software. The moat Andreessen spent fifteen years championing had a leak in it, and the market repriced accordingly.
Most conversations about AI investment focus on who’s building at the frontier; Anthropic, OpenAI, and the infrastructure companies feeding compute to both. That’s a legitimate category of value creation, and I’m not arguing against it. What I am arguing is that the second-order conversation, what happens to the rest of the software economy, is being badly underestimated.
The trigger was a market realization that productivity gains from agentic AI were accruing to end users and AI model providers, not the software vendors sitting in between. The per-seat licensing model, the engine that powered a decade of software valuations, suddenly looked like a rounding error in a world where autonomous agents can do the work of ten seats without adding one. This wasn’t a panic sell. It was a structural repricing.
The release of autonomous coding agents in late 2025 demonstrated that software can be built at a fraction of its former cost. That lowers the barrier to entry for new products while simultaneously eroding the defensive moat of every incumbent that didn’t own something beyond code. Investors are now explicitly sorting companies into those with “assets and positions that accumulate value over time and can’t be conjured from scratch” versus those that can be replicated overnight. The former category is winning. Institutional money is rotating into what Seeking Alpha has started calling HALO stocks — “heavy assets, low obsolescence” — and away from software-only plays that suddenly look fragile against the capabilities of any frontier model.
If you’re a software business without proprietary data, regulatory moats, or physical infrastructure, you are in a race you cannot win. The question is just how long it takes to lose it.
The Claude Problem
I’ll be direct about something that might seem odd coming from someone whose company runs partly on AI tools: I think most SaaS businesses face legitimate extinction risk, and I think many of them will simply become a feature of a Claude conversation rather than a product anyone pays for on its own.
That’s not pessimism, it’s pattern recognition. The companies most at risk are those selling point solutions for tasks that a general-purpose AI can now do in response to a single prompt. Summarization tools, basic analytics dashboards, lightweight workflow automation — these aren’t products anymore, they’re prompts. The companies that survive will be those where the software is downstream of something harder to replicate: a proprietary data asset, a regulated workflow, a physical process that requires a body in a room.
And here’s the thing that grounds the whole thesis: Claude cannot create a physical asset. Not yet, anyway. It can write the contract, build the model, design the deck; but it cannot run the event, fill the arena, or make 15,000 people feel something at the same time. That irreducibility is exactly what makes physical, experience-based businesses interesting right now.
Ari Emanuel has been spending billions acquiring live sports and entertainment assets on this exact logic. His argument: as AI-boosted productivity makes digital content increasingly cheap to produce, demand for live experiences; concerts, sporting events, things that can only happen once, in real life — must rise, because the supply is fixed. That’s the scarcity premium. Sports franchises, by definition, have a finite supply. There are only so many legitimate promotions, only so many fighters, only so many licenses to run events in a market. Demand for all of it goes in one direction.
The global sports industry generated $521 billion in 2024. It’s growing at 8% annually. Funds like RedBird Capital have built their thesis’s around sports media and live event ownership specifically because they see live content as the anchor of a media portfolio in an AI-abundant world.
The argument isn’t that AI won’t touch sports, it will, but rather that the core asset, the live competitive event with real human stakes, remains AI-proof in the way that matters: it can’t be synthesized.