Latest from AI Roller Coaster: Disruptive Nausea
That funny feeling when everything's changing
AI continues to advance at an unprecedented pace and I find myself getting a funny feeling in my stomach whenever I read about the roller coaster. Some here's me trying to connect the dots from the AI universe (so you can save on the Dramamine), where GPUs are as valuable as gold, startups are sprouting like weeds, and everyone’s trying to figure out if their job will soon be replaced by a really smart toaster. Buckle up—this ride’s only getting started!
AI Infrastructure: The Crowded Road Ahead
We’ve all heard that AI infrastructure is hot. I mean, the cloud is packed with GPUs, foundation models are all the rage, and venture capitalists are throwing money around like it’s candy. But guess what? This space is getting so crowded, it’s starting to look like a Taylor Swift concert. It’s possible that the market leaders have already emerged, with the cost of infrastructure services, like GPU hourly rates, potentially reaching their peak (H100s, for example). As value shifts toward consumer and business applications, the pace is accelerating far faster than in previous tech cycles. The fabled “trough of disillusionment” for AI infrastructure seems near but perhaps shallow given the warp speed of evolution. New infrastructure investments require a unique angle or untapped niche (or gobs of money, if you have it!).
The Explosive Growth of AI Observability and Eval Startups
AI observability has quickly become one of the hottest investment areas. Startups like Braintrust, Galileo, Patronus, and MaximAI are securing significant funding, focusing on helping companies evaluate and monitor AI models. These startups are capitalizing on the fact that while we have tons of fancy open-source ML tools, figuring out if your AI model is actually working is harder than it sounds. (Spoiler alert: sometimes it’s not.)
The workflows are vastly different from existing software practices and everything needs to be re-wired. I'm curious to see how enterprises and consumers evaluate the value they get from AI tools, both at purchase time and during ongoing usage. I would expect we are still in the first innings of this game where we don't even know how to measure the score.
AI Employees: Because Who Needs Human Co-workers Anyway?
Remember that viral meme of an “AI version” of every function in a company? It’s not a joke anymore. Engineers and salespeople are top targets for AI startups because, let’s face it, they’re the largest headcount at most tech companies. These roles, often generalist in nature, can benefit from AI tools that help streamline processes from end to end. Human employees will likely need to own more end-to-end outcomes and become more generalist. A "build" employee can generate product briefs using ChatGPT, generate designs using Vercel's V0, write and deploy code using Cursor, and evaluate the results too - for those of us with a DevOps mindset, we are unleashing the mother of all Shift Left.
The shift toward AI-driven workflows could create a “barbell effect” in knowledge work, where highly specialized skills and strong generalist capabilities are rewarded. Multiple job functions might get merged with AI tools providing function-specific workflows.
Empire Strikes Back: SaaS Giants Embrace AI
AI startups thought they’d have the space all to themselves, but the incumbents had other plans. The big SaaS players are moving faster than we expected (I know, shocking). They’ve slapped AI layers on top of their existing offerings like they’re redecorating a room and now, voilà, they’re “AI companies.” For example, Qualified.com took their SaaS products, gave them a quick AI makeover, and relaunched them as Piper AI SDR. It’s the SaaS equivalent of putting on a new outfit and cosplaying an AI company.
Why are they doing this? Simple. They already have distribution, customer relationships, and more data than any startup could dream of. Startups will have to think bigger, faster, and—let’s be honest—weirder to beat them at their own game. So, how disruptive do you have to be to outsmart the giants? Very. Like, “let’s reinvent the entire business model” disruptive. No pressure, though.
AI-Native Innovation: Time to Rethink Everything (Yes, Everything)
If you think the future of AI is just slapping an “AI-powered” sticker on an old product and calling it a day, think again. The real disruptors are still in the making, and they’re going to do more than just toss AI into the mix—they’re going to completely reinvent the way businesses operate. (Think of it as the difference between putting lipstick on a pig versus breeding a whole new species of super pig.)
Consider this: Will we still have salespeople running around making cold calls, or will AI handle that? Will onboarding new employees still take months, or will AI plug into their brains (metaphorically, of course) and get them up to speed in weeks? The truly AI-native companies will be the ones that rethink the entire user experience and business model from the ground up. And those companies? Well, they’re still a few years away. But when they get here, they’re going to blow our minds.
Like NotebookLM - last weekend, I fed it my kid’s second grade math homework, and it read out a podcast - how cool is that?
AI P&L Spoiler Alert: The Economics Are Changing
If you thought AI companies would have the same cushy gross margins as SaaS, think again. While traditional software companies have been enjoying 95% gross margins (hello, easy money), AI-first companies are going to see something a bit different. AI models are expensive to run, and if companies start charging for outcomes instead of just software licenses, their P&Ls will look… well, let’s just say “less cozy.”
Will these companies look to international expansion to lower costs? Maybe. Will they blend services with software to optimize outcomes? Probably. One thing’s for sure: if you’re a VC, you’ll need to adjust your Excel sheets and rework your margin expectations.
In conclusion, the AI world is in full-on transformation mode. Whether you’re a founder, engineer, or investor, the rules are changing fast—and if you don’t keep up, you might find yourself left behind by the bots. So, how are you thinking about building products in this crazy AI-first landscape? Drop me a line—I’d love to know how you’re staying ahead of the curve (or at least trying to!).
References:
1. Sequoia Capital: Generative AI Act Two
2. Eugene Cheah on the GPU Bubble
3. The Gross Margin Imperative in the Age of AI – Lightspeed Venture Partners