AI startups are headed for a pricing cliff, and many founders (and investors) aren’t ready for it. For someone who has lived through multiple tech shifts, you might be getting déjà vu. Major incumbents are releasing AI “agents” and features that reset customers’ expectations to zero (or close to it) on price. The result: today’s AI startups likely won’t grow ACVs (annual contract values) like SaaS companies did, because offering key features for free is becoming the competitive norm. Here’s why this trend is real, and what it means.

OpenAI and others have anchored the price of core AI functionality shockingly low. Remember when ChatGPT launched? OpenAI effectively set the anchor price for a state-of-the-art AI assistant at $0 (free tier) and $20/month for premium. That’s for tech many of us would have gladly paid much more for. When something so powerful is offered at the cost of a few lattes, it drastically undercuts what customers expect to pay elsewhere. In fact, the widespread availability of ChatGPT’s generous free tier drove a huge surge in business AI usage, without many companies spending a dime.

Now they’re pushing further with agent-like systems. (If ChatGPT can plan trips, write code, and analyze data as a sidekick, how much will users pay another vendor for similar capabilities? Likely a lot less than traditional software.) Big tech is training customers to expect AI-rich features for free or bundled in, not as expensive add-ons.

Two hard truths:

1. Commoditization & Skepticism: If your AI product looks like a thin wrapper on GPT, buyers will ask, “Why am I paying you $$ when I can use ChatGPT or an open-source model for free?” Many startups are finding that customers will test your product, then try to rebuild it in-house or wait for a cheaper copycat. The plethora of similar LLM-based tools has already sparked a race-to-the-bottom price war. Lower prices benefit consumers, but make it unsustainable for vendors (margins shrink, and it’s hard to fund R&D).

2. Efficiency Shrinks Spend: Ironically, AI’s superpower is to reduce workloads, and that can reduce software spend. AI products that automate work can eliminate the need for extra seats or tools. In SaaS, the playbook was often “land small, then expand seats to 100s or 1000s.” In AI, the expansion might not come, an AI might do the job of 10 users, so who needs 10 seats? This caps the revenue a single customer generates.

Bottom line: AI startups today just don’t have the pricing leverage that SaaS companies enjoyed a decade ago. Customers have been conditioned to expect lower costs for similar value. And that brings us to a crucial point: many AI features that startups charge for will soon be considered baseline freebies.

This isn’t the first time a tech wave hit a pricing wall. Previous generations of SaaS saw similar “race to free” dynamics once big players and competition showed up. A few examples from the trenches:

• Cloud Storage Wars (late 2000s): Dropbox built a great business selling cloud storage until the giants (Google, Microsoft, Amazon) decided storage would be ultra-cheap or free. Google Drive launched with 15 GB free, dwarfing Dropbox’s 2 GB free offering. Microsoft went so far in cutting prices and boosting OneDrive free limits that analysts said Microsoft might have “killed Dropbox” by making storage essentially a giveaway. The core feature, GBs of storage hit a price cliff. Dropbox had to pivot to focus on collaboration and enterprise features, because basic storage had been commoditized to $0.

• Email Marketing & Marketing Automation (2010s): There was a time you’d pay serious money for email newsletter software. Then freemium changed the game. Mailchimp famously drove adoption by offering a robust free plan (today, 500 contacts and 2,500 emails/month are free). They turned email marketing into a low-cost utility for the masses. Later, all-in-one platforms like HubSpot started bundling email tools into their suite, often with a free tier as well. When HubSpot made its CRM free and included basic email sends, it set the expectation that “simple email campaigns should come standard, at no extra cost.” Standalone email tools had to either move upmarket with advanced automation or compete on price (many did both).

• CRM Software (mid-2010s): SaaS CRM (think Salesforce) used to be a cash cow sold seat-by-seat. Then a new wave of products like HubSpot CRM came in and literally made core CRM features free. HubSpot’s free CRM (launched 2014) was fully functional for an unlimited number of users, an unheard-of move at the time. This flipped the script: if you were a startup building a basic sales CRM, suddenly you couldn’t charge $50/user/month for it, the baseline price had become $0. The only way to make money was layering on premium features or going after niche vertical needs, because the market now saw CRM as something you could start using for free.

In each of these cases, major platform players or a frenetic market crowded with competitors forced prices down to the point that the core functionality became essentially free. Startups had to adapt by finding new value above and beyond that core, or by monetizing in different ways (upsells, enterprise services, etc.).

Sound familiar? This is exactly what’s beginning to happen in AI.

Fast-forward to today. We’re at the early stages of a similar shift:

• Generative AI everywhere: Microsoft is baking GPT-4 into Office and Windows. Google is putting AI into Workspace apps. These incumbents might charge a bit extra now (e.g. Microsoft 365 Copilot add-on), but over time such features tend to get bundled or standard. It’s easy to imagine that what an AI startup sells as a product today (say, an AI writing assistant) is just a built-in feature of Google Docs or Outlook tomorrow, effectively free for the user.

• Open-source & API commoditization: Meta released LLaMA 2 for free use, and a new open-source model seems to drop every other week. If you’re charging high prices for, say, AI image generation or document summarization, there’s likely an open model or inexpensive API that can do 80% of the job. This open-access trend means every new AI capability diffuses into the community fast. Startups don’t have a multi-year window to monetize a unique tech advantage; within months there might be a free alternative or a cheaper API competitor.

• “Agents” as a Service: OpenAI, Anthropic, and others are racing to offer agent-like AI that can perform complex tasks (coding, researching, customer support flows, etc.). OpenAI’s plans (if reports are accurate) even include high-end agents at surprisingly affordable rates, e.g. a “software dev” agent for ~$10K/month, which at first sounds high, but to a Fortune 500 is trivial (and that’s for top-tier; cheaper tiers could be just a couple grand). If OpenAI itself offers a $2K/month agent that handles a big chunk of knowledge work, how could a smaller AI startup charge $200K/year for something similar? The pricing anchor is reset far lower. Even if those exact prices are in flux, the direction is clear: more capability for less cost, year over year.

In short, what AI startups consider their “secret sauce” today might be table stakes (and free) in a year. Any narrow feature that’s just applying a public model to a use-case is vulnerable to this. Founders I speak with are already feeling it: customers ask them “why wouldn’t I just use [ChatGPT or XYZ]?”. It’s a daunting question, especially for early-stage companies seeking differentiation.

So, how do we navigate this harsh reality? A few thoughts for those of us in the arena:

1.Investors: Start preparing to reset your expectations on monetization and growth. The old SaaS playbook of steadily expanding ACV per customer may not apply in AI in the next few years, at least not for companies built purely on current gen AI capabilities. When evaluating AI startups, pay extra attention to product differentiation and user engagement over pure revenue numbers. If a startup has millions of users on a free plan and low revenue, that might be normal in this environment (whereas in SaaS 10 years ago, that’d be a red flag). Diligence should ask: Is this solving a sticky, mission-critical problem? Does it leverage proprietary data or network effects? Because if it’s just a clever UI on top of an LLM, the moat is shallow, and pricing power will be near zero. Be cautious with revenue projections that assume easy upsells or high pricing, if anything, assume downward pressure on pricing until proven otherwise. (Also, brace for business models that monetize indirectly, like advertising or data network effects, in lieu of straight SaaS fees.)

2. Founders: Build for value, not valuation hype. In a world where key features tend toward free, you need to answer: “What will we charge for, and why will people pay?” That might mean focusing on unique data, proprietary algorithms, or integration that a big platform can’t easily replicate. It could mean becoming the best at a specific vertical/task so that your performance justifies paying for. Or adopting a “freemium + upsell” model: give the core AI away free to get distribution (since everyone else is doing it), and monetize premium features (advanced analytics, workflow integration, human-in-the-loop services, custom training, security/compliance, something the big generic tools won’t do out-of-the-box). Also, be ready to iterate on pricing models. Usage-based pricing, outcome-based pricing, hybrid models, all are on the table. The only wrong move is assuming you can charge high prices just because you deliver “AI”; you must deliver business value that stands above what free alternatives offer. (And when you do charge, consider tying it to that value e.g., if you save a company $100K, maybe a $10K/month price can stick. But a $10K price on a generic tool that might save $100K is a much harder sell now.) Lastly, tighten your belts: if you have to offer a lot for free, ensure your burn and infrastructure costs make sense. Margins matter, because price wars kill companies that can’t sustain the fight.

3. Customers (and end-users): Enjoy the benefits but choose wisely. Short-term, this is great news for you, lots of AI tools competing, many free or cheap options, and big players bundling new goodies into products you already use. Take advantage of it! Experiment with features from incumbents and startups alike. But also, be aware: the tool that’s free today might not be around tomorrow if it can’t find a sustainable footing. If you’re adopting an AI solution that’s mission-critical, ask yourself how that company will survive long-term. (If you really love it, consider supporting it in ways that count feedback, paid plans, etc.) We will likely see some fantastic AI products die or stagnate because they couldn’t make money in this environment. Don’t let “free” be the only reason you pick a solution if it’s central to your business, reliability, support, and evolution matter too. And when budgeting, realize that the true cost of AI may shift from “license fees” to things like integration effort, data validation, or premium support.

I don’t share this to be a downer on AI startups, in fact, I’m incredibly excited about the innovation happening. But history is instructive: Every platform shift creates hype, then commoditization, then a new equilibrium. We saw it with storage, with email, with core software features, and we’re seeing the signs with AI right now. The companies that survived those past waves adapted their models, they doubled down on what made them unique and found creative ways to monetize beyond the commoditized core. The same will be true here.

For investors and founders, the message is clear: be realistic about pricing power. The era of easy big-ticket SaaS sales in AI might be short-lived. But those who solve real problems and build moats will still capture value, even if it’s via unconventional strategies.

And hey, sometimes giving away the magic is the best way to become indispensable. Just ask the folks who built fortunes on “free” products that became global standards.