Last week’s announcements from OpenAI marked a turning point — not just for their business but for how we think about generative AI economics more broadly. At Dellecod, we’ve been reflecting on the implications for companies building on top of, alongside, or even orthogonally to this ecosystem.
This isn't a story about a product release. It’s about the scaffolding that will support — or constrain — the future of commercially viable AI.
Three closely timed moves told a cohesive story: surging revenue tied to GPU availability, a low-cost $8/month offering positioned for scale, and a strategic embrace of advertising. Each of these by itself might have raised a few eyebrows. Together, they form a blueprint for OpenAI’s long game: grow user base, build infrastructure capacity, monetize efficiently, repeat.
At first glance, the numbers are staggering. From $2B in revenue last year to an expected $20B next year — a tenfold leap in under 24 months. But buried in that growth is a more revealing truth: the link between revenue and compute. OpenAI’s CFO, Sarah Friar, laid it out clearly. The more GPUs they access, the better their models become. The better the models, the more users engage. And the more usage, the more revenue to invest back into infrastructure. It’s a modern flywheel, running not on distribution or demand, but on silicon and electricity.
This puts pressure not just on model development but on global chip supply. Their partnership with Cerebras, reportedly worth $10B, is more than a vendor deal — it’s a bet that alternative chip architectures can unlock exponential gains in inference cost efficiency. And if it works, they’ll be able to keep scaling users without blowing out their loss sheet. Because make no mistake: losses are part of this strategy. Significant ones. Token and inference costs aren’t falling fast enough to offset user growth, and GPU rental costs remain stubbornly high. Their current model likely loses money per user — possibly up to $20B annually.
Which brings us to the $8 offering — ChatGPT Go. Ostensibly a solution for global affordability, it may also serve as a test of tier-based performance degradation. Key features are unlocked (including the newer GPT-4.5 Instant), but heavy costs are absorbed. Why? Because it's a user acquisition engine. A data funnel. A lock-in device. The economics don’t have to work now. What matters is reach and retention at scale across regions that haven’t historically been able to afford access.
Then there’s the ads.
It’s hard to overstate how much of a cultural shift this is for OpenAI. Ads in AI products have long provoked backlash — both from purists worried about model integrity and users concerned about privacy creep. But OpenAI seems to have studied the history of monetization closely: all dominant platforms, from Facebook to YouTube, eventually embraced ads to fund their free experiences. If executed well, this move could be one of the largest revenue levers since the App Store.
Here’s what’s interesting: they’re not expecting to monetize at Meta’s level right away. Instead, they’re modeling value at a fraction of it. At just 9% of Meta’s average ARPU, OpenAI could generate $5B annually — essentially monetizing goodwill and global usage at low marginal cost. If they ever close that gap, even partially, their ad revenue could eclipse all other lines.
Taken together, this is a company transitioning from technical breakthrough to economic durability. For much of the past year, success meant "better models." Now, it’s more nuanced. It’s about pricing strategies that reward volume. It’s about tiered access that matches cost curves. It’s about preserving user trust despite increasing complexity in the business model.
As people building our own AI systems and tooling, we take away a few things.
First, the substrate matters. Compute isn’t a commodity — it’s the new topography of competition. If you're building products that rely on inference, don't treat capacity as a given. Build a model around how much compute you can sustainably afford, and plan for volatility.
Second, access isn’t just about openness — it's an ecosystem question. The Go plan is a masterclass in using pricing strategy for geographic expansion. But it also positions OpenAI as a data aggregation engine. Any product built on top may be shaped as much by what users do inside ChatGPT as what they do in your app.
Third, new monetization models will shape user expectations fast. Ads in AI may have once been taboo, but they’ll normalize quickly if done with care. OpenAI is betting that users value access more than purity, as long as there is transparency. That opens — and closes — strategic paths for others. If you're not charging for usage, ask yourself how long your runway is before you're asked to.
Finally, the sequencing of OpenAI’s news is telling. The revenue was placed first. Ads came last. It’s a signaling story, meant not just to avoid backlash but to anchor what everything else depends on: model quality and scale. Without those, monetization is moot. With those, almost anything is possible.
We're watching closely — not just what this means for OpenAI, but what it tells us about the future of AI-native business. It's moving fast, but the direction feels increasingly clear. Whoever can win and sustain the most usage — across markets and use cases — will define the next chapter.
Quietly and steadily, OpenAI is building the infrastructure to do just that.