OpenAI Set to Jump from $13B to $100B - Are We Cool With This?

We need to talk about OpenAI's revenue projections the way you'd talk about a friend's clearly doomed startup pitch. With love, but also with math.

The company is reportedly forecasting revenue of $100 billion by 2028 or 2029—up from $13 billion in 2025. That's not a growth curve. That's a religious experience. According to Epoch AI, only seven US companies in the past fifty years have scaled from $10 billion to $100 billion in a decade or less. OpenAI wants to do it in three.

The fastest anyone's actually pulled this off? Seven years, achieved by both Tesla and Meta. Google—OpenAI's closest software analog—needed ten full years. Even pandemic darling Moderna topped out around $40 billion and has yet to breach triple digits. For context, OpenAI's plan would require growing faster than companies that literally put cars in space or connected three billion humans.

The Infrastructure Trap

Here's where the numbers get uncomfortable. OpenAI has locked itself into massive infrastructure commitments with Nvidia, AMD, and Broadcom—the kind of deals you make when you're confident your scaling laws apply to revenue as reliably as they do to model performance. But here's the thing about hardware: you can't refactor a data center. You can't pivot a chip fab. You can't patch your way out of overcommitment.

The infrastructure costs for training and serving frontier models are growing faster than most revenue projections can support. OpenAI isn't just betting it can 10x its revenue—it's betting it can do so while maintaining margins that justify infrastructure spending that compounds annually.

If they miss these targets—and to be clear, missing could mean hitting "only" $40 billion in revenue, which would still be extraordinary—the ripple effects hit everyone. The entire AI supply chain is now woven through with assumptions about OpenAI's trajectory. Nvidia's projections, cloud providers' capacity planning, enterprise software vendors' AI strategies—all of it implicitly prices in OpenAI's continued exponential growth.

The Revenue Reality Check

OpenAI expects about half its 2028 revenue to come from ChatGPT subscriptions. The other half? New ventures in advertising, shopping, and automating knowledge work. Let's unpack that math.

To hit $50 billion in subscription revenue, OpenAI would need roughly 200 million paying subscribers at $20/month—or some combination of higher prices and enterprise deals that nets out the same. For reference, Spotify took 15 years to reach 220 million paid subscribers. Netflix needed two decades to hit 230 million. Both offered content libraries that cost billions to build but served an infinite number of users for essentially free after creation.

AI inference doesn't work that way. Every query has a marginal cost. Every response burns tokens. Every interaction with o3 or future models requires compute that scales with complexity. OpenAI isn't selling access to a fixed library—it's selling real-time manufacturing of unique outputs. The unit economics are fundamentally different, and not in a favorable direction.

The New Ventures Gamble

The other $50 billion is supposed to come from advertising, e-commerce, and workflow automation. These aren't just new product lines—they're entirely different businesses requiring different infrastructure, different partnerships, and different competitive dynamics.

Advertising means competing with Google and Meta's targeting machines, built over decades and trained on behavioral data OpenAI doesn't have. Shopping means logistics, payment processing, merchant relationships—basically building Amazon but with an AI wrapper. Workflow automation means displacing established enterprise software vendors who've spent years embedding themselves into corporate IT stacks.

Each of these markets is governed by different scaling laws than model training. Network effects matter more than parameter counts. Distribution beats raw capability. Integration depth trumps inference speed. OpenAI's advantage in language models doesn't automatically translate to advantage in ad auctions or supply chain management.

What This Means for the Rest of Us

If OpenAI actually delivers on these projections, it rewrites what's possible in enterprise software and validates the entire AI infrastructure buildout we're witnessing. Markets will respond accordingly—expect every AI company's Series B deck to include "OpenAI comparable growth" slides through 2027.

But if they fall short—and history suggests most companies announcing revolutionary growth eventually meet gravity—we're looking at a reset moment for AI expectations. Not a collapse, but a recalibration. The technology remains transformative; the timelines and TAM (total addressable market) assumptions just get more realistic.

For marketers and growth leaders, this matters because your AI vendor strategies shouldn't be built on OpenAI's most optimistic projections. Plan for the technology to work. Plan for costs to remain high. Plan for competition to intensify. Don't plan for one company to quintuple the market while also quintuple its revenue while also maintaining margins.

The models will keep improving. The applications will keep expanding. But scaling laws—the real ones, the ones governed by customers and cash flow rather than parameters and compute—still apply. Even to OpenAI.

Ready to build AI strategies based on reality rather than revenue projections? Winsome Marketing's growth experts help you navigate AI adoption with economics that actually work.