The Wall Street Journal recently spotlighted the scale of today's AI investment boom and growing doubts about whether it will pay off. In 2024, global AI infrastructure spending hit $235 billion, with projects like OpenAI's "Stargate" envisioning $500 billion over four years. OpenAI—the industry's most prominent player—has about 700 million weekly users but is on track for only $13 billion in annual revenue, or roughly $17 per user. This gap between significant investment and modest monetization lies at the heart of AI's business challenge.
Revenue Illustration
Using OpenAI as a case study highlights the math problem. The company has reportedly committed to paying Oracle around $60 billion annually for servers and floated plans for data-center investments approaching $1 trillion. Against that backdrop, $13 billion in annual revenue is a drop in the bucket. Optimists point to projected growth in OpenAI's paid subscribers—currently around 20 million and expected to rise sharply—but even rapid adoption won't close the gap unless per-user revenue climbs substantially (a tall order when many competitors continue offering powerful AI tools for free).
For the economics to work, users would need to value the product much more highly. Say 100 million users subscribed at $30 per month (or $1/day, about half the cost of a cell phone plan), OpenAI could generate $36 billion annually—a figure that begins to cover multi-billion-dollar infrastructure and R&D costs. But at $7.50 per month ($0.25/day), the same base yields just $9 billion annually, which doesn't scratch the surface against a single $60 billion compute contract.
Unlike social media, where costs flatten as scale grows, OpenAI's expenses rise with usage. Every new customer brings additional compute, data, and energy requirements—meaning profitability only arrives if users are willing to pay much more than they do today. As we alluded to in A Fundamental Approach to AI Profitability, the question doesn't only hinge on explosive user growth but whether the product is valuable and sustainable enough to command sufficient, long-term recurring payments.
History's Warning Signs
While comparisons to the dot-com bubble can be overused or overstated, they remain instructive for dealmakers. In that era, many companies showing impressive user growth but weak monetization were primarily acquired or funded on the basis of "eyeballs" rather than cash flow. When projected cash flows failed to materialize, investors and buyers faced securities litigation and shareholder suits. So deal terms evolved, and parties more frequently implemented earn-outs, material adverse effect (MAE) clauses, and indemnities within their transaction agreements to help allocate risk between optimistic projections and financial realities.
Tools for Today's Acquisition Agreements
For dealmakers evaluating AI targets, many of the same contractual protections can be applied today:
- Earn-outs tied to realized revenue and margins: Structure contingent payments around actual monetization, not user counts or engagement metrics.
- Robust MAE clauses: Capture material changes in monetization assumptions, regulatory actions, or infrastructure cost escalations.
- Purchase price adjustments: Reflect the volatility of compute and power commitments in working capital and other adjustments.
- Comprehensive indemnities: Require seller coverage (and/or R&W insurance) for misstatements or omissions about key risks, such as the stability of revenue streams or long-term obligations under compute and chip supply contracts.
The Bottom Line
Debates about whether trillions in AI spending is too much will play out over years. But acquisition agreements allow dealmakers to protect themselves today. The dot-com era serves as a reminder that transformative technology alone does not guarantee profitable business models. For buyers and investors in today's AI transactions, the lesson is clear: structure deals so that the upside is shared, but the downside risk is contractually managed.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
