OpenAI on Track to Double Revenue Ahead of IPO

OpenAI is reportedly on track for $40B+ annualized revenue, about double its late-2025 run rate, driven by growth in paying ChatGPT users, enterprise adoption, and the coding assistant Codex. The article flags that extremely high computing/power costs remain a key constraint on the outlook, tempering the growth story despite strong momentum.

Analysis

The market will likely read this as confirmation that AI demand is real, but the more important mechanism is value transfer: incremental monetization is still being pulled through a very expensive compute stack. If that revenue requires outsized inference spend, the economic winners are the suppliers of accelerated compute, networking, and datacenter power, not necessarily the model layer itself. That favors semis and infrastructure with pricing power, while keeping a lid on the long-term margins of frontier-model businesses unless usage efficiency improves materially.

Second-order effects matter more than the top-line number. Enterprise adoption and coding assistants should support a broader AI refresh cycle across cloud, networking, and storage, but they also intensify competition among model providers and hyperscalers, which can force lower per-token pricing or richer customer incentives. The risk is that revenue growth is bought with margin dilution, so the next catalyst is not more revenue but proof of gross margin expansion and lower compute cost per unit of usage.

Over the next 1-3 months, the key watch item is whether hyperscaler capex commentary and GPU supply commentary accelerate in response to this demand signal. Over 6-18 months, the falsifier is simple: if OpenAI-like growth does not translate into improving unit economics, the current AI re-rating shifts back toward a narrow picks-and-shovels trade rather than a broad software multiple expansion. That makes this more of a barbell setup than a clean bullish read-through.

The contrarian view is that consensus may be overestimating the durability of margin capture at the application layer. Paying users and enterprise uptake prove willingness to spend, but they do not prove sustainable profitability; in fact, they may imply the opposite if usage scales faster than efficiency gains. If compute stays scarce and expensive, the eventual beneficiaries may be the infrastructure vendors and hyperscalers that control capacity, while pure AI software names face rising customer expectations without matching economics.

Key Decisions for Investors

• Long NVDA and AVGO on any broad AI pullback over the next 1-3 weeks; thesis is sustained accelerator and interconnect demand, with upside if hyperscaler capex guides higher. Falsify on evidence of demand deceleration or meaningful pricing pressure in next quarter commentary.
• Overweight MSFT versus a basket of high-multiple AI application/software names for the next 3-6 months; Microsoft captures enterprise AI demand with better distribution and balance-sheet support, while the article’s message suggests the model layer may struggle to keep all the economics.
• Add ANET as a secondary beneficiary over 1-3 months if AI workloads keep scaling; networking becomes a larger bottleneck when inference traffic grows, and that can expand wallet share even if GPU supply stays tight.
• Avoid chasing broad AI software longs at current multiples until there is proof of gross-margin expansion rather than just revenue growth; use the next earnings season to screen for names with AI revenue but deteriorating incremental margins.
• Watch for a bearish reversal in any of the above if hyperscaler capex or data-center power commentary softens; that would signal the revenue growth is not translating into durable infrastructure demand.