OpenAI spending hit $34 billion last year ahead of planned IPO, FT reports

OpenAI reportedly spent $34 billion last year to strengthen its position in the AI market ahead of a planned IPO. Audited figures cited by the Financial Times show about $19 billion of R&D spending in 2025 and nearly $6 billion on sales and marketing, plus other costs. The disclosure underscores the capital intensity of competing in artificial intelligence, but the article contains no direct market reaction or new operational guidance.

Analysis

The more important signal is not the absolute spend, but the implied willingness to keep subsidizing model quality long before monetization is fully proven. That tends to compress the competitive gap in the near term: the leaders can absorb losses to lock in developer mindshare, while smaller frontier labs may be forced into a capital-markets reset within 12-18 months if they cannot match training cadence and inference economics. For public-market comps, this is less about a single company and more about the bar rising for every AI entrant that needs continued external funding.

Second-order beneficiaries are the picks-and-shovels layers with scarce capacity: advanced semicap, high-bandwidth memory, networking, and datacenter power infrastructure. The spend pattern suggests the bottleneck is shifting from model research to deployment at scale, which favors suppliers with pricing power and long backlog visibility rather than software names trying to defend margins. The risk is that as capex intensity normalizes, the market may rotate away from “AI software optionality” toward the physical enablers that can actually convert demand into revenue.

The key contrarian point is that aggressive spend ahead of an IPO can be read as confidence, but it also raises the hurdle for the listing: public investors will likely underwrite a path to operating leverage, not just growth. If investor appetite weakens, the company could be forced to slow hiring and marketing within two to three quarters post-IPO, which would ripple into cloud spend, contractor demand, and experimental AI ecosystem funding. That creates a tactical window where hype remains high but the secondary effects are already showing up in lower-quality funding markets.

Key Decisions for Investors

• Long semicap/picks-and-shovels basket over AI software beta for 3-6 months: prioritize ASML, AMAT, LRCX, and NVDA suppliers; thesis is that capex intensity and HBM/network demand stay elevated even if model monetization wobbles.
• Pair trade: long infrastructure beneficiaries (VRT, ETN, SMCI on pullbacks) vs. short unprofitable AI application names with high burn and no pricing power; enter on any post-earnings AI exuberance, target 15-25% relative performance over 2 quarters.
• Avoid chasing pre-IPO AI private secondary exposure unless priced at a discount to the last round by at least 20-30%; the risk is a financing overhang once public investors force operating leverage discipline.
• For public market event risk, buy 6-9 month put spreads on the most crowded AI software proxies if they trade at >15x forward revenue; the trade works if the market starts differentiating between model spend and monetizable workflow adoption.
• Watch cloud and datacenter power names for confirmation: if order books extend, add to long infrastructure exposure; if management commentary turns cautious, cut high-multiple AI software longs first.