June 25, 2026, (Inside AI) — Corporate spending on artificial intelligence software and models may reach $680 billion next year, according to a new Reuters Breakingviews podcast. That staggering figure is now forcing CFOs to rethink how they budget for AI tools from providers like OpenAI and Anthropic.
The Viewsroom podcast, hosted by columnists Aimee Donnellan and Una Galani, examined whether volatile pricing will benefit the biggest AI labs or push companies toward cheaper alternatives. The discussion comes as enterprises face unpredictable costs for using large language models and other generative AI services.
Donnellan and Galani debated if new pricing strategies from major vendors can lock in long-term enterprise clients. They also questioned whether sticker shock will accelerate adoption of open-source models or smaller, specialized systems that reduce per-query expenses.
No direct quotes from the podcast were provided in the source material, but the episode description frames a core tension: AI giants are racing to monetize their technology, yet many corporate buyers are already seeking restraint. This dynamic could reshape the competitive landscape within months.
Pricing Pressure Tests Enterprise Loyalty
The $680 billion projection covers spending on both software and models. It reflects how deeply AI has penetrated business operations, from customer service chatbots to complex data analysis. But the figure also masks sharp fluctuations in actual costs per task.
Some companies report that AI inference costs can swing by 30% month over month. That unpredictability makes annual budgeting difficult and irritates finance chiefs accustomed to stable SaaS contracts. One solution gaining traction is multi-provider routing, where systems automatically select the cheapest model for each request.
Anthropic and OpenAI have both introduced tiered plans with volume discounts. Yet the podcast suggests such moves may not be enough. As Galani and Donnellan noted in related columns, the AI frenzy contains echoes of past tech bubbles where exuberance outpaced practical economics.
Historical context matters here. The reference to Galbraith’s bezzle—the inventory of undiscovered embezzlement that builds during speculative booms—hints at hidden financial risks. When costs are opaque and rising fast, companies may overinvest without realizing the true expense until later.
Efficiency Seekers Gain Ground
Smaller AI firms and open-source communities stand to benefit if enterprises pivot toward cost control. Models from Mistral, Meta, and others can run on-premises or in private clouds, offering predictable pricing. Fine-tuning these models for specific tasks often yields performance close to larger commercial systems at a fraction of the cost.
The podcast also touched on the AI wealth carve-up, a topic explored in another Breakingviews piece. That analysis argues that distributing AI’s economic gains fairly remains an urgent, unsolved problem. If only a few vendors capture the $680 billion, market concentration could stifle innovation and keep prices high.
Donnellan, who covers pharma and consumer goods, brings a cross-industry lens to the debate. Her experience tracking cost pressures in regulated sectors informs the view that AI spending cannot grow unchecked. Galani, based in Hong Kong, adds perspective on how Asian markets are adopting cheaper AI alternatives faster than Western peers.
Listeners can find the full episode on Apple, Spotify, or the Reuters app. The hosts emphasized that their opinions are their own, not those of Thomson Reuters. The podcast is part of a series that regularly challenges prevailing market narratives.
For now, the message from Breakingviews is clear: the AI cost surge will force a reckoning. Whether that means a shift to leaner models, stricter vendor negotiations, or entirely new procurement frameworks, the era of blank-check AI spending may be ending.
