AI ROI Series | EP27 | Meta Spent $100M+ in 28 Days on Claude Code. Then It Cut Half the Users.
How do you measure AI ROI and AI spend? Paul Brzozowski, Co-Founder of Olakai, the Enterprise AI System of Record, shows the math. Meta reportedly spent more than $100 million on Claude Code in a single 28-day stretch, then cut its internal Claude Code users from about 60,000 to about 30,000. Microsoft cut the monthly AI budget for many of its cloud and AI engineers from about $100,000 to about $10,000. This episode breaks down what happened, why both companies made the right call, what it means for Anthropic, how OpenAI just halved the value of its $200 Pro plan, and how enterprises should measure AI spend and AI ROI before a vendor makes the decision for them. If you sign off on an AI budget, or you are the one who has to defend it, this one is for you. The numbers are directional, so check my math. Chapters 0:00 Meta spent $100M+ on Claude Code in 28 days 0:54 Who this episode is for 1:20 What you'll get in the next ten minutes 1:47 What actually happened at Meta and Microsoft? 2:22 Is Meta pulling back on AI? 3:03 Why did Microsoft cut, and what is it using instead? 4:08 Does this hurt Anthropic? The math 4:51 How did AI pricing just change? The $200 Pro plan explained 6:43 How do most enterprises measure AI spend today? 7:35 How do you measure AI properly? Capture, Structure, Act 8:22 What does an AI FinOps and AI ROI report look like? 9:30 Why you need an Enterprise AI System of Record 10:10 The one line to take to your board 10:36 Want one of these reports on your own data? Key numbers Meta: more than $100M on Claude Code in 28 days; internal users about 60K to about 30K; 2026 capex guidance $130B to $145B, roughly double 2025. Microsoft: per-engineer monthly AI budget about $100K to about $10K for many cloud and AI engineers; projected internal Claude spend (at least $1B a year) cut by more than a third. Anthropic: $65B revenue run rate in July. Microsoft about 1.5% of it, Meta about 2% annualized, together about 3.5%. OpenAI Pro $200: 20x Plus down to 10x Plus. That is $10 per Plus becoming $20 per Plus, at the same $200. Benchmarkit survey of 372 enterprises: about a quarter missed their AI cost forecast by more than 50%. The reports shown were built from the Olakai demo environment on synthetic data. Want one of these reports on your own data, or a look around our demo environment? DM me on LinkedIn: [your LinkedIn profile URL] Learn more about Olakai, the Enterprise AI System of Record: https://olakai.ai Your AI is an investment. Let's measure it like one. Sources: The Information (October 5, 2026) via trade press; Meta 2026 capex guidance; Microsoft Build 2026 and the Microsoft Copilot launch (September 25, 2026); CNBC on Anthropic's run rate (August 17, 2026); OpenAI Help Center on ChatGPT Pro allowances; The Register on Anthropic Enterprise billing (April 16, 2026); Benchmarkit for Mavvrik, 2025 State of AI Cost Management. #AIROI #AIFinOps #EnterpriseAI #ClaudeCode #Anthropic #Meta #Microsoft #OpenAI #Olakai Tags: AI ROI, AI FinOps, enterprise AI, AI spend, AI budget, Claude Code, Meta Claude Code, Microsoft Claude, Anthropic revenue, ChatGPT Pro 200, AI cost forecasting, AI governance, enterprise AI system of record, Olakai, measure AI ROI
