The Information

🚀 AI Deep Dive Episode 7: The Unexpected AI Costs Catching Companies Off Guard What does AI really cost beyond the token bill? EY Global Consulting AI Leader Dan Diasio joins @rocketalignment to explore the unexpected costs of putting AI to work—from retraining employees and building safety systems to planning for outages when agents run critical processes. What does it take to make those investments pay off? And when should companies own a model rather than rent one? Shot at The Information’s AI Agenda Live event at the San Francisco Museum of Modern Art, this conversation also examines the full cost of AI, model routers and why a company’s org chart can get in the way of change. The topics discussed are based on EY Consulting research analysis. Learn more by visiting ( Supported by EY. Chapters 00:00 Intro 02:05 Why AI adoption is difficult 05:26 Culture, budgets and the org chart 07:55 Redesigning workflows around AI 12:21 How much can today’s models already do? 17:10 Multi-agent systems and human judgment 21:54 The full cost of AI 26:11 When to own your AI models 30:51 How model routers work 34:59 Leading change and making agents reliable

AI models can solve complex math problems. So why can't CFOs rely on them for financial forecasting? @anaplan CEO Charlie Gottdiener explains why LLMs still fall short of the accuracy businesses require: "LLMs can't do it. They're not architected to do it. They're language models, they're really good at predicting the next word, not doing arithmetic at scale."

Microsoft is offering voluntary buyouts to 60,000 employees, more than a quarter of its workforce. The Information's @aaronpholmes explains how the offers fit into the company's massive AI spending plans: "There's been sort of a recognition by leadership at the company that Microsoft can probably operate with significantly fewer employees than it has right now."

Broadcom's $63 billion financing arrangement tied to Anthropic raises a bigger question. Is the AI boom becoming too dependent on debt? "Using debt to finance speculative investments. Never a good idea. That's how you get in trouble. That's why everybody's warning about a bubble." — @gilluria, Head of Technology Research at D.A. Davidson

OpenAI’s cheapest model costs 80% less to run than Anthropic’s cheapest offering, but price alone may not win over business customers. Anthropic’s IPO filing reveals just how deeply its models are taking hold at major companies. Read more:

Pro Report: The companies run by The Information’s 50 Most Influential People in Tech are worth more than $34 trillion, roughly equivalent to a year of U.S. GDP. Just four executives—Jensen Huang, John Ternus, Sundar Pichai and Satya Nadella—lead companies accounting for 57% of that value. Read more with The Information Pro:

What if your 50-step workflow can be vastly improved by reimagining it with AI? EY Global Consulting AI Leader Dan Diasio explains why his team starts with the jobs a process needs to accomplish, then rebuilds it around people and agents. “And we can reconstruct that flow of work into something that looks way different than the 50 steps that we had previously.” Supported by EY 🚀 Watch the full episode of AI Deep Dive:

Exclusive: Instacart has held preliminary sale discussions with DoorDash and Uber, potentially setting up one of the biggest acquisitions in grocery delivery. Full story:

Data center insurance premiums could approach $24 billion by 2030 as the AI infrastructure boom exposes developers to soaring penalties, power delays and equipment risks. Full story:

Exclusive: More than a quarter of Meta’s 10 million Muse users are shopping through the AI app, giving Stripe an early foothold in a fast-growing payment channel. Full story:

Anthropic’s IPO filing reveals an 85% gross margin on running existing models, even as it relies heavily on rented computing power. The filing also outlines ways shareholders could eliminate the powers of its unusual AI safety trust. Read AI Agenda:

Anthropic’s pricing power helped it overtake OpenAI in revenue, but its reluctance to extend discounts is now driving some major customers toward its rival. Full story:

Today on TITV: Microsoft is offering voluntary buyouts to 60,000 employees, more than a quarter of its workforce. What does the move signal about the company's priorities? Plus, DoorDash's talks to buy Instacart for around $17 billion, SpaceX and Broadcom's multibillion-dollar AI chip financing deals, Anaplan's enterprise AI strategy and General Compute's $100 million raise.

The Information | TITV | October 27, 2026

RT @Jessicalessin: Kurzweil is replicating himself. Sounds right. Great profile on singularity founder by @AmyDMarcus

Exclusive: Futurist Ray Kurzweil, 78, is building an AI replica of himself that can speak, reason and learn from conversations. He hopes the digital version, called Rai, will continue interacting with people long after his death. Full story:

Today on TITV: Microsoft is offering voluntary buyouts to 60,000 employees, more than a quarter of its workforce. What does the move signal about the company's priorities? Plus, DoorDash's talks to buy Instacart for around $17 billion, SpaceX and Broadcom's multibillion-dollar AI chip financing deals, Anaplan's enterprise AI strategy and General Compute's $100 million raise. 📺 Tune in at 10 am PT / 1 pm ET on:

The AI data center boom is creating enormous new risks for developers and their lenders. A two-month construction delay could trigger $300 million in penalties, threatening projects backed by billions in borrowed money. Full story:

Anthropic generated $3.5 billion in free cash flow on $16.3 billion in revenue in the first half of 2026, reversing a $1.6 billion cash burn a year earlier. Its 85% gross margin on running existing AI models challenges skeptics who questioned the profitability of selling AI access. Read AI Agenda:

Exclusive: Instacart is weighing a sale as its growth trails larger rivals DoorDash and Uber. The company has hired an investment bank to explore its options, though a deal would likely face antitrust scrutiny. Full story:

Exclusive: Stripe’s early lead in AI shopping payments is putting pressure on Visa and Mastercard, which are racing to promote competing technology that preserves their control over transactions. Full story: