Good morning,
This week we discuss how AI agents broke two industries in a single week — and the insurance sector's turn came faster than anyone expected. Plus India AI Impact Summit, Elon Musk “new” vision for Orbital Data center and more.
It’s reprice time
Act I: Software gets repriced. On Feb 5th, Anthropic released Opus 4.6, and swiftly highlighted its impact on finance work notably by updating their Cowork product - autonomous partner, capable of reading, writing, and organizing files directly on your computer to complete multi-step tasks. compan
The same day, OpenAI launched GPT-5.3 Codex (an upgrade to its coding agent) but also Frontier “a new platform that helps enterprises build, deploy, and manage AI agents that can do real work.” Early customers include Uber, State Farm, and Cisco.
The market's verdict was swift: roughly $550 billion in market cap was erased from software companies in 2 days. ServiceNow fell 28% YTD, Salesforce 24%, Intuit 36%, Thomson Reuters 31%. The average forward earnings multiple for software companies collapsed from roughly 39x to about 21x in just a few months.
The thesis is simple: if an AI agent can do the work of 10 people using a SaaS tool, the per-seat pricing model that underpins the entire software industry is broken. Analysts are calling it the shift from "software-as-a-service" to "service-as-software" — you don't buy tools anymore, you buy outcomes.
Act II: Insurance distribution gets repriced. Then, on February 9, the disruption came for us. Two AI-powered apps went live directly inside ChatGPT. One, from digital insurer Tuio and infrastructure firm WaniWani, lets users get a real-time, bound home insurance quote through a natural conversation. The other, from Insurify, compares motor policies side-by-side in the chat window. No website. No aggregator. No broker.
The S&P 500 Insurance Index posted its worst single-day fall since October, dropping 3.9%. Willis Towers Watson lost 12%, Arthur J. Gallagher 9.9%, Aon 9.3%. We dropped ~4%.
The Takeaway: These two selloffs are the same story. AI agents are disintermediating any business built on humans using tools; whether that tool is Salesforce or a broker comparison site. For us, this means two things. First, many commercial clients are SaaS companies or depend on SaaS, this repricing will hit their balance sheets and risk profiles. Second, AI is no longer a backend efficiency play for our industry; it is actively disintermediating distribution. The "point of sale" is shifting from aggregator websites and human brokers to conversational AI agents. This could have many implication on our business model, but at the moment one thing is for sure: the first one of the global player to build distribution via GenAI will have a great advantage.
Competitor Watch: State Farm beta test OpenAI Frontier
State Farm is an early OpenAI Frontier customer. The insurer is among the first wave of companies deploying OpenAI's new enterprise agent platform. Details are thin, but the signal is clear: a top-3 US insurer is integrating AI agents into its operations at the platform level, not just as a bolt-on tool.
ElevenLabs launches the first insured AI agent. The AI voice company (used by 75%+ of the Fortune 500) became the first to secure an AI-specific insurance policy covering the actions of its voice agents - meaning if an AI agent gives a customer wrong information, the enterprise is covered.
More top news
The $650 Billion Capex Arms Race: in the last newsletter, I talked about Microsoft ($140B) and Meta ($125B) impressive expected CAPEX for 2026. Since then, Alphabet ($175-185B) and Amazon ($200B). Meaning that from those four companies alone we should expected about $650 billion capex in 2026. Considering that this picture is missing the like of Nvidia, Apple, OpenAI, Anthropic… it is not unlikely that the actual industry AI spending will reach the trillion mark in 2026.
The India AI Impact Summit kicked off today in New Delhi, acting as a massive convening for the Global South. With leaders like Sundar Pichai and Sam Altman attending, the summit heavily focused on "Sovereign AI"—the idea that nations must build their own localized data centers and models to avoid reliance on the US. India is expected to launch its own sovereign large language model.
A $1.25 Trillion Space Merger. In a move to save the financial of xAI and prepare for a potential IPO, Elon Musk combined his aerospace company SpaceX (valued at $1T) with his artificial intelligence startup xAI ($250B). Musk stated that the move was to prepare for data center to go to space (then on the moon) “in the next 3 years”. While this might sound crazy, it is an option that has been increasingly consider at a solution to the energy bottleneck on earth as space would provide amazing exposure to solar energy. Google is already exploring the idea with Project Suncatcher; Jeff Bezos’ space company, Blue Origin, too; and some startup like Starcloud are already planning launches.
ChatGPT gets Ads: OpenAI began testing ads in ChatGPT, pricing them at a staggering $60 CPM (cost per 1,000 views)—roughly triple Meta's typical rates. The ads are dynamically matched to the context of your conversation. This was the perfect moment for Anthropic to release an Anti-Ads in AI campaign, aired on Super Bowl Sunday!
The International AI Safety Report 2026 — authored by 100+ experts led by Turing Award winner Yoshua Bengio — found that over 700 million people now use AI weekly, some models can detect when they're being tested and alter behavior, and AI has been used in real-world cyberattacks by state-linked groups. Nature published an editorial calling 2026 "the year the world comes together for AI safety."
Tool to try: “Vibe Coding”
Let me finish this newsletter by advising you experience one of the biggest shifts in tech right now: "vibe coding.”
Never touched code before? No problem. Vibe coding means you no longer write code manually; instead, you describe what you want in plain English, and the AI builds the software for you. It's moving the bottleneck from programming skill to strategic vision.
You can find many tools to experiment with it. For the less technical people: Lovable, Replit, Vercel or Bolt are great option. If you want to get your hands dirty: OpenAI Codex, Claude Code, Cursor, GitHub Copilot or Windsurf will enable your most extravagant ideas.
Not sure why it can be relevant? As professionals, you often need to visualize concepts, dashboards, or new customer portals. With tools like v0 or Lovable, you don't need to wait weeks for IT or design teams to build a proof-of-concept. You can literally "talk" a prototype into existence in an afternoon.