Good morning from the wAIve team in London.

Today, Friday 23 October 2026: A day for very big valuations and rather smaller returns. OpenAI wants $30 billion from investors at a price tag of $1.25 trillion, Barclays is letting 5,000 people go and giving AI the credit, and America's banks have started changing the locks on AI agents that turn up with a customer's password. Meanwhile, only one company in eight can actually point to AI making it any money.

  INDUSTRY  

OpenAI is in advanced discussions to raise $30 billion at a valuation of $1.25 trillion. If confirmed, it would make OpenAI the most highly valued private company in history, surpassing every previous benchmark for a pre-IPO business. OpenAI's current annual revenue is estimated at around $15 billion, meaning investors would be pricing the company at roughly 83x current revenue.

THE PLAY: Competition from Anthropic, Google DeepMind, and DeepSeek is intensifying, with compute costs compounding. Despite staggering growth, from a $24 billion run rate in March to $50 billion in September, OpenAI's finances are heavily weighed down by capex and compute.

  BREAKING  

Barclays announced it is cutting 5,000 jobs, roughly five in every hundred of its 93,000 global staff, with executives explicitly claiming AI productivity gains. Around 1,000 roles have already been lost since January. The bank reported a jump in quarterly profits alongside the announcement. Other banks watching their own AI investments will now face pressure from investors to follow suit.

THE PLAY: This is one of the largest AI-attributed headcount reductions announced by a major financial institution and the first in the UK to be framed so directly. There's a strong AI wind blowing through the banks and fintech, and it's only just beginning.

  RESEARCH  

Morgan Stanley has blocked AI agents from accessing customer accounts using client credentials. The issue emerged as customers began granting AI tools their login details to perform tasks such as portfolio rebalancing and payment authorisation, sometimes without customer permission.

THE PLAY: Most consumer-facing AI agents weren't designed with banking authentication in mind. They use the same username and password flow as a human. The banks' response is to lock the door. The agentic economy needs a security infrastructure that doesn't yet exist at scale.

Chinese AI laboratory DeepSeek has reportedly closed a $15 billion funding round, one of the largest in Chinese technology history, coming just days after the US government ordered federal agencies to remove DeepSeek software over national security concerns.

Why it matters: European companies procuring AI infrastructure now face a choice that is no longer purely commercial. It is geopolitical.

Walmart is introducing technology that allows AI agents to make purchases on customers' behalf, with user-controlled spending limits and transaction approvals. At the same time, Amazon has moved to block AI agents from shopping on its platform. The divergence reflects fundamentally different theories about who, or what, the customer of the future is.

Why it matters: If AI agents become the primary interface through which consumers shop, the platform that agents choose becomes the dominant retailer. This is a strategic decision about whether to compete for human attention or agent preference.

Closing prices, Thursday 22 October (US markets). Not investment advice.

Anthropic's new always-on AI agent that works through your Gmail, Slack, calendar, and over 6,000 connected services, even while you are offline. Best for executives who want an AI that acts proactively on tasks rather than waiting to be asked. Available via waitlist for Claude Pro and Max subscribers. Pro plan starts at $20 per month.

THE PLAY: Sign up for the waitlist today and define one recurring workflow (weekly report compilation or meeting follow-up emails) that you want Duet to handle autonomously. Evaluate whether the output requires your edit or simply your approval.

Canva has released its first proprietary image generation model, available to all users, promising more customisable results and lower costs than third-party models it previously relied on. Best for marketing, communications, and brand teams who need on-brand visuals without a designer in the room. Free tier available; Canva Pro from £13.99 per month.

THE PLAY: Open Canva today and generate three on-brand images for an upcoming campaign or presentation using the new model. Compare the output with your existing visual assets and assess whether your design workflow needs to change.

A Gartner survey of 2,002 chief information officers, published this week, found that only 13 in every 100 companies are generating genuine business value from their AI investments. At the same time, more than half of respondents, 51 in every 100, expect AI to actually increase their operating costs. Yet three quarters of chief executives plan to raise their technology budgets further in the coming year.

Most organisations are spending more on AI, most expect it to cost them more, and fewer than one in seven are seeing real returns.

Organisations that run AI tools without redesigning the processes around them tend to add cost rather than remove it. A language model that drafts emails faster still requires a human to review and send them. An AI that surfaces data insights still requires an analyst to interpret and act on them. The result is that any productivity gain is lost. The organisations that do report returns are treating AI as a chance to redesign process, not a software procurement decision.

What this data also reveals is a failure of board-level oversight. Boards are approving spend based on strategic narrative rather than operational evidence.

Winning organisations identify the highest-friction process in a specific workflow, test whether AI reduces that friction measurably, quantify the saving, and use that evidence to justify the next investment. Losing organisations deploy broadly, measure loosely, and report activity rather than outcomes.

Instead of running with the crowd, AI should be seen as an opportunity to stop, and reinvent.

TAKEAWAY: Instead of running with the crowd, AI should be seen as an opportunity to stop, and reinvent.

When you need to compare options (vendors, strategies, scenarios), most AI outputs default to prose that buries the differences. Instead, ask the model to produce a structured table and specify exactly what columns you need. This forces the AI to be explicit and parallel in its reasoning.

This works because gaps and weaknesses that prose can obscure become immediately visible when they must sit in a column alongside their alternatives. The output is also directly usable in a board pack or briefing document without further reformatting.

Thanks for reading, the wAIve Daily Team. Found this useful? Forward it to someone who needs it. Tell us what you need to know. Reply with your questions.

wAIve Daily is published with Centai Strategy