Send us Fan Mail

OpenAI and Dell Technologies announced a partnership yesterday to bring Codex and ChatGPT Enterprise into hybrid and on-premises environments through the Dell AI Data Platform and Dell AI Factory. The deal lands at the moment Codex weekly developer usage crossed four million and gives OpenAI a serious enterprise distribution answer to Claude Code.

Ramp's May 2026 AI Index, published yesterday, shows Anthropic overtaking OpenAI in US business spend for the first time. Anthropic's adoption climbed to 34.4% of tracked businesses in April; OpenAI fell to 32.3%. The engine is Claude Code — now reportedly the fastest-growing product in Anthropic's history and authoring roughly 4% of all public GitHub commits.

xAI quietly opened Grok Build to public beta over the weekend, an agentic CLI coding tool initially limited to SuperGrok Heavy subscribers at $300 a month, with a $99 introductory price for the first six months. Grok joins the coding-agent race against Claude Code, Codex and Google's developer stack — and is now the third paid CLI competing for engineering wallets.

Send us Fan Mail

Google I/O 2026 opens tomorrow morning, Tuesday 19 May, at the Shoreline Amphitheatre. The Android Show on 12 May already locked in Gemini Intelligence, Googlebook on Aluminium OS, Android 17 with generative UI, and Android XR glasses with Samsung, Warby Parker and Gentle Monster. The keynote tomorrow is expected to land a new Gemini model and the developer tooling that flows into Google Marketing Live on Wednesday.

OpenAI quietly turned ChatGPT advertising into a measurable performance channel on 5 May. CPM pricing has been replaced with cost-per-click bidding at $3-$5 per click. The self-serve Ads Manager is in beta. Minimum advertiser spend has dropped from $250,000 to $50,000. OpenAI is publicly forecasting $2.5bn in ad revenue this year, scaling to $11bn by 2027.

Snap and Perplexity have "amicably ended" the $400m AI search partnership originally signed in November 2025, per TechCrunch and Engadget on 6 May. The integration never moved beyond limited testing inside Snapchat's Chat interface. It is the first nine-figure AI distribution deal between two named, well-funded companies to publicly fail, and the cleanest case study yet for the limits of platform-scale AI partnerships.

Coty has tied up with Pencil — owned by Brandtech Group, sister company to Jellyfish — to build an end-to-end generative AI content system across CoverGirl, Rimmel, Sally Hansen and Max Factor. An embedded Pencil team starts inside Coty on 1 July. It is the most credible enterprise GenAI content deployment we have seen in beauty.

Pattern across all four stories: AI is becoming the operating layer of marketing, not a feature of it. ChatGPT is now priceable. Snap is showing what happens when an AI deal fails to ship. Coty is showing what an in-house GenAI production line actually looks like. And Google is about to reset the infrastructure on Tuesday and the ad framework on Wednesday.

Watchlist: Google Marketing Live on Wednesday 20 May, and Mustafa Suleyman's 18-month white-collar automation framing recirculating across Fortune, FT and Yahoo Finance over the weekend.

Send us Fan Mail

YouTube used its 2026 Brandcast on 13 May to ship the most credible AI ad stack in television this year: Custom Sponsorships matched dynamically by AI, Multimodal Video Creation powered by Gemini, Veo and Nano Banana, and Buy with Google Pay on connected TVs. It is the cleanest preview of what Google Marketing Live will land on Wednesday and the most useful direct-response brief any advertiser will read this weekend.

Marc Benioff said on the All-In podcast, published Friday, that Salesforce expects to spend roughly $300 million on Anthropic tokens this year, almost entirely on coding. It is the single largest disclosed enterprise AI procurement figure of 2026 to date and gives operators a concrete reference for what frontier-model spend looks like inside a large software company.

Anthropic is in advanced talks to acquire developer tools startup Stainless for at least $300 million, per The Information. Stainless writes the SDKs that ship with OpenAI's, Google's and Meta's APIs. The structural read is that Anthropic is buying the layer through which its rivals reach their own developers.

Apple heads into WWDC on 8 June with iOS 27, a redesigned Siri running on Gemini under a Google Cloud partnership confirmed at Cloud Next, and an expected extension framework that lets third-party assistants plug into Writing Tools and Image Playground. For brands and apps living inside Apple's distribution, this is the planning event of the summer.

Underneath all four stories is the same pattern: AI is now a procurement, distribution and infrastructure question, not a model question. The economic and creative consequences are landing inside marketing budgets, ad platforms and developer ecosystems this week.

Watchlist: Google I/O on 19–20 May and Google Marketing Live on 20 May. Gemini 3.2 Flash has surfaced in LM Arena and inside iOS Gemini app strings, and is the most likely model launch of the next 48 hours.

Send us Fan Mail

OpenAI moved ChatGPT into regulated retail finance on Friday with a Plaid-powered Personal Finance experience for Pro users in the US, reading from more than 12,000 financial institutions. It is the most consequential consumer surface OpenAI has shipped since ads, because it pulls account-level data into the model and reframes ChatGPT as a financial dashboard, not just a chatbot.

Anthropic is in talks to raise between thirty and fifty billion dollars at a valuation of up to $950bn, against a stated $30bn annualised revenue run rate as of April. If priced, it overtakes OpenAI's $825bn mark and reorders the capital map of the sector. Enterprise share, not consumer share, is doing the work.

HubSpot has shipped AEO Sensor as a free public dashboard tracking citations, mentions and AI-referred traffic across ChatGPT, Gemini and Perplexity. It lands in the same week that ChatGPT referral traffic hit a twelve-month low — useful timing for anyone still building an AI search measurement layer from scratch.

Meta has launched Incognito Chat for Meta AI inside WhatsApp and the Meta AI app, running on its Private Processing trusted execution environment. The chats are not saved, not visible to Meta, and not retained by default. For brands using WhatsApp as a commerce or service channel, this changes the data exhaust assumption.

Underneath all four stories is a single pattern: AI is moving from horizontal chat into vertical, account-aware product surfaces — finance, distribution, search measurement, private messaging — and the commercial implications now sit firmly inside marketing, CRM and ecommerce planning, not innovation teams.

Watchlist: We will also flag Google I/O on 19–20 May and Google Marketing Live on 20 May as the week ahead's main pressure point, plus the Gemini Omni leak as the model story to watch.

Send us Fan Mail

Anthropic moved down-market on 13 May with Claude for Small Business, a Cowork-based bundle with built-in workflows across finance, ops, sales, marketing, HR and customer service and out-of-the-box connectors to QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace and Microsoft 365. There is no extra licence fee on top of existing Claude plans. A ten-city US training tour starts today.

At SAP Sapphire 2026 in Orlando, Christian Klein used the 13 May keynote to brand SAP as "a business AI company" and named Claude as a primary reasoning and agentic model embedded in the new SAP Business AI Platform via Joule. Mistral is live on the platform now, Cohere North arrives in June, and Nvidia's OpenShell sits under the Autonomous Suite as the agent control layer.

Google used the Android Show: I/O Edition on 12 May to reposition Android as "an intelligence system" with Gemini Intelligence at the core — agentic actions across apps, multi-step task execution, screen-context awareness — rolling out first on Samsung Galaxy and Pixel this summer, then watches, cars, glasses and the new Googlebook laptop line in the autumn.

Anduril closed a $5bn Series H on 13 May at a $61bn valuation, led by Thrive Capital and Andreessen Horowitz. Valuation doubled in roughly a year. 2025 revenue doubled to $2.2bn. The capital backs Arsenal-1, Anduril's Ohio manufacturing facility, on top of a $20bn ten-year Pentagon contract signed in March. Defence is now where the biggest applied-AI cheques are landing.

Watchlist: Google's Threat Intelligence Group disclosed the first known case of an AI-generated zero-day used in the wild — a 2FA bypass in an open-source admin tool — flagged before a mass exploitation event. And ChatGPT Ads Manager, OpenAI's self-serve CPC ad platform, is now widely accessible in the US with Dentsu, Omnicom, Publicis and WPP signed as agency partners.

The pattern across the day: the frontier labs are no longer chasing one another for model supremacy — they are racing to embed inside the workflows their buyers cannot easily replace. Anthropic at the SMB desktop. Claude inside SAP. Gemini inside Android. Anduril inside the Pentagon. The model is the input. The integration is the moat.

Send us Fan Mail

OpenAI has turned its GPT-5.5-Cyber model into a named platform. Daybreak launched on 12 May with a defender-first pitch — find vulnerabilities, generate patches, validate fixes — and a launch roster that includes Akamai, Cisco, Cloudflare, CrowdStrike, Fortinet, Oracle, Palo Alto Networks and Zscaler. Two days after we covered the model, OpenAI now has the brand, the workflow and the partner stack to sit inside the enterprise security tooling layer.

Vapi raised $50M Series B at a $500M valuation led by Peak XV, with M12, Kleiner Perkins and Bessemer participating. The headline detail is operational, not financial: Amazon Ring evaluated more than 40 voice-AI vendors and now routes 100% of inbound calls through Vapi, in production within two weeks. Vapi has crossed one billion calls handled. Voice AI for customer service has quietly cleared the production bar.

JPMorgan initiated coverage on Mistral with a $430bn sovereign-AI thesis on 12 May. The bank says more than 60% of European enterprises plan to increase sovereign-AI spend over the next two years, pegs Mistral's annualised recurring revenue at about $400M as of February with enterprise at roughly 95% of revenue, and cites the recent Koyeb acquisition and a 13,000-GPU Nvidia GB300 deployment as the operational base. Europe is being priced as a real third pole.

Isomorphic Labs, the DeepMind drug discovery spin-out, closed a $2.1bn Series B on 12 May led by Thrive Capital, with Alphabet, GV, CapitalG, MGX, Temasek and the UK Sovereign AI Fund participating. It is one of the largest private rounds in AI-for-drug-discovery on record and a useful reference point for where the deep-pocketed capital is choosing to compound.

Watchlist: An unusually busy day for agentic commerce launches — Shoplazza's Athena admin agent, Marqo's Sibbi unified shopping agent and Preciso's Ultima Ads DSP for Shopify all shipped 12 May. And the brand-safety conversation is moving from theoretical to operational, with regulators flagging AI-generated impersonations of Nike, Amul and others.

The unifying theme: Capital and capability are now flowing to whoever can sit inside an enterprise workflow and quietly take work off the desk. Daybreak inside the SOC. Vapi inside the contact centre. Mistral inside the European regulated stack. Isomorphic inside the drug pipeline. The model layer is no longer the prize. The integration into a real production process is.

Send us Fan Mail

OpenAI has formally entered the enterprise services business. The new OpenAI Deployment Company launches with $4bn, 19 investor and integrator partners — including TPG, Bain Capital, Brookfield, Advent, Goldman Sachs and SoftBank — and the acquisition of Scottish consultancy Tomoro. The hard read for agencies and consultancies: OpenAI now sells the implementation layer, not just the model.

Meta has opened its ad ecosystem to outside AI. The official Meta Ads MCP server and Meta CLI are now in open beta, with day-one write access from Claude, ChatGPT and Perplexity. Campaigns, ad sets and catalogues can be created and edited in natural language from a third-party assistant. This is the first time a major ad platform has handed write access to a competitor's model.

Alibaba is integrating Qwen with Taobao and Tmall. The Qwen app will get an end-to-end shopping agent across more than four billion products, with virtual try-ons, 30-day price tracking and post-sale handling inside the same conversation. It is the most ambitious agentic commerce launch from any platform in the world.

xAI is in fresh restructuring. The Information reports another round of layoffs, ongoing co-founder exits and Cursor staff already meeting xAI teams ahead of SpaceX's $60bn option on Cursor. The talent map at the frontier is being redrawn, again.

Watchlist: Anthropic's $1.8bn Akamai compute deal and reported lease of SpaceX's Colossus 1 data centre point to a scale-or-suffocate quarter on compute. And Perplexity's Comet browser, now live on iOS globally and integrated into Samsung's S26 line, is quietly becoming a real distribution channel.

The unifying theme: the AI majors are no longer selling tools. They are selling deployment, distribution and transactions. The boundary between platform, agency and storefront is dissolving in the same week.

Send us Fan Mail

Google has dropped its pre-Marketing-Live bidding and budgeting update two weeks ahead of the 20 May livestream. Journey-Aware Bidding enters beta for Target CPA Search, Smart Bidding Exploration extends to Performance Max with product feeds and Shopping, and Demand-Led Pacing rolls out across Search and Shopping in the coming months. These are not features. They are removals of the manual levers performance teams still rely on for governance.

Cloudflare announced the first mass layoff in its 16-year history alongside a record Q1: revenue $639.8M, up 34%, while cutting 1,100 staff, roughly 20% of headcount, on the explicit grounds that agentic AI made the work redundant. The stock fell 24%. The market is not punishing the cuts — it is punishing the margin profile that survives them.

Qualcomm CEO Cristiano Amon told Fortune the company is working with "pretty much all" major AI players, naming OpenAI and Meta, on secret hardware that is not a phone — glasses, jewellery, pins, pendants. This is the most explicit on-record confirmation yet that the post-smartphone interface for AI agents is being engineered now, not five years out.

OpenAI has expanded GPT-5.5-Cyber in limited preview to vetted defenders under its Trusted Access for Cyber programme. The model carries fewer guardrails, can generate and validate vulnerability exploitation plans against simulated systems, and ships with mandatory Advanced Account Security from 1 June. The commercial read is that frontier model access for security workflows is being tiered, gated and audited — not democratised.

Watchlist: OpenAI's B2B Signals research sets a new internal benchmark — frontier firms now use 3.5x the AI intelligence per worker that typical firms do, up from 2x a year ago, and only 36% of that gap is explained by message volume. The rest is depth. Google Preferred Sources also expanded to all supported languages globally, a quietly important SEO signal as brand recognition becomes a discoverability lever.

The pattern of the day: tooling is moving faster than governance. Bidding control is collapsing into AI-managed beta surfaces. Cost lines are being restructured around agents before measurement frameworks catch up. The interface itself is being redesigned around wearables before anyone has agreed what an "ad" looks like on a pin. The operator question this week is not which feature to test — it is which controls you still own.