Get tomorrow's brief in your inbox
May 1, 2026
Today: Google Cloud just posted its strongest growth on record (63%) as enterprise AI spending went from promise to payday. Stripe launched Link, a digital wallet that lets your AI agents spend money on your behalf with approval controls. Meta is quietly processing 10 million business AI conversations per week while flush startups rent massive Manhattan offices with only a handful of employees.
Google Cloud revenue hit $20 billion this quarter, up 63% year over year. That is the fastest growth rate Google has ever posted since they started breaking out cloud numbers in 2020. Source
What changed? Enterprise AI solutions became Google Cloud's primary growth driver for the first time. Companies are no longer experimenting with AI. They are buying it at scale. Google's revenue from products built with generative AI models grew 800% this quarter. AWS saw similar momentum, with customers using 170% more tokens on its Bedrock service than last quarter. Microsoft Azure grew 40%, with adoption of Anthropic and OpenAI models doubling.
If you run a business, this tells you two things. First, your competitors are already buying AI infrastructure and building tools on it. Waiting is falling behind. Second, the three big cloud providers (AWS, Azure, Google Cloud) are now in a full sprint to win your business. That means better pricing, more flexible contracts, and aggressive customer support. If you have been thinking about migrating workloads to the cloud or testing AI models, this is the best negotiating environment you will see for years. Source
The flip side: Big Tech is collectively spending close to $600 billion this year on AI capital expenditures. Analysts now project that number will top $1 trillion in 2027. Source That money flows to chipmakers, data center providers, and gear suppliers. If you sell to those companies, you are in the right place at the right time.
Stripe Link: A Digital Wallet Built for the AI Agent Era
Stripe released Link, a digital wallet that works on web, iOS, and Android. You can connect payment methods (cards, banks, crypto, buy now/pay later), track spending, and manage subscriptions. The standout feature: you can give your AI agents permission to spend money on your behalf without exposing your payment credentials. Source
Here is how it works. You grant your AI agent access to Link via OAuth. The agent creates a spend request and waits for your approval. You review the transaction on mobile or web before the payment credential is shared. Stripe says spending limits and approval-free transactions are coming soon. For now, Link supports traditional payment methods, with support for agentic tokens and stablecoins coming "soon."
If you are experimenting with autonomous AI agents (like OpenClaw or similar tools), Link solves the trust problem. You get the convenience of automation without handing your credit card to a black box. Stripe also offers Issuing for agents, which lets you issue virtual cards with spending controls and real-time authorization. Developers building agents can use Link instead of building their own wallet from scratch. Free to use, no pricing announced yet. Source
Google Gemini Is Coming to Your Car (If You Have Google Built-In)
Google announced that Gemini is rolling out to cars with Google built-in, replacing the current Google Assistant. This is not just new cars. Existing vehicles will get the upgrade via software update, starting with English in the U.S. GM already confirmed Gemini is coming to 4 million model year 2022 and newer vehicles (Cadillac, Chevrolet, Buick, GMC). Source
Gemini lets you speak more naturally. Instead of memorizing specific commands, you can say something like "find a highly rated sit-down restaurant with outdoor seating along my route" and Gemini will pull suggestions from Google Maps, then handle follow-up questions about parking or menus. It can also turn on the heat, recommend music, retrieve vehicle information (like EV battery levels), and summarize or respond to messages hands-free.
There is also Gemini Live, currently in beta. You activate it by tapping a button or saying "Hey Google, let's talk." This enables open-ended, real-time conversations for brainstorming or learning while you drive. Google plans to expand support to more languages and regions, and eventually integrate Gmail, Calendar, and Google Home. Source
If you have a compatible vehicle, watch for the upgrade prompt when you are signed into your Google account. This is free with Google built-in.
Codex CLI Adds /goal Command for Autonomous Coding Loops
OpenAI's Codex CLI (version 0.128.0) now supports a /goal command. You set a goal, and Codex keeps looping until it evaluates that the goal has been completed or the configured token budget runs out. This is similar to the Ralph loop pattern. Source
The feature is implemented through automatically injected prompts (goals/continuation.md and goals/budget_limit.md). If you are already using Codex for coding work, this adds a layer of autonomy. Instead of iterating step by step, you define the end state and let Codex figure out the path. Be aware that autonomous loops can burn through token budgets fast if the goal is complex or ambiguous. Set clear goals and monitor usage.
Goodfire's Silico: A Tool for Debugging AI Models During Training
Goodfire, a San Francisco startup, released Silico, a mechanistic interpretability tool that lets you look inside an AI model and adjust its parameters during training. This is the first off-the-shelf tool of its kind that works across all stages of development, from building a dataset to training a model. Source
Silico uses agents to automate much of the complex interpretability work that used to require humans. You can zoom in on specific neurons or groups of neurons, run experiments to see what they do, and trace pathways upstream and downstream to understand how neurons affect each other. Goodfire claims to have reduced hallucinations in LLMs using this technique. With Silico, they are packaging their in-house methods as a product.
This is for developers and researchers who train their own models or work with open-source models. You cannot use it to poke around inside ChatGPT or Gemini, but you can use it on models like Qwen 3. Pricing not announced. If you are training custom models and struggling with unexpected behaviors, Silico might help you understand what is going on under the hood. Source
Meta's Business AI Is Processing 10 Million Conversations Per Week
Meta's business AI tools are now facilitating 10 million conversations per week as of late March, up from 1 million at the beginning of the year. The company recently expanded its business AI assistant beta in the U.S., EMEA, APAC, and LATAM. Source
Meta is offering these tools for free to small businesses to achieve scale, but CEO Mark Zuckerberg hinted that monetization is coming. The company also said 8 million advertisers are using at least one of its GenAI ad creative tools, with strong adoption among small and medium-sized businesses. Advertisers using the video generation feature saw 3% higher conversion rates in tests.
If you advertise on Facebook or Instagram, you should test the AI creative tools. Free right now, measurable lift in conversions, and it takes minutes to generate variations. Meta is also launching Meta Ads AI Connectors this week, which lets you connect your Meta ad account to an AI agent for automated campaign management. Source
Apple Sold Out of Mac Minis and Studios Due to AI Workload Demand
Apple reported Mac revenue of $8.4 billion in Q2, beating Wall Street's $8 billion estimate. Mac sales were up 6% year over year. CEO Tim Cook said customer demand for running local AI models (like OpenClaw) on Mac mini and Mac Studio devices was higher than Apple expected. Source
Both Mac mini and Mac Studio sold out in recent weeks. Cook said it may take "several months" to reach supply-demand balance. The Mac mini was the top-selling desktop in China, where OpenClaw adoption is exploding. Apple also saw strong enterprise demand, with companies like Perplexity using Mac as their platform for building enterprise AI assistants.
If you need a machine for running local AI models, expect Mac mini and Mac Studio to remain supply-constrained through summer. The Neo MacBook (colorful, consumer-focused) also saw strong demand, but availability is better there. For business use cases, the mini and Studio remain the best value for local AI workloads. Source
Legal AI Startup Legora Hits $5.6B Valuation, Harvey at $11B
Legora, a Swedish legal tech startup, raised a $50 million Series D extension at a $5.6 billion valuation. The company crossed $100 million in annual recurring revenue and claims 1,000 law firms and in-house legal teams as customers across 50 markets. Nvidia's NVentures led the extension, marking Nvidia's first legal AI investment. Source
Competitor Harvey raised at an $11 billion valuation last month and claims 100,000 lawyers across 1,300 organizations as customers. Both companies are spending heavily on marketing (Harvey partnered with actor Gabriel Macht, Legora launched a campaign with Jude Law) and expanding globally.
If you run a law firm or in-house legal team, this is a good time to negotiate. Two well-funded competitors are fighting for market share, which means aggressive pricing and feature rollouts. Both platforms launched only 18 months ago, so they are still proving themselves. Ask for pilot periods, volume discounts, and clear ROI metrics before committing. Source
Big Tech AI Spending Hits $600 Billion This Year, $1 Trillion by 2027
Wall Street analysts now estimate total AI capital expenditures will exceed $1 trillion in 2027, up from $600 billion in 2026. This year's spending projections increased across the board: Google up 4% to $185 billion, Amazon up 1% to $200 billion, Meta up 8% to $135 billion, and Microsoft up 24% to $190 billion. Source
Meta faces the most investor scrutiny because it has no cloud infrastructure business to monetize AI investments. Google, Amazon, and Microsoft can turn their spending into revenue by selling access to customers. Meta's free cash flow dropped to $1.2 billion in Q1 from $26 billion last year. JPMorgan downgraded Meta to neutral, citing a "challenging path" to generating returns on AI spending. Source
Alphabet stock surged 34% in April, its best month since 2004. Meta shares dropped 9% after earnings. The diverging moves show Wall Street is not guaranteed to reward every AI spending spree. You need a clear path to revenue. Source
ChatGPT Images 2.0 Sees Strong Adoption in India, Modest Growth Elsewhere
OpenAI's ChatGPT Images 2.0 launched last week with improved handling of complex prompts and accurate text rendering across multiple languages. India became the largest user base, with about 5 million downloads during launch week compared to 2 million in the U.S. However, global engagement gains were modest: daily active users and sessions up only 1%. Source
Emerging markets like Pakistan, Vietnam, and Indonesia saw sharper spikes in downloads (up to 79% week-over-week). Early use cases in India focus on self-expression: stylized portraits, social media visuals, fantasy newspaper covers, and photo restoration. The tool is free for ChatGPT users, but adoption patterns suggest it is resonating more in markets where professional design tools are less accessible. Source
Elon Musk Confirms xAI Used OpenAI Models to Train Grok
In federal court testimony, Elon Musk confirmed that xAI used OpenAI's models to improve its own Grok model through a practice called model distillation. When asked if xAI distilled OpenAI's technology, Musk said "partly," adding that "generally all the AI companies" do it and that it is "standard practice to use other AIs to validate your AI." Source
Model distillation is when one larger AI model acts as a "teacher" to pass knowledge to a smaller "student" model. It is legitimate when used within companies to train their own models, but controversial when used to mimic a competitor's model. Companies like OpenAI and Anthropic have accused Chinese firms (DeepSeek, Moonshot, MiniMax) of distilling their models. Source
Test AI-Generated Ad Creative on Meta (Free, 3% Conversion Lift)
If you advertise on Facebook or Instagram, try Meta's GenAI ad creative tools. They are free right now, take minutes to use, and advertisers using the video generation feature are seeing 3% higher conversion rates in tests. Source Log into Ads Manager, look for the GenAI tools under creative options, and generate 3-5 variations of your best-performing ad. Run them as split tests for a week and see if the lift holds.
Give Your AI Agent a Wallet (Stripe Link, Free)
If you are running AI agents for automation, set up Stripe Link to let them make purchases on your behalf with approval controls. It is free, supports OAuth, and gives you transaction visibility without exposing payment credentials. Source This is especially useful if your agent books travel, orders supplies, or subscribes to services. You approve each spend before it goes through, then set spending limits later when you trust the workflow.
Google's 63% cloud growth and the $1 trillion spending forecast tell the same story: AI infrastructure is no longer speculative. Companies are buying it, deploying it, and seeing returns. The next phase is differentiation. Stripe is building wallets for agents, Google is putting Gemini in cars, and Meta is processing 10 million business conversations per week. The winners will be the companies that embed AI into workflows customers already trust, not the ones building standalone AI products nobody asked for. If you have been waiting for proof that AI spending pays off, this quarter was it. Now the question is whether your business is positioned to compete with companies that are already all-in.