Get tomorrow's brief in your inbox
Today: Anthropic bought all compute capacity at xAI's Colossus 1 data center, doubling Claude Pro rate limits by partnering with SpaceX for orbital infrastructure. Google launched a free 5-day AI Agents course starting June 15, teaching "vibe coding" (natural language programming) for small business owners. Meanwhile, frontier companies are using AI 3.5 times more deeply per worker than typical firms, while Snap's $400M Perplexity deal collapsed after disagreeing on rollout strategy.
The compute crunch is reshaping how AI companies operate. This week, Anthropic announced it bought the entire output of xAI's Colossus 1 data center (300 megawatts of capacity) to immediately double rate limits for Claude Code Pro, Max, and Enterprise users. Elon Musk's xAI had moved training to Colossus 2, leaving the first data center excess capacity. The move signals a bigger shift: xAI is repositioning itself as a compute provider rather than just an AI model company.
What this means for small business: The rate limit doubling (Claude Code going from 5 hours to 10 hours per day for Pro users) translates directly to lower friction for developers and entrepreneurs relying on AI coding. If you're using Claude Code, you get more uninterrupted work time. Beyond the rate limits, this deal proves compute is becoming as strategic as the models themselves. Anthropic is also leaning on SpaceX to use orbital data center capacity, signaling that the hyperscalers believe physical AI infrastructure is a bottleneck that land-based data centers alone can't solve.
The broader story: Multiple sources this week confirmed that chip shortages, energy constraints, and space limitations are the real bottlenecks slowing AI deployment. Frontier companies (the top 5% of AI users) are already working around these constraints by delegating more work to agents and optimizing usage patterns. If you're planning AI infrastructure investment, the takeaway is clear: access to compute will matter more than model choice in 2026.
Sources: Anthropic/SpaceX compute partnership, AI infrastructure constraints, xAI data center details
Google's Free AI Agents Course (June 15-19) - Google and Kaggle are running a redesigned 5-day intensive on building AI agents, free, for the second time. Focus: "vibe coding" (natural language as programming interface). Participants build "10x agents" that can automate customer service, scheduling, market analysis. Includes a capstone project. What it costs: Nothing. Who should apply: Non-technical small business owners and IT professionals wanting to understand agent-building without coding background. Why it matters: This is the first mainstream education pushing non-programmers toward agent workflows. The course last year attracted 1.5 million learners.
Action: Mark your calendar for June 15. Early signup usually fills up fast.
Anthropic Managed Agents: Multi-Agent Orchestration - New features shipped this week: multi-agent coordination (create fleets of agents for complex tasks), outcomes definition (tell Claude what success looks like, it iterates), and "dreaming" (agents inspect past sessions to self-improve). Intended for teams building production systems. Impact: Teams using this report 10x faster shipping speed according to Anthropic. If you're planning agent-based workflows, waiting for stable versions of these features will pay dividends over hand-rolling your own coordination logic.
Source: Anthropic Code w/ Claude event
Google AI Search Updates - Google's AI Overviews now pull in quotes and perspectives from Reddit, forums, and firsthand sources. Adds creator attribution and context links. Also: new feature highlights links from your news subscriptions. Why it matters: If your business relies on search visibility, expect Google to credit Reddit and discussion boards more prominently. Plan content strategy around helping users find your expertise in forums where your audience actually hangs out. Example: If you run a service business, make sure you're answering questions in industry forums where Google will now surface that content.
Sources: The Verge, TechCrunch, Google announcement
How Frontier Firms Are Actually Using AI - OpenAI published B2B Signals research showing that frontier firms (top 5% of enterprise usage) now use 3.5x as much "intelligence per worker" as typical firms, up from 2x a year ago. But here's the real insight: message volume only explains 36% of the gap. Most advantage comes from depth and complexity. Frontier firms send 16x as many Codex (coding) messages per worker. They use agents for delegated work, not just chat-based Q&A.
Translation for small business: You don't need to be OpenAI or Stripe to move ahead. The firms pulling away from the pack are those treating AI as a work multiplier (taking on harder tasks, providing richer context, asking for more substantial outputs) rather than a fancy search box. If your team is only using ChatGPT to summarize emails, you're in the typical zone. If you're using it to generate proposals, analyze customer data, build APIs, and iterate on code - you're moving toward frontier behavior. The depth matters more than the frequency.
Source: OpenAI B2B Signals report
Uber's AI Assistant for Drivers - Uber deployed an AI-powered assistant helping drivers decide where to position themselves, when to switch between ride and delivery, and how to optimize earnings. Impact: New drivers ramp up 10-100x faster than learning through trial and error. The assistant turns complex marketplace data (heatmaps, trends, earnings comparisons) into plain-language insights. Experienced drivers also use it repeatedly, proving it's a long-term tool, not just an onboarding help.
Lesson for service businesses: If you're managing a marketplace or distributed workforce, AI assistants that translate complex operational data into plain-language guidance are high-ROI. Drivers get better earnings, Uber gets retention and faster ramp. Build that feedback loop.
Source: OpenAI / Uber case study
Match Group Slows Hiring to Fund AI Tools - The parent company of Tinder and other dating apps is cutting hiring and not filling 300 open positions to pay for AI enablement across the company. CFO Stephen Bailey said every employee gets access to cutting-edge AI tools and training. The company expects productivity gains to offset the headcount reduction and ultimately drive revenue. Also: Match is pivoting strategy away from app-centric dating toward in-real-life events, responding to Gen-Z burnout with digital dating.
Takeaway: If you run a team and want to move to AI-native workflows, budget for tool costs and plan for productivity gains over time. The cost-neutral math only works if adoption is real and tools are actually used. Also watch: traditional business models (dating apps) are being disrupted by generational behavior shifts, not just AI. AI tooling alone won't save a product if market fit is deteriorating.
Claude Code Rate Limits Double (Today) - Anthropic doubled the 5-hour daily limit for Claude Code Pro, Max, and Enterprise customers to 10 hours per day. This is live now. Higher limits mean less context switching, fewer interrupted workflows, and more time for complex tasks like code review, multi-file refactoring, and long-running agent loops.
If you're a Claude Code subscriber: This directly increases your productive capacity. If you hit the old limit regularly, start planning longer, more complex tasks.
Source: Anthropic Code w/ Claude 2026
Perplexity-Snap Deal Ends - Snap and Perplexity "amicably ended" their $400 million partnership in Q1. The companies couldn't agree on how to roll out Perplexity's AI search engine inside Snapchat beyond a limited test. Snap is now assuming zero contribution from the partnership in Q2 guidance.
What this signals: Revenue-share partnerships between AI startups and platforms are harder to execute than either party expected. Integration timelines slip. Feature roadmap disagreements cascade. If you're evaluating partnerships with AI startups, expect negotiation timelines to slip and be conservative on revenue timing.
DeepSeek Valuations Surge - The Chinese AI lab that shocked the market in early 2025 with efficient open-weight models is raising its first VC round at a $45 billion valuation, up from $20 billion just weeks ago. The round is led by China's state-backed chip fund, with Tencent and Alibaba reportedly interested. DeepSeek has optimized models to run on Huawei chips, positioning China to have an AI pipeline independent of U.S. technology.
Implication: The AI competition is genuinely bifurcating. U.S. and Chinese companies are building parallel stacks. Expect tariffs, export controls, and supply chain fragmentation to matter more for AI infrastructure pricing and availability going forward.
Sources: TechCrunch, FT reporting
EU Considers Restricting U.S. Cloud for Sensitive Data - The European Commission is drafting rules (part of a "Tech Sovereignty Package" due May 27) that would restrict member governments' use of U.S. cloud providers for sensitive data (health, financial, judicial). The move reflects U.S.-EU tension and Europe's desire for digital autonomy. Private sector companies are not affected by the proposal.
For SMBs: If you serve government or highly regulated clients in Europe, watch for this regulation. It may accelerate demand for European cloud alternatives (AWS, Azure, Google Cloud likely have compliant local offerings, but startups offering "European cloud" may see tailwinds).
OpenAI Expands ChatGPT Ads - OpenAI is expanding ChatGPT Ads beyond pilot, introducing cost-per-click bidding. SMBs can now run ads directly in ChatGPT. Early uptake suggests this is a new growth channel (similar to Google Ads, Reddit Ads, TikTok Ads) for reaching users actively seeking information.
Action: If you run paid ads, test a small ChatGPT campaign. Cost is likely high CPL initially, but audience intent is probably strong.
Prompt: Use AI to Build a Positioning Worksheet Open Claude Code and ask: "Build me an interactive HTML form where I answer 5 questions about my business (target customer, biggest pain, competitor differentiation, pricing model, revenue model) and generate a 200-word positioning statement in real-time as I type. Make it professional but casual. No dark colors." Takes 10 minutes, zero code knowledge required. You get a tool you can reuse with clients or team members. Save the HTML file and share. This is "vibe coding" in action: you're delegating the code entirely, keeping the thinking.
Automation Idea: LinkedIn + CRM Signal Sync If you use LinkedIn for lead gen, set up a workflow (Zapier, Make, n8n) that catches LinkedIn Ads CAPI conversions and logs them directly into your CRM with campaign attribution. This closes the "signal gap" (mentioned in several articles this week). Most demand gen teams waste hours on manual reconciliation. Automating this 30-minute setup saves your team 2-3 hours per week. (See Zapier blog links in Pricing section.)
The AI economy is stratifying along two axes: compute access and usage depth. Frontier companies are pulling ahead not by having fancier models, but by using AI for harder work and delegating more tasks to agents. Compute is now the bottleneck, not capability, and Anthropic's move to buy SpaceX data center capacity signals that even the biggest labs are scrambling for power. Meanwhile, you can get started on AI skills (Google's free course), AI infrastructure (Anthropic's doubled rate limits), and AI distribution (ChatGPT Ads) all at zero or minimal cost. The real differentiator in the next three months will be whether you're using AI to do the same things faster, or using it to do new, harder things that weren't possible before.
Watch this week for: Anthropic's rate limit usage patterns (high demand signals more infrastructure pressure), DeepSeek model releases (the China-U.S. AI divergence accelerating), and early results from Google's agents course (on what real non-technical people actually build with "vibe coding").
Word count: 1,487