Get tomorrow's brief in your inbox
Today: China blocked Meta's $2B acquisition of AI startup Manus, halting the "Singapore-washing" model that let Chinese AI companies avoid Beijing and Washington scrutiny. Google invested $40B in Anthropic to strengthen its cloud ties to Claude AI. Oracle added AI-driven workflows to Primavera Unifier for automated project management.
China blocks Meta's $2 billion Manus acquisition, ends "Singapore-washing" workaround
China's National Development and Reform Commission ordered Meta to unwind its $2 billion purchase of Manus, a Singaporean AI startup that was originally founded in China before relocating. This marks Beijing's first major intervention against the "Singapore-washing" model, where Chinese AI companies relocate to Singapore to dodge restrictions from both Beijing and Washington.
For small businesses, this signals that cross-border AI acquisitions are getting riskier. If you're building AI products that rely on tools or APIs from startups with Chinese ties (even indirect ones), expect supply chain uncertainty. Manus had hit $100 million in annual recurring revenue just eight months after launch with a general-purpose AI agent that handled market research, coding, and data analysis. Companies like this won't disappear, but their access to Western capital and cloud infrastructure just got harder. If you were counting on a particular AI tool staying available long-term, diversify your stack now.
Meta wanted to integrate Manus into its Meta AI assistant and enterprise products. That won't happen. For businesses evaluating AI agents, this means the big platforms (OpenAI, Anthropic, Google) just got relatively safer bets compared to fast-moving startups with murky ownership structures.
Google bets $40 billion on Anthropic to compete with its own Gemini
Google invested $40 billion in Anthropic, the company behind Claude AI, which directly competes with Google's own Gemini chatbots. The deal strengthens Google Cloud's ties to Anthropic, positioning Claude as a key product for Google's enterprise AI customers.
This is unusual: Google is funding its own competitor. For your business, it means Claude AI (which runs on Google Cloud infrastructure) is now a safer long-term bet. If you've been hesitant to commit to Claude because you weren't sure about Anthropic's funding runway, this $40B investment answers that question. Google gets to monetize Claude's growth through cloud usage even while competing with Gemini. You benefit from having two well-funded options instead of one.
Meta signs deal for solar power from space
Meta inked an agreement with Overview Energy to reserve up to 1 gigawatt of power from satellites that beam infrared light to solar farms at night. The first satellite launches in January 2028, with full deployment starting in 2030 using 1,000 spacecraft in geosynchronous orbit.
For businesses running AI workloads, this is a signal that data center power costs might stabilize in the 2030s. Meta used 18,000 gigawatt-hours in 2024 (enough to power 1.7 million homes) and needs 30 gigawatts of renewable capacity to keep scaling AI. If space-based solar works, it changes the economics of running large-scale AI infrastructure. For now, expect cloud providers to keep raising prices as they scramble for power. If you're building AI-heavy products, factor in rising compute costs over the next 3-5 years before this infrastructure matures.
Oracle Primavera Unifier adds AI-driven project workflows
Oracle launched AI capabilities in Primavera Unifier that automate work prioritization, create audit trails, and streamline compliance workflows. The no-code uDesigner tool lets teams build automated workflows without technical expertise. Pricing wasn't disclosed, but Primavera is typically enterprise-focused.
If you run a project-heavy business (construction, engineering, regulated industries), this addresses a real problem: managing compliance paperwork without hiring more admins. The AI summarizes progress, captures timestamps and decisions automatically, and integrates with ERP and scheduling systems to reduce manual data entry. The real-time dashboards help you spot project risks earlier. The downside is implementation overhead. Oracle's tools require upfront integration work, so this makes sense for firms managing multiple projects simultaneously, not one-off contractors. If you're currently drowning in project documentation for compliance, request a demo. If you run 1-2 projects at a time, stick with simpler tools like Monday.com or ClickUp.
AI agent called 3,000 pubs to verify prices, changed an entire market
An AI system called 3,000 UK pubs to verify drink prices, ending the assumption that "no one will check" pricing claims. The automation made price verification economically viable at scale for the first time.
This is the business model shift to watch. Tasks that were too expensive to do manually (calling thousands of businesses, checking claims at scale, verifying compliance across suppliers) are now cheap enough to automate. If your business depends on competitors not checking your claims, that window is closing. On the flip side, if you compete against businesses that cut corners, you can now verify and call them out cheaply. Build this into your ops: what could you verify at scale that your competitors assume you won't bother checking?
Canva AI 2.0 helps creators build systems, not just content
Canva AI 2.0 shifts from speeding up individual design tasks to building reusable systems. Instead of generating one-off graphics, users can now create templates, workflows, and brand systems that generate variations automatically.
If you're still using Canva as a faster Photoshop, you're missing the point. The value is in building a system once (brand kit, templates, automated resizing for different platforms) and letting AI generate variations. For small businesses that need social media graphics, email headers, and ad variations weekly, this saves 5-10 hours per month. Set up your brand system properly and train one team member to use it. Stop paying freelancers $50-100 per graphic when you can generate consistent variations in-house.
Use AI to find bottlenecks in your own workflow, not just customer-facing tasks. Most small businesses automate customer service or content creation first, but the biggest time sinks are often internal: repetitive data entry, status updates, compliance documentation. Run an experiment this week: track how much time you spend on one recurring internal task (invoicing, project status reports, vendor follow-ups) and ask an AI assistant (Claude, ChatGPT, or a no-code tool like Zapier AI) to draft a workflow that automates 80% of it. Even if you only implement part of it, you'll learn what's actually automatable versus what needs human judgment.
The AI supply chain is fragmenting. Chinese AI companies relocating to avoid restrictions just lost their workaround, while Western cloud providers are doubling down on competing AI models simultaneously. For small businesses, this means picking tools with clear, stable funding and infrastructure (Claude, ChatGPT, Gemini) rather than betting on fast-growing startups with unclear ownership. Watch for power costs to pressure cloud pricing over the next 3-5 years as Meta and others scramble for renewable energy capacity. The immediate opportunity is in workflow automation that makes previously uneconomical verification tasks viable at scale.