Presented by
Good morning, AI enthusiasts. Microsoft just built a $2.5 billion business dedicated entirely to making sure AI actually works inside your company, and it's the fourth major tech giant to make this exact bet in a matter of weeks.
Amazon, OpenAI, and Anthropic have all launched something similar in the past two months, betting that AI engineering services, not the models themselves, is where the real enterprise money sits. Is Microsoft's head start with Fortune 500 customers enough to win a race everyone just entered at once?
In today's recap:
Microsoft launches $2.5B AI deployment business
Anthropic explores custom AI chip with Samsung
Alibaba bans Anthropic's Claude Code at work
Debug your website automatically with Safari's MCP server
4 new AI tools, prompts, and more
MICROSOFT
Microsoft launches $2.5B AI deployment arm
Microsoft
Recaply: Microsoft just launched Microsoft Frontier Company, a new AI deployment business backed by $2.5 billion, with 6,000 experts helping customers build and improve their AI systems.
Key details:
The new business mixes industry knowledge with hands-on AI engineering. It goes further than what's called Forward Deployed Engineering. Teams work directly with customers to design, build, and keep improving their AI systems over time.
Microsoft is putting in $2.5 billion and 6,000 experts. That's bigger than Amazon's plan. Amazon announced $1 billion for a similar AI team just two days before Microsoft's news.
Judson Althoff runs Microsoft's Commercial Business. He said the new team will be "the largest, most capable" engineering group in the industry, bigger than rival efforts.
Microsoft launched this week. It's already working with big clients like London Stock Exchange Group, Unilever, and Novo Nordisk. More partners like Accenture and EY will help it grow worldwide.
Why it matters: Microsoft thinks AI engineering, not just AI models, is where the real money is. It's not the only one. Amazon put $1 billion into a similar team just two days earlier. OpenAI and Anthropic have their own versions too. Microsoft already works with huge companies, so it has a head start here. But there's a catch. Microsoft says it will protect each customer's data and know-how. Whether that promise holds up is still an open question.
PRESENTED BY FIN
Scale AI support on AWS, see how July 9
Customer expectations keep rising. Support budgets don't. On July 9, Fin and AWS are hosting a live executive session on how leading enterprises close that gap: scaling AI-powered support while simplifying how they buy it.
You'll see how to resolve an average 76% of conversations with Fin on AWS enterprise-grade infrastructure, procure through AWS Marketplace to put committed cloud spend to work, and turn the Fin and AWS collaboration into lower support costs. Register for the live session to see how.
ANTHROPIC & SAMSUNG
Anthropic explores custom AI chip with Samsung
MLQ
Recaply: Anthropic just opened talks with Samsung to build a custom AI chip. The company hasn't decided yet what the chip will do or how powerful it needs to be.
Key details:
Anthropic wants to use Samsung's new 2-nanometer chip process and its advanced packaging. The goal is to rely less on Nvidia's dominant hardware.
OpenAI unveiled its own custom chip with Broadcom just last week. Samsung is also planning to spend $518 billion on four new chip plants in South Korea.
Anthropic recently hired Clive Chan, an engineer who previously worked on OpenAI's own chip project. He's expected to help build Anthropic's new chip team, according to The Decoder.
The chip project is still in its early stages, with no finalized design yet. Anthropic told TechCrunch this week that Nvidia, Google, and Amazon chips still matter most to its plans.
Why it matters: Samsung is playing both sides here. It already makes chips that Nvidia needs to train and run its own AI models, while using Nvidia's software to make its own chips. Now Samsung might also help Anthropic build a chip meant to cut Nvidia out. Every major AI lab wants its own chips, but almost none of them can build one without leaning on the very companies they're trying to move away from.
GUIDES
Debug your website automatically with Safari's MCP server
Recaply: In this tutorial, you will learn how to connect your coding agent to Safari's new MCP server, so it can see your site's console logs, network requests, and screenshots on its own.
Step-by-step:
Download Safari Technology Preview from Apple's developer site. Open Safari Settings, then Advanced, and turn on "Show features for web developers."
Go to Safari Settings, then Developer, and turn on "Allow remote automation and external agents" so MCP clients can connect to your browser.
If you use Claude, open your terminal and run the "claude mcp add" command with the Safari driver path to register the server.
Open your project and ask your agent, "Find bugs on my site in Safari," and it will check the DOM, console, and network on its own.
Ask follow-up questions too, like "How accessible is my site in Safari?" or "How does my site perform?" to catch more bugs.
Pro tip: The Safari MCP server runs only on your machine and never sends data to Apple, so it's safe to use on private projects.
TOGETHER WITH TABS
Two Minutes to Know What Slow Billing Is Costing You
Most SaaS finance teams know their billing process is slow.Most SaaS finance teams know their billing process is slow. Few know what it's costing them.
The Tabs Billing Lag Calculator puts a dollar figure on it in two minutes — benchmarked against top SaaS companies.
TOOLS
Trending AI Tools
🎨 Pocket - Meta's app for generating small interactive AI-made games and mini-apps from text prompts, free on iOS and Android
🤖 Venice AI - Privacy-first platform with 200+ AI models and no data retention, free tier plus paid plans for extra features
📊 GeneBench-Pro - OpenAI's new benchmark testing AI judgment on messy computational biology problems, 10 sample questions free on Hugging Face
⚙️ Safari MCP Server - Apple's free MCP server letting coding agents see and debug your site directly in Safari
NEWS
What Matters in AI Right Now?
Alibaba banned workers from using Anthropic's Claude Code at work. Staff must now use Alibaba's own coding tool, Qoder, instead. The ban follows Anthropic's claim last month that Alibaba illegally copied Claude's abilities. Reuters
Anthropic, Google DeepMind, and Meta are hiring psychology and philosophy experts. They want to study whether AI could become conscious. Anthropic says it's testing models for signs of "panic" and "anxiety." Financial Times
Meta launched Pocket, an app that lets people make small games using AI prompts. It grew out of Meta's earlier purchase of the Gizmo team. Gizmo had already reached 635,000 downloads before the switch. TechCrunch
Google introduced Gmail Live. It lets people search their inbox by voice, asking things like "what's my gate number?" The feature rolls out to Google AI Pro and Ultra users this summer. X
Palantir CEO Alex Karp slammed OpenAI and Anthropic's pricing on CNBC. He said something has "gone completely wrong" with how labs charge for AI. The comments came two days after Palantir and Nvidia launched a rival system that lets customers keep their own model weights. CNBC
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
