Salesforce bets $8B on agentic

And Nvidia strips down

Welcome back, everyone! Starting today with some wild news: There are now 200,000 of us hanging out every Wednesday and Friday here at the Future Tools newsletter. I’m grateful to each and every one of you for hitting subscribe, opening these emails, and sharing your thoughts with me. The future of this industry is bright and I’m lucky to get to talk about it with smart folks like you.

And now that we’ve hit 200K…why not send this newsletter to a friend or coworker? 300K readers sounds pretty cool.

Now? Let’s get to the big stories of the week.

Salesforce’s $8B data power play is all about agentic AI

Via WIRED

Salesforce dropped a whopping $8 billion to acquire Informatica, one of the oldest names in enterprise data. The deal positions Salesforce to enhance its backend data stack at a time when every enterprise player is racing to make their data AI-ready.

  • The agentic AI angle: Informatica recently unveiled a slate of agentic AI offerings aimed at automating core data workflows (ingestion, transformation, metadata management) with minimal human input. 

  • The upside: By acquiring Informatica, Salesforce gets access to a modern, mature data architecture built to support AI agents at scale. Combined with its existing tools like Data Cloud and MuleSoft, Salesforce is building a full-stack platform where data becomes the fuel and interface for AI.

The bigger picture: Salesforce is betting that whoever controls data quality, lineage, and governance will control the AI layer, too—and this move is its clearest shot yet at owning that foundation.

Nvidia’s low-cost Blackwell chip signals major China push

All eyes are on Nvidia this week. The company will report its earnings in just a few hours. And one thing we know so far? Nvidia is releasing a stripped-down AI chip for China, expected to enter production in June. Price tag? $6,500–$8,000, compared to the $10K–$12K H20 model now restricted by US export bans.

The new playbook:

  • Lower specs, legal limits: The new chip swaps out advanced high-bandwidth memory for simpler construction, staying just under US thresholds.

  • Massive losses to recover: Nvidia’s CEO said export bans cost the company $5.5 billion in inventory write-downs and forced it to walk away from $15 billion in potential sales.

  • Pressure from Huawei: With Huawei’s Ascend 910B gaining ground, Nvidia needs to hold the line—not with performance, but with ecosystem lock-in. CUDA is still its biggest moat.

Looking to the future: By staying inside the regulatory lines and retaining developer mindshare, Nvidia’s trying to make sure it still owns the lane when the rules shift again.

Meta and Google vets want to crack the code to spatial AI

A new AI startup wants to make 3D worlds as easy to create as a text prompt. SpAItial, founded by Synthesia co-founder Matthias Niessner, just raised a $13 million seed round to tackle one of AI’s toughest technical frontiers: generating photorealistic, interactive 3D environments from scratch.

The technical challenge: Niessner’s team includes veterans from Google’s Beam and Meta’s 3D asset teams, and they’re going after what he calls “behavioral realism,” or environments that not only look real but behave like the real world. The goal is to build dynamic spaces that support interaction, physics, and agency. 

Why it matters: The first company to solve text-to-3D with usable interfaces could lay the groundwork for AI-native game engines. If vision models defined the past two years, spatial models may define the next decade.

Your entire content team, in one tool

Blaze AI

Blaze AI helps marketers generate, repurpose, and design content across channels while keeping brand voice consistent and SEO in check.

How you can use it:

  • Repurpose blogs into social posts, carousels, and newsletters

  • Maintain tone and design standards across formats

  • Eliminate tool-hopping with one unified platform

Pricing: Starts at $34/month

AI agents that test your app for you

Posium

Posium uses multiple coordinated agents to handle the entire QA lifecycle, from discovery to test planning, code generation, debugging, and maintenance.

How you can use it:

  • Identify flaky tests and auto-fix issues before they ship

  • Adapt test coverage as your app evolves

  • Run tests across browsers, devices, and frameworks from one dashboard

Pricing: Free plan available; paid starts at $80/month

Jobs, announcements, and big ideas

  • Claude gets a voice as Anthropic launches mobile voice mode.

  • China hosted a robot kickboxing match with humanoid fighters.

  • UN study finds women are three times more likely to lose jobs to AI.

  • WordPress launches a new AI team to future-proof the platform.

  • The Browser Company explains why it paused development of Arc browser.

  • Mistral releases Agents API to help developers build AI-powered apps.

Did the AI race just hit its iPhone moment? I break down why Veo 3, Claude 4, and OpenAI’s latest drops have everyone in the industry losing their minds.

That’s a wrap! Here’s one more big question for the road: Should AI be opt-in only?

Nick Clegg, Meta’s former global affairs chief, said at a UK event that asking artists for permission to train AI models on their work would “basically kill the AI industry” overnight. But opponents say it’s theft. 

The battle playing out in Parliament (with big names like Elton John and Dua Lipa pushing for disclosure rules) raises a tough question: Should scale excuse consent?

Hit reply and tell me what you think.

—Matt (FutureTools.io)

P.S. This newsletter is 100% written by a human. Okay, maybe 96%.