Alibaba bets on chips

Meta struggles with scale

vent

Welcome back! Taco Bell is rethinking where and when to lean on AI. After rolling out voice AI ordering at 500+ locations, viral clips started surfacing—like one customer “ordering” 18,000 water cups just to bypass the bot and reach a real person.

Some locations may still go all-in with AI, but others might stick with human cashiers during busy hours.

What do you think about AI in drive-thru?

Login or Subscribe to participate in polls.

Alibaba Bets Big on AI Chips and Cloud Growth

Eddie Wu intends to transform Alibaba for the AI era / Photo credit: Alibaba Group

Alibaba is doubling down on its cloud computing ambitions and developing a new AI chip to power next-gen infrastructure. Its cloud division revenue grew 26% year-over-year in the June quarter, outpacing previous quarters and marking a bright spot for the tech giant.

The chip details: Alibaba’s latest chip is designed for inference workloads, where models create predictions or generate answers in real time. The chip will reportedly be compatible with Nvidia software and manufactured using a 7 nanometer process.

Investor reactions: Shares surged nearly 19% in Hong Kong trading after Alibaba’s earnings release, building on a 13% gain in New York the week prior. While overall revenue rose just 2% year-over-year to $34.7 billion, investors are buying into Alibaba’s AI-fueled future—particularly as the company leans into its cloud unit and chip innovation.

The takeaway: Control of the chips and infrastructure behind AI models is becoming just as critical as software. As Microsoft, Google, and Nvidia expand their silicon strategies, Alibaba is positioning itself as a major player in Asia’s AI race—one where hardware, not just algorithms, determines who leads.

Meta’s $14B Scale AI Deal Shows Early Cracks

Meta’s partnership with Scale AI, once touted as a cornerstone of its superintelligence ambitions, is looking shaky. Just two months after investing $14.3 billion and bringing Scale CEO Alexandr Wang and top executives on board, Meta is diversifying its data labeling vendors and leaning on competitors like Mercor and Surge.

How they got here: In June, Meta poured billions into Scale to secure high-quality labeled data—the lifeblood of large AI models—and lure talent to its newly formed Meta Superintelligence Labs (MSL). Wang was tapped to lead the effort, and several Scale execs followed.

What’s driving the tension: Inside MSL, researchers reportedly view Scale’s data as lower quality compared to competitors, preferring to work with Surge and Mercor. Newcomers are frustrated navigating Meta’s bureaucracy (one Scale executive left Meta after two months), while longtime GenAI staff see their roles diminished.

The big question: Meta’s largest AI investment to date was meant to help it catch up to OpenAI and Google. Does Scale still deliver strategic value—or did Meta make the wrong bet on its path to AI dominance?

China Calls for AI Cooperation on a Global Stage

At the Shanghai Cooperation Organization (SCO) Summit, Chinese President Xi Jinping urged member nations to deepen collaboration on AI while rejecting a “Cold War mentality” in tech. Speaking to more than 20 world leaders, including Russia’s Vladimir Putin and India’s Narendra Modi, Xi pitched AI as a shared opportunity rather than a zero-sum race.

“Global governance has come to a new crossroads,” Xi said. “The house rules of a few countries should not be imposed upon others.”

What’s driving this: China is positioning itself as a global AI convener, investing $84 billion across SCO nations and launching initiatives in new energy, digital economy, and tech innovation. The summit also highlighted China’s warming ties with India, signaling a potential realignment of alliances as AI infrastructure becomes increasingly strategic.

Why it matters: While Washington doubles down on export controls and domestic chip manufacturing, Beijing is leveraging multilateral frameworks like the SCO to pull emerging economies into its AI orbit.

Keep Humans in the AI Loop

HumanLayer

HumanLayer is an AI governance API that builds human oversight into automated AI workflows at critical decision points. It lets AI agents pause before high-stakes actions, ensuring trust and control without sacrificing automation.

How you can use it

  • Add human-in-the-loop checks for business-critical workflows.

  • Get real-time notifications via Slack, email, or Discord.

  • Set up requests for approval, feedback, or expert input.

  • Integrate seamlessly with LangChain, CrewAI, LlamaIndex, or OpenAI.

Pricing: Free & Paid

Turn Text into Emotional Narration

Copilot Audio Expressions

Copilot Audio Expressions is Microsoft’s experimental AI tool that converts text into high-fidelity, emotionally expressive audio narration in seconds. Using the MAI-Voice-1 model, it lets you pick from nearly a dozen synthetic voices and choose between conversational emotive mode or narrative-focused story mode.

How you can use it:

  • Generate natural-sounding narration for videos, podcasts, and prototypes.

  • Preview and download MP3s instantly without logging in.

  • Add emotional nuance to presentations or accessibility features.

  • Experiment with expressive storytelling using multiple voice styles.

Pricing: Free

Jobs, announcements, and big ideas

  • Anthropic raises $13B in Series F, hitting a $183B valuation.

  • Marketing is a game. We help you play it smarter. Masters in Marketing is your cheat code for fresh insights, expert advice, and strategies that actually work. Subscribe today.*

  • Amazon debuts Lens Live, an AI-powered visual search tool for instant product discovery.

  • Reese Witherspoon calls on women to embrace AI as the future of filmmaking.

  • OpenAI adds parental controls and sensitive chat routing to improve ChatGPT safety.

  • ElevenLabs launches v2 SFX model with higher quality and seamless looping.

  • Tencent unveils HunyuanWorld-Voyager, the first open-source ultra-long-range world model.

*This is sponsored advertising content.

Your new Photoshop cheat code. I show you how to use Nano Banana to turn complex image edits into one-click prompts (for free!)

That’s a wrap! See you on Friday.

—Matt (FutureTools.io)

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