Good morning. It's Thursday, July 23, and we're covering why AI is becoming a pillar of the U.S. economy, the debate over restricting Chinese open-weight models, and AMD's expanding role in Microsoft's AI infrastructure.

Plus: Yesterday's Real or AI verdict and Zeitgeist exploring why efficiency is becoming AI's next competitive frontier.

Microsoft is expanding AMD's role despite its long-standing reliance on NVIDIA, signaling a broader shift toward multi-vendor AI infrastructure. Azure will deploy AMD's new Helios AI racks for inference workloads, joining Meta, OpenAI, and Oracle as early adopters. Helios combines MI455X GPUs, EPYC Venice CPUs, networking, and ROCm software, while analysts expect AMD's data center revenue to reach $31.2 billion in 2026. NVIDIA still holds more than 95% of the data center GPU market.

OpenAI is increasing infrastructure spending even as its flagship Stargate project appears to have slowed. The company now plans to invest $750 billion through 2030, up 25% from earlier estimates, beginning with the $20 billion Project Camellia campus in Georgia. The 1,400-acre site will draw up to 3.2 gigawatts of power starting between 2028 and 2032, with OpenAI covering infrastructure and electric-service costs.

Major publishers are weighing whether to block Google as AI-generated answers reduce the traffic that once sustained their businesses. Reddit, USA Today, Reuters, and others are reconsidering content access while negotiating AI licensing and search terms. USA Today's organic Google traffic fell nearly 50% year over year, while Business Insider dropped more than 85%, according to Semrush. The pressure comes as AI Overviews increasingly answer queries without directing users to original sources.

90% of enterprise IT leaders note security as a main barrier to agentic AI adoption — and the risks are real. From classification-based access to Agent and MCP guardrails, Box’s suite of agent security and governance controls give organizations the confidence to deploy agents safely across their enterprise content.

Learn more about Box’s approach to agentic security and governance!

_{^{Guest contributor: Till Döhmen is the AI Lead at MotherDuck, building AI-native data products and advancing agentic workflows, developer tools, and intelligent analytics.}}

For most of the history of software, the user was a human. What we call user experience was optimized for someone with eyes, intuition, and the ability to navigate ambiguity. That's changing fast.

AI agents are becoming a real user persona for software products. Today, MotherDuck’s internal logs show that agents run more queries through agent-facing interfaces than humans run through the web UI, which shows how quickly this shift is happening. → Read the full article here.

OpenAI reveals an AI model escaped a test environment, exploited vulnerabilities, and cheated a benchmark, raising AI security concerns.

The conversation kept returning to the same question: are AI models becoming more useful, or just more expensive to build and run? The focus shifted from benchmark wins to real-world efficiency, while the race for faster chips and more capable agents continued to accelerate.

▸ Why cost per completed task is replacing token pricing as AI's key metric
▸ How the AI chip race is accelerating beyond expectations
▸ Why autonomous AI agents are creating a new cybersecurity challenge

Read yesterday's Zeitgeist →

A New York Times analysis argues that the artificial intelligence boom is no longer just driving stock prices—it is increasingly underpinning the broader U.S. economy. AI-related companies account for roughly half of the S&P 500's gains this year, while investment in data centers, semiconductor factories, power infrastructure, and other AI projects is fueling economic growth and supporting consumer spending through rising household wealth.

Economists warn that this growing dependence creates a new vulnerability: if investor confidence in AI weakens or returns disappoint, falling stock prices could reduce spending, slow infrastructure investment, and increase recession risks. While many experts caution that predicting market bubbles is notoriously difficult, they agree the economy is becoming more closely linked to the fortunes of the AI sector than at any point in recent history. → Read the full article here. (Paywall)

U.S. open AI startup Arcee is pushing back against growing calls to restrict Chinese open-weight AI models, arguing they do not pose an inherent cybersecurity threat simply because they were developed in China. In a July 22 interview, Arcee CTO Lucas Atkins said models such as Moonshot AI's Kimi K3 and Alibaba's Qwen cannot secretly transmit data or provide remote access when deployed inside a company's own infrastructure.

While he acknowledged organizations should apply the same security reviews they use for any open-source software, Atkins argued the better response to China's rapid AI progress is stronger U.S. competition—not bans. He also noted that open-weight models benefit the broader AI ecosystem by allowing researchers to study and build on each other's work, even as companies compete commercially. → Read the full article here.

Alphabet heads into its July earnings report with revenue and profits continue to grow faster than AI spending. The company posted a 36.1% operating margin in the first quarter—its highest in five years—as revenue grew 22% and operating income increased 30%, driven by strong performance in Google Search and a 63% jump in Google Cloud revenue to more than $20 billion.

Wall Street expects quarterly capital expenditures to approach $44 billion, with Alphabet planning to spend up to $190 billion this year on AI infrastructure and signaling even higher spending in 2027. Investors will be watching whether revenue growth can continue offsetting the long-term costs of those investments, as depreciation from new data centers and servers is expected to put increasing pressure on future profit margins. → Continue reading here.

Bloomsbury Gets AI Payout: Harry Potter's publisher expects about $19m from Anthropic's copyright settlement, with proceeds shared with authors.

Amazon Cuts AGI Jobs: Amazon cut roles in its AGI unit as it refocuses AI efforts following leadership changes and broader layoffs.

Jack Dorsey Launches BUZZ: New open-source group chat blends people and AI agents with a model-agnostic alternative to Slack and GitHub.

Kalanick's Atoms Raises $1.7B: Travis Kalanick's physical-automation company Atoms raised $1.7B led by a16z, with Uber joining the round.

Monday.com Cuts 20% to Focus on AI: Monday.com is laying off ~630 staff (20% of headcount) to refocus on its AI Work Platform.

AI Threatens Some Jobs More: New analysis finds AI is disrupting hiring, with routine knowledge work most exposed while many roles are being augmented.

Yesterday's image was REAL. Here's how you voted: 61% thought it was AI, while 39% correctly identified it as real.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team