TL;DR

Trump accelerates U.S. AI buildout, bars “woke” AI in contracts; GPT-5 nears release; Nvidia chips breach China curbs.

Highlights

• Trump administration unveils a 28-page AI Action Plan, signing executive orders to accelerate U.S. data center and semiconductor buildout, roll back federal oversight, and boost AI exports 1.
• New rules bar “woke” or ideologically biased AI from federal contracts, requiring models to be “ideologically neutral” 6.
• U.S. to supply full-stack AI systems to allies while tightening export controls on advanced chips to China; $1B in Nvidia chips reportedly reached China via intermediaries despite curbs 2.
• Trump drops idea of breaking up Nvidia , instead allowing sales of lower-tier H20 accelerators to China; Nvidia’s dominance remains under scrutiny 3.
• Alphabet beats Q2 estimates with $96.43B revenue, raises 2025 capex outlook to $85B to support AI/cloud demand 4.
• OpenAI’s GPT-5 in advanced testing, described by CEO as “smarter than humans”; release timing remains uncertain, with Gemini 2.5 as main competitor 5.
• Tesla targets end-of-2025 rollout for unsupervised Full Self-Driving in U.S. cities and sets 2026 mass-production goal for Optimus humanoid robot 78.
• AMD CEO notes U.S.-made TSMC chips cost 5–20% more than Taiwan-made; strong AI chip demand from OpenAI and Musk-linked firms 9.
• Meta unveils non-invasive AI-powered wristband for muscle-based computer control, targeting assistive tech and AR/VR 10.
• Oracle partners with Bloom Energy to deploy on-site fuel cells for U.S. AI data centers, addressing energy and sustainability needs 11.
• FDA’s generative AI tool Elsa faces scrutiny for hallucinating drug approval studies, raising concerns about AI reliability in regulatory settings 12.
• Perplexity AI launches $200/month Max plan for power users, expanding premium offerings amid rising cloud costs 13.
• Pew study finds Google AI Overviews halve link click-throughs; Google disputes impact on publisher traffic 14.
• Walmart hires Instacart ’s Daniel Danker to lead AI overhaul, consolidating dozens of systems into four unified “super agents” 15.

Commentary

The Trump administration’s AI Action Plan marks a significant deregulatory shift, prioritizing rapid buildout of U.S. data center and semiconductor capacity through streamlined permitting, tax incentives, and federal land access 1. The policy also introduces categorical exclusions from environmental review for large-scale projects and seeks to position the U.S. as a leading AI exporter, with a focus on supplying allies while tightening controls on adversaries 1. The new federal procurement rules—barring AI deemed “woke” or ideologically biased—will reshape vendor strategies and model development for government contracts, with compliance guidance pending across agencies 6.

Export enforcement remains a challenge: despite tightened controls, $1 billion in Nvidia ’s latest AI chips reportedly reached China via intermediaries, highlighting persistent gaps in global supply chain oversight 2. The administration’s nuanced approach—permitting lower-tier chip exports while restricting top-end models—aims to maintain U.S. leverage in the AI hardware stack, even as Nvidia ’s market dominance continues without imminent antitrust action 3.

On the enterprise side, Alphabet ’s robust Q2 results and increased capex signal sustained demand for AI infrastructure 4, while Oracle ’s move to on-site fuel cells addresses the growing energy needs and sustainability pressures of hyperscale AI data centers 11. AMD ’s comments on U.S. chip cost premiums reinforce ongoing supply chain and margin challenges for AI hardware buyers 9.

In the AI model race, OpenAI’s GPT-5 is nearing release, with claims of superhuman reasoning, though reliability and deployment timing remain open questions 5. Tesla is pushing toward unsupervised FSD and scaling its Optimus robot program, further blurring lines between software, hardware, and automation 78. Meanwhile, Meta ’s muscle-sensing wristband and Walmart ’s system consolidation reflect ongoing efforts to integrate AI into new interfaces and enterprise workflows 1015. However, the FDA’s Elsa tool highlights unresolved risks in deploying generative AI for critical regulatory functions, underscoring the need for robust validation 12.