Nvidia Breaks Records but Leaves Wall Street Wanting More

Nvidia’s July quarter revenue hit $46.7 billion and net income jumped 59% to $26.4 billion. The company projected $54 billion for the current quarter, only slightly above analyst expectations.

Data-center sales grew 56% to $41.1 billion but narrowly missed estimates, marking the second straight shortfall for the unit that drives 89% of revenue. The soft guide and segment miss shaved nearly 3% off the stock after hours even as Nvidia unveiled a $60 billion share-repurchase plan.

Huang said hyperscalers may pour $3-$4 trillion into AI infrastructure within five years and claimed Nvidia could secure up to 35% of that spend. Export curbs on its H20 chips already chopped $4 billion from the quarter and forced the company to assume zero H20 revenue for the next period.

Generative AI automates 30–60% of repeatable legal work, fracturing the economic base of law firms that equate hours with value. A $1.5 billion firm with $500 million in such billings risks losing about $250 million as diligence, drafting, and research collapse to minutes.

Adopting AI slashes billable hours and revenue, while resisting it drives clients to faster, cheaper rivals. Partners face compensation fights, traditional utilization metrics break, and most firms lack the retained earnings to fund multi-year AI programs.

General counsels already demand AI-driven discounts and threaten 20% fee cuts for non-compliance. The piece estimates AI-native, fixed-fee providers can unlock a $72 billion market across 32 million underserved businesses.

Big investors back startups for outsized return potential. But regular investors have to wait for companies to go public before they can invest. Thanks to regulatory updates, some companies are doing things differently.

Take Revolut. In 2016, 433 regular people invested an average of $2,730. Today? They got a 400X buyout offer from the company, as Revolut’s valuation increased 89,900%.

No wonder 10K+ regular people are taking the chance on Pacaso. Founded by a former Zillow exec, Pacaso’s vacation home co-ownership tech earned $110M+ in gross profit to date. They even reserved the Nasdaq ticker PCSO. But you don’t have to wait to invest.

Anthropic has released a research preview of Claude for Chrome, its first extension that lets the AI read pages, click buttons, and fill forms directly in the browser. The controlled pilot invites 1,000 Claude Max subscribers to join a wait-listed test before broader rollout.

Internal trials show the agent can handle tasks like scheduling meetings, drafting emails, and filing expense reports, but also revealed vulnerabilities to prompt-injection attacks. Newly added permissions, confirmations, and content filters cut the attack success rate from 23.6% to 11.2% overall and to 0% on four browser-specific attack types.

Anthropic argues that real-world testing is the fastest way to uncover novel threats and harden the agent’s defenses. It says it shares any safety findings with developers building browser agents on its API.

Google released Gemini 2.5 Flash Image, a new image generation and editing model accessible now through the Gemini API, Google AI Studio, and Vertex AI. The preview is live today, with a stable release promised in the coming weeks.

The model blends multiple images, keeps character details consistent, and performs pinpoint edits via plain language prompts. Pricing is set at $30 per million output tokens—about $0.039 per image—and the updated AI Studio “build mode” lets developers deploy custom apps or export code to GitHub.

Gemini 2.5 Flash Image taps the broader Gemini knowledge base to interpret diagrams, answer real-world questions, and obey complex editing instructions in a single step. Every output carries an invisible SynthID watermark, underscoring Google’s bid to label AI-generated visuals by default.

LinkedIn cofounder Reid Hoffman has published Superagency, a book arguing that artificial intelligence amplifies personal decision-making power. He presents its thesis in a newly re-aired hour-long interview on Every’s AI & I podcast.

The book speaks to both skeptics and builders, urging them to make agency a core design value and advocating “private commons” to give individuals greater control over their data. Hoffman links today’s AI anxieties to historical fears around the printing press, showing how perception shapes whether technology feels liberating or limiting.

The discussion shifts the AI conversation from job displacement to the scope of control users gain. By grounding the debate in centuries of tech adoption patterns, it casts human perception of agency as the factor that will determine AI’s societal role.