Blog

Nvidia’s quarter is now the single best read on whether the AI buildout is real — and it just doubled again. Whatever you think about the AI trade, this is the number every interviewer will assume you know. On August 26 Nvidia reported second-quarter fiscal 2027 revenue of $96.2 billion, up 18% from the prior quarter and up 106% from a year ago, with Data Center revenue of $89.0 billion, up 117%. Gross margin held at 75.0%. The company guided third-quarter revenue to $108.0 billion plus or minus 2% while explicitly assuming no Data Center compute revenue from China. Jensen Huang’s framing is the line to remember: “AI has reached its inflection point… Now, compute is revenue.” The strategically interesting disclosure is buried in the highlights: Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute financing platforms intended to mobilize over $500 billion of third-party capital for AI infrastructure over time, subject to definitive agreements. Read that alongside last week’s $105 billion Nvidia debt backstop: the chipmaker is systematically moving the financing of its own demand off its balance sheet and onto private credit’s.

Salesforce made Claude the default brain of its entire stack, and the market rewarded it with one of the largest one-day gains in the company’s history. If you are recruiting into enterprise software, tech strategy, or anything touching go-to-market, this is the week’s most consequential product decision: the biggest enterprise SaaS vendor just abandoned model-agnosticism. On August 26 Salesforce and Anthropic announced Claudeforce, which runs in both directions. Salesforce goes into Claude as a plugin with 37 prebuilt sales skills — meeting prep, deal health review, pipeline review — routed through Salesforce so business rules are enforced on every action. Claude goes into Salesforce as a reasoning model for the Atlas Reasoning Engine, powers Agentforce Vibes and Agentforce Coworker by default, and becomes the default model for Slack, Slackbot and Slack AI. Salesforce in Claude is in pilot now with open beta expected in September. The strategic claim Benioff is making is that the interface itself is disappearing: “the UI is the AI.” Shares rose roughly 23% on August 27 — though be precise about causality in an interview, because the move reflected a Q2 earnings beat and raised guidance alongside the partnership, not the Anthropic deal alone.

Bill Gates says the entry-level job squeeze is real, arrives first, and nobody has a plan — including him. This is the week’s future-of-work item, and it matters less as new evidence than as a signal about where elite opinion has moved: the most reliably optimistic technologist of the last thirty years is now arguing for limits. In an essay published August 26 on Gates Notes, running close to 6,000 words, Gates argues AI differs from prior automation waves because it performs cognitive work, spreads through infrastructure people already use, and requires no specialized skill to access. He names three risks: jobs vanishing first for young people in white-collar roles such as customer support and software engineering; criminals empowered by fraud, deepfakes and worse; and stunted child development from AI companions. He calls the coming period “one of the most turbulent times in human history” and writes that “there is not even a plan to have a plan,” proposing national coordinating bodies, a new international organization, and a “Human Reserved” category of work set aside for people. Treat this as argument, not evidence — Gates offers no new data and the causal question remains contested — but if you are interviewing this fall, expect to be asked what you think of it.

A federal judge ruled the Pentagon illegally blacklisted Anthropic for criticizing the administration. This is a live case study in political risk for any company whose product decisions carry a public position — and a reminder that “national security” is a contestable justification, not an automatic one. On August 27, U.S. District Judge Rita Lin ruled that the Defense Department’s designation of Anthropic as a supply-chain risk, and Defense Secretary Pete Hegseth’s order that military contractors not do business with the company, exceeded his authority and violated Anthropic’s First Amendment rights. Lin called the designation “baseless and illegal” and wrote that national security is “not a blank check to punish and retaliate against government critics.” The February ban followed Anthropic’s refusal to permit military use of its models for domestic surveillance or autonomous weapons. Two caveats worth carrying: the ruling does not require the Pentagon to resume working with Anthropic, and the government is expected to appeal.

OpenAI cut off Cursor because Elon Musk bought it, turning model supply into a competitive weapon. Supplier concentration risk just stopped being a theoretical slide in a strategy deck. On August 28 OpenAI notified SpaceX that it will wind down the contract supplying its models to the coding editor Cursor, with a proposed shutoff of November 12, 2026, after SpaceX acquired Cursor’s parent Anysphere earlier in the month. OpenAI’s stated reason is trust: it cited Twitter breaking contract terms after Musk’s acquisition and Musk’s sworn admission this year that xAI violated OpenAI’s terms of service. OpenAI says it is giving the maximum notice its change-of-control clause allows while withholding all future models from Cursor in the interim. The lesson for anyone building a product on someone else’s foundation model is blunt: your supplier’s view of your owner is now a live input to your product roadmap, and a change of control can terminate your core dependency.

Anthropic published a standard for letting AI agents run physical lab and factory equipment. Software agents have been stuck inside software; this is the plumbing that gets them onto a factory floor, and it is the kind of unglamorous standard-setting move that decides who captures value in a category years later. On August 27 Anthropic opened a research preview of the Model Hardware Standard (MHS), a specification that lets AI agents discover and operate microscopes, liquid handlers, robotic arms and similar instruments in parallel. Anthropic says MHS cuts integration work that normally takes weeks or months down to hours or minutes; it is model-agnostic, works with any device that has a programmable interface, and is accessible through the Model Context Protocol, with plans to open-source it. In a Genentech proof-of-concept, Claude orchestrated a liquid handler, robotic arm and plate reader to autonomously tune pipetting flow rates for a standard protein assay and recovered from tip-pickup and fluid-detection errors on its own — while still struggling, Anthropic notes, with physical intuition problems like bubble formation. Anthropic paired the release with an expansion of its science program, including 10,000 free and discounted Claude seats for researchers. Note that MHS is a vendor-authored standard in a research preview, and the performance claims are Anthropic’s and its partners’ own.