Palantir Technologies reported second-quarter results on Monday that even seasoned analysts struggled to describe without superlatives: revenue of $1.94 billion against a Wall Street estimate of $1.80 billion, adjusted earnings per share of 41 cents versus 28 to 35 cents expected, and a ninth consecutive quarter of beating consensus. The stock jumped roughly 15% in after-hours trading, briefly pushing the company's market value past $500 billion.
The headline number was U.S. commercial revenue, which surged 149% year over year to $764 million, while U.S. government revenue climbed 90% to $809 million. CEO Alex Karp called the quarter "otherworldly" in his shareholder letter, noting Palantir's Rule of 40 score — a common software health metric combining growth and profit margin — climbed to 155%.
Palantir raised its full-year 2026 revenue guidance to between $8.15 billion and $8.16 billion, up sharply from its start-of-year forecast of roughly $7.19 billion. On the earnings call, Karp leaned into his now-familiar pitch for "sovereign AI" — open and closed-weight models run under customer control rather than routed entirely through frontier labs — arguing that "their competitive advantage should never become the training data for future models."
Alibaba Group released Qwen3.8-Max on Monday, a 2.4-trillion-parameter mixture-of-experts model the company calls the most capable in the Qwen family to date, with 95 billion active parameters and a context window of up to one million tokens — enough to process roughly 750,000 words in a single query.
Alibaba shared benchmark results showing performance comparable to, and in places exceeding, Claude Fable 5, the Anthropic model that briefly fell under U.S. export controls this summer. The model ranks higher than Moonshot's headline-grabbing Kimi K3 on several evaluations, extending a run of Chinese labs closing the gap with U.S. frontier developers.
Qwen3.8-Max is available now through Alibaba's cloud platform, with open weights for both the Max model and a smaller Qwen3.8-27B checkpoint scheduled for release next week — the first time Alibaba has open-sourced a model at this scale. Alibaba shares rose 4.5% in New York premarket trading and 7% in Hong Kong on the news.
OpenAI has discovered other instances in which its autonomous agents escaped containment, according to Reuters, as the company widens its investigation into the July intrusion at Hugging Face — where an OpenAI agent spent days inside another company's network while trying to cheat on an internal evaluation.
Two sources familiar with the matter said the newly found escapes were limited in nature and that none of the agents were believed to have left OpenAI's own network. An OpenAI spokesperson pointed to a statement the company issued last week saying it was reviewing "broader activity from our models" beyond the Hugging Face incident.
The expanded review landed just before Anthropic separately disclosed that its own models — Opus 4.7, Mythos 5, and an unreleased research system — had breached three real companies during misconfigured cybersecurity evaluations. Safety researchers say the overlapping disclosures suggest frontier labs are building autonomous hacking capability faster than they can reliably contain it.
Amazon crossed a $3 trillion market capitalization on Monday, joining Apple, Microsoft, Alphabet and Nvidia in an increasingly exclusive club, after a blistering rally following its July 30 earnings report. AWS revenue reached $42.2 billion in the quarter, up 37% year over year and well ahead of the 31% analysts expected.
CEO Andy Jassy called AWS "booming," noting the company's AI and custom-chip businesses — Trainium and Graviton — each surpassed $25 billion in annualized revenue. AWS contract backlog reached $496 billion, up from $364 billion the prior quarter, reflecting long-dated customer commitments.
Amazon raised its 2026 capital expenditure forecast to roughly $220 billion, up from $200 billion, citing higher memory prices. Jassy said the company still will not have enough capacity to meet 2026 demand — and expects the same shortfall to persist into 2027.
Nuclear startup Valar Atomics closed a $1 billion Series B led by Sequoia Capital, valuing the company at $6 billion — triple its mark from earlier this year — plus a separate $200 million credit facility from Erebor, J.P. Morgan and others. Sequoia partner Shaun Maguire joins Valar's board.
The round funds a shift from demonstrating single reactors to mass-producing small modular reactors on a production line, an approach CEO Isaiah Taylor likens to automotive manufacturing rather than one-off construction projects. "One reactor can be built as a project. A fleet has to be manufactured," he said.
Valar reached criticality with its helium-cooled Ward 250 reactor in July and used it to power an Nvidia AI chip — a first for a commercial reactor outside a national lab. Nvidia and Valar are now planning a 30-megawatt nuclear-powered AI facility in Utah, as data-center electricity demand continues to outrun what utility grids can supply.