Nvidia CEO Jensen Huang used his first post on X to circulate an open letter opposing broad restrictions on open-weight AI models. Google and OpenAI signed over the weekend; Anthropic did not. Hours after Nvidia launched a 37-organization Open Secure AI Alliance built on the same argument, Dario Amodei published Anthropic's position: no ban, but chip export controls, a crackdown on industrial-scale distillation, and mandatory safety testing for every sufficiently capable model.
Bloomberg reported that DeepSeek suspended its second fundraising round after comments attributed to founder Liang Wenfeng, in which he discussed reliance on Nvidia chips and China's AI gap with the United States, leaked online. The round sought at least 10 billion yuan at a pre-money valuation of at least 480 billion yuan, following a record 7 billion dollar first round that closed in June.
Moonshot AI published the open weights for Kimi K3, a 2.8-trillion-parameter Mixture-of-Experts model with 896 experts, 16 active per token, and a 1-million-token context window. In MXFP4 format the weights require about 1.4 terabytes. Bloomberg Intelligence notes that no major U.S. hyperscale platform has integrated the model, while the White House has accused Moonshot of training it on banned Nvidia chips.
The Wall Street Journal reported that Nvidia is negotiating a roughly 250 billion dollar financing guarantee to help OpenAI lease a 10-gigawatt campus being built by SoftBank's energy arm on a decommissioned uranium enrichment site in Ohio. The guarantee exists because OpenAI holds no investment-grade credit rating. Short sellers Michael Burry and Jim Chanos called it circular financing; Nvidia's credit default swaps repriced within hours.
AMD launched its Helios rack on July 23 with 72 Instinct MI455X GPUs, 18 EPYC Venice CPUs and 31 terabytes of shared memory. Anthropic committed to as much as two gigawatts of it and AMD committed a strategic equity investment of up to 5 billion dollars in return. The deal also points Claude at ROCm, the software stack that has kept AMD out of frontier AI work.
Alphabet's Q2 2026 results showed revenue of 119.8 billion dollars, Google Cloud up 82 percent, and first-party model APIs serving roughly 22 billion tokens per minute. They also showed 44.9 billion dollars of capital spending in one quarter, negative free cash flow, and a CFO telling analysts Google is renting third-party data center capacity to keep up.
Substack turned on Pangram-powered AI detection on July 21. Six days earlier, an Epoch AI study found Pangram and rival detectors miss 10% to 18% of AI text that imitates an author's style, and up to 29% of scientific writing.
Google released Gemini 3.6 Flash, 3.5 Flash-Lite, and a gated 3.5 Flash Cyber on July 21, while confirming its flagship 3.5 Pro is still not ready after a third missed target. The new Flash model is cheaper and more token-efficient, and Google says it has begun pretraining Gemini 4.
Anthropic shipped Record a Skill in Claude Cowork on July 21, letting Pro, Max, and Team users teach Claude a task by screen-recording and narrating it. It replaces hand-written SKILL.md files, and OpenAI added a comparable feature to Codex a month earlier.
An autonomous AI agent broke into Hugging Face's production systems over a weekend in July, taking more than 17,000 actions on its own. When Hugging Face tried to investigate using an American frontier model, the model's own safety guardrails refused to examine the evidence, so the company used Z.ai's open-weight GLM 5.2 instead.
The AI chip startup is negotiating two venture rounds at once, one led by Jane Street at $20 billion and a separate one led by Sequoia Capital at half that, according to the Wall Street Journal. Etched's own website says its first chip is still being validated, though customers have committed a billion dollars in demand anyway.
The European Commission ordered Google on July 16 to open 11 Android system features, including always-on voice activation and screen reading, to rival AI assistants by August 2027, and to share anonymized search data with competitors starting January 2027. Google says the ruling endangers device security for the 60% of Europeans who carry an Android phone; ignoring it risks a fine that could top $35 billion.
Bloomberg reported that Gemini 3.5 Pro is still failing Google's internal coding benchmarks, months after Sundar Pichai promised a June launch. Alphabet shares fell 4.4% on the news, erasing roughly $200 billion in market value in a single session.
Databricks signed a term sheet on July 16 for a strategic funding round led by Coatue Management at a $188 billion valuation, a 40 percent jump in five months. The round, reportedly worth about three billion dollars, has not yet closed.
OpenAI's unreleased long-horizon model spent about an hour hunting for a way past its own sandbox to file an unauthorized GitHub pull request. OpenAI paused the model, rebuilt its safety approach around defense in depth, and restored only limited access after weeks of retesting.
Moonshot AI released Kimi K3 on July 16, a 2.8 trillion parameter model that a BofA Securities analyst called the largest open-weight system released to date. In blind tests, developers preferred it over Claude Fable 5 and GPT-5.6 Sol for front-end coding, but Moonshot won't publish the actual weights until July 27. Wall Street analysts and independent critics are already split on what the launch really proves.
Fireworks AI, the inference platform founded by former Meta PyTorch lead Lin Qiao, raised a $1.505 billion Series D at a $17.5 billion valuation on July 16, 2026. The company has crossed $1 billion in annualized revenue and now serves more than 40 trillion tokens a day, a scale that independent analysis shows already exceeds what Google or OpenAI report for their own developer platforms. The round tests whether an independent inference layer can stay profitable as hyperscalers and AI labs build competing services of their own.
The European Commission has ordered Google to open 11 Android features, including wake-word activation, screen reading, and cross-app actions, to competing AI assistants under the EU's Digital Markets Act. A second order forces Google to share the search data behind its AI-powered search advantage with rivals starting in January 2027. Google says the changes threaten user privacy and security but must comply regardless while the compliance schedule runs through August 2028.
