OpenAI Reaches 10-Gigawatt Compute Goal Ahead of Schedule

OpenAI has announced that it achieved its goal of 10 gigawatts of AI compute capacity in the United States years ahead of its original schedule. The company has been rapidly scaling out its compute infrastructure through massive data center investments, and this milestone marks a major step toward securing the compute resources needed to train and run AI models.

The backdrop is surging demand for the company's services, including the GPT series and Codex, along with an industry-wide trend in which inference demand is beginning to outpace training. OpenAI's Sam Altman said last month that "we need to become an intelligence factory," underscoring that securing compute capacity sits at the heart of the company's strategy.

White House Rejects Expanded Access to Anthropic's "Mythos"

The White House has rejected Anthropic's plan to expand access to its AI model "Mythos" to roughly 70 companies. According to the Wall Street Journal, the rejection stems from concerns about insufficient compute resources.

The decision contrasts with the White House's move last month to restore access for Anthropic, a sign that debate over the allocation of AI compute resources continues within the government. With demand for compute outstripping supply, how much of the resource goes to which companies is becoming a policy flashpoint.

AI Investment by Big Four US Tech Firms to Hit $725 Billion a Year

Based on the latest quarterly earnings from Alphabet, Meta, Microsoft, and Amazon, the combined 2026 capital spending of the four biggest US tech companies is projected to reach as much as $725 billion. The bulk of the investment is going toward AI data center equipment.

Alphabet and Meta raised their full-year guidance, while Microsoft disclosed its spending outlook through the end of the year for the first time, matching Alphabet at $190 billion. Amazon held its figure at $200 billion, though a surge in first-quarter spending has squeezed its free cash flow. According to Bloomberg, AI demand has also turned memory chips into some of the highest-margin products in the world.

Alongside this scale of investment, questions persist over whether it is actually translating into revenue. The LA Times analyzed the progress on payback in an article titled "Is big tech's massive AI investment working?"

Tencent Releases 440MB Offline Translation Model

Tencent has released an open-weights AI translation model that runs fully offline on smartphones. The model is just 440MB in size, supports 33 languages, and claims translation quality that surpasses Google Translate.

Compact AI models that run on edge devices carry significant implications for privacy protection and for use in regions with limited connectivity infrastructure. The release is drawing attention as an example suggesting that the trend of running AI agents on personal devices is reaching a practical level.

Anthropic Releases a Bioinformatics Benchmark

Anthropic has released "BioMysteryBench," a benchmark showing that Claude can solve real-world bioinformatics problems at an expert level. While the results are promising, important caveats are also attached.

Efforts to evaluate the scientific applicability of AI models are meaningful for gauging just how practical AI is as a research tool. That said, the benchmark's design and evaluation criteria await independent verification.

FDA Launches Pilot for Real-Time AI Monitoring of Clinical Trials

The US Food and Drug Administration (FDA) has launched a pilot program for real-time monitoring of clinical trials using AI and cloud computing. It is positioned as part of an organizational rebuild following staffing cuts driven by DOGE (the Department of Government Efficiency), and the agency says it could substantially shorten the time to new drug approval.

Stanford HAI Releases 2026 AI Index Report

Stanford University's Institute for Human-Centered AI (HAI) has released the 2026 edition of its annual "AI Index Report." Widely cited across the industry as a comprehensive analysis of trends in AI, this year's edition is believed to cover falling inference costs, the rise of open-source models, and AI regulatory developments across countries.

---