Anthropic has just eclipsed OpenAI in valuation, closing a staggering $65 billion Series H funding round that values the AI startup at $965 billion. The massive influx of capital arrives precisely as the company releases Claude Opus 4.8, a new flagship model boasting improved coding capabilities and a novel “Dynamic Workflows” tool for coordinating swarms of AI subagents.
The AI price war in China is reaching new extremes, even as a backlash brews against the unsustainable corporate costs of “tokenmaxxing.” Xiaomi has announced a permanent 99% price cut for its MiMo-V2.5 API, effectively commoditizing access to top-tier models despite its own Q1 operating profits tumbling by 59.5%. Meanwhile, major companies are starting to question the ROI of these massive token expenditures; a tech lead at gaming giant miHoYo recently revealed that testing a multi-agent NPC system burned through 2 million RMB ($275,000) of tokens in a single night, echoing Uber’s internal struggles to justify massive AI budgets that haven’t translated to proportionate consumer value.
The most fascinating watch today is the deep-dive interview with Xpeng’s CEO, He Xiaopeng, on his massive bet on “physical AI” and humanoid robots. He argues that applying digital AI paradigms to the physical world is a major mistake, and details why developing a general-purpose humanoid robot is up to 100 times harder than building an EV company. He Xiaopeng: Robot IRON’s Birth, The Accident, GX, His Big Bet & Swimming in Blood
This week, engineering organizations aggressively shifted away from unconstrained, single-agent architectures toward highly deterministic, platform-governed execution loops. A clear consensus emerged that scaling AI requires decoupling stochastic reasoning engines from strict, sandboxed execution environments, while simultaneously optimizing the underlying “boring machinery” of data pipelines to feed these models without bottlenecking real-time inference.
How Snapchat Serves a Billion Predictions Per Second · Snapchat Snapchat reduced its data plane costs by 10x and halved inference latency by transferring features as raw bytes and delaying deserialization until inside the inference engine. At the scale of a billion predictions per second, this proves that optimizing network transport and hardware-specific execution graphs (e.g., isolating dense matrix multiplications on GPUs while keeping embedding lookups on CPUs) is far more critical than tuning the ML model itself.
SK Hynix Inc. and Micron Technology Inc. have both surged past $1 trillion market valuations as an unrelenting artificial intelligence frenzy drives record demand for advanced memory chips. Driven by robust corporate earnings momentum, Goldman Sachs has lifted its year-end target for the S&P 500 Index to 8,000, helping power global equities to fresh highs.
The AI boom continues to create new titans, with South Korea’s SK Hynix and U.S. chipmaker Micron both surging past the $1 trillion market capitalization mark in today’s session.
Despite the AI rally pushing indices to fresh records, major institutions are urging caution, with Bank of America advising clients to prepare for a “summer correction” due to stretched technical indicators. Across the Atlantic, a top European Central Bank official echoed this sentiment, warning of elevated correction risks triggered by high market valuations, private credit vulnerabilities, and the ongoing conflict in Iran. Meanwhile, the Minneapolis Fed’s Neel Kashkari reaffirmed that defeating inflation remains the central bank’s priority, noting that persistently high prices could force tougher policy action while the labor market stays “in decent shape”. In the commodities space, U.S. crude oil fell below $90 following reports that an Iran framework agreement could restore traffic through the Strait of Hormuz within a month, helping drag the 10-year Treasury yield down to 4.465%.
According to a cnbeta report on Qwen3.7-Max, Alibaba’s latest AI model has stunned the global developer community by breaking into the top two of the coding arena, trailing only Anthropic’s Claude. Running 35 hours continuously without context degradation, Qwen3.7-Max cements China’s place as a definer, not just a follower, in the Agent AI race.
ByteDance plans to spend up to $70 billion on AI data centers this year and next, preparing to challenge top US AI companies globally. Meanwhile, Nvidia CEO Jensen Huang announced that Nvidia will invest $150 billion annually in Taiwan, cementing the island as the absolute center of the AI revolution.
Google has officially laid the groundwork for an autonomous agentic economy, announcing the Agent Payments Protocol (AP2) and the Universal Commerce Protocol (UCP). These frameworks provide standard languages and strict, tamper-proof digital mandates for AI agents to securely make purchases and execute transactions on a user’s behalf.
Matrix Multiplications on GPUs Run Faster When Given “Predictable” Data Matrix multiplications are supposed to be fully deterministic, executing the same number of operations and memory accesses regardless of the tensor’s contents. Yet, initializing matrices with zeros or ones yields measurably faster performance than using normally distributed random data. The culprit is dynamic switching power: predictable data minimizes transistor state flips, reducing power consumption and preventing the GPU’s Voltage Regulator Module from aggressively throttling clock frequencies under heavy load.
Nvidia CEO Jensen Huang announced the chipmaker will pour $150 billion a year into Taiwan to cement its position as the undisputed “epicenter of the AI revolution”. The massive investment, which includes a new headquarters slated to be operational by 2030, serves as a sharp reality check to the US government’s ongoing, highly subsidized efforts to onshore semiconductor manufacturing.