The Memory Crunch Finally Reaches Your Cart
Apple raised prices across roughly fifteen product lines this week — Macs, iPads, the Vision Pro, HomePod, and Apple TV — the most sweeping mid-cycle price action in the company's modern history. The MacBook Neo jumped from $599 to $699, the MacBook Air rose to $1,299, and the 14-inch MacBook Pro climbed $300 to $1,999.
Outgoing CEO Tim Cook told The Wall Street Journal the situation had become unsustainable, describing the global memory and storage shortage as a "hundred-year flood" unlike anything he'd seen in over four decades in the business.
The root cause sits upstream. DRAM and high-bandwidth memory suppliers are redirecting capacity toward AI data centers, where demand has become effectively bottomless. Counterpoint Research pegs the price increase at roughly fourfold over the past three quarters, and memory maker Micron's gross margin has rocketed from 39% to nearly 85% — eclipsing both Nvidia and Meta.
Analysts read the move as a preview, not an isolated event. With the iPhone untouched for now, expectations are building for a price increase when the iPhone 18 launches this fall — DRAM that cost Apple roughly $39 per unit in last year's Pro model could run closer to $145 this time around.
Washington Puts a Leash on GPT-5.6
The White House's Office of the National Cyber Director and Office of Science and Technology Policy asked OpenAI to release GPT-5.6 only to a small group of government-approved partners, rather than the general public, citing the model's advanced capabilities.
CEO Sam Altman told staff the government would be approving access "customer by customer" during a preview period, with a broader rollout expected in a couple of weeks. Commerce Secretary Howard Lutnick reportedly pressed Altman directly to hold off until relevant agencies signed off.
The intervention follows the Commerce Department's June 12 export-control order forcing Anthropic to pull its Fable 5 and Mythos 5 models offline for foreign nationals, after officials judged both labs' newest models to carry comparable cybersecurity capability.
Altman was blunt that the arrangement isn't one he wants to repeat for future launches — even as he agreed to it for now. The episode lands without any finished federal framework: Trump's AI security executive order calls for a voluntary 30-day pre-release review, but the mechanics remain unbuilt weeks later.
Anthropic Tells the Senate: Alibaba Ran the Biggest Claude Heist Yet
In a June 10 letter to the Senate Banking Committee, Anthropic accused operators tied to Alibaba's Qwen lab of running what it called the largest known distillation attack against Claude to date — nearly 29 million exchanges over six weeks through roughly 25,000 fraudulent accounts.
Anthropic's head of policy told senators the campaign was carried out "illicitly, systematically, and at industrial scale" to harvest U.S. AI capability without bearing the training cost involved.
The scale dwarfs Anthropic's prior disclosures. In February, the company flagged a combined 16 million exchanges across DeepSeek, Moonshot AI, and MiniMax. The alleged Alibaba operation alone exceeds that combined total by nearly 75%.
Two days after the letter, the Commerce Department issued the export-control order pulling Anthropic's own Fable 5 and Mythos 5 offline for foreign nationals — a timing coincidence fueling debate over whether national-security framing is settling a commercial dispute Congress, not courts, is now being asked to referee.
From Fortnite to Forklifts: $320M for an AI That Learned to Act by Watching Gamers
General Intuition closed a $320 million round at a $2.3 billion valuation, led by Khosla Ventures with Jeff Bezos, Eric Schmidt, and General Catalyst joining in — just three months after a $134 million seed.
The startup, spun out of gaming clip platform Medal, trains its models on hundreds of millions of hours of gameplay footage carrying exact button-press action labels — not just video — teaching spatial and temporal reasoning that transfers from virtual worlds into physical ones.
The proof point: the same underlying model that's been playing a Fortnite-style game for over 100 hours straight also drives a quadrupedal robot that learned to navigate an unfamiliar office after just eight minutes of real-world fine-tuning.
CEO Pim de Witte has turned down a reported $500 million data-acquisition offer from OpenAI and multiple later approaches, betting that the company's action-labeled gameplay dataset makes it a foundational model provider for robotics and autonomous systems rather than an acqui-hire target.
A Physicist Says Microsoft's Quantum Breakthrough Might Just Be Noise
University of St Andrews physicist Henry Legg published a formal critique in Nature arguing that Microsoft's February 2025 Majorana paper relied on a flawed detection protocol and selective presentation of data that, in raw form, looks more like disorder than proof of a topological qubit.
Legg's analysis says Microsoft's plotting code was set to highlight only the most favorable region of its results, and that when reviewers asked about other regions, the company said it had examined only the one that passed its own protocol.
Microsoft is standing by its work. Technical Fellow Chetan Nayak said the company remains confident in its roadmap and pointed to continued evaluation under DARPA's Quantum Benchmarking Initiative, which has moved Microsoft into its final assessment phase.
It's the third round of controversy for Microsoft's two-decade bet on topological qubits, following two prior Nature retractions in 2018 and 2021 — and it lands just as rivals Google and IBM keep advancing a different, superconducting-qubit architecture.
Groq Loses Its Founder to Nvidia, Then Raises $650M Anyway
Groq confirmed a $650 million round led by existing backers Disruptive and Infinitum, roughly six months after Nvidia paid an estimated $20 billion for a non-exclusive license to Groq's chip technology and hired away founder-CEO Jonathan Ross, president Sunny Madra, and several senior engineers.
Rather than competing head-on with Nvidia in hardware, new CEO Doug Wightman is pushing Groq's "neocloud" business — renting out inference capacity across 13 data centers that already serve more than five million developers and process trillions of tokens a week.
The rebuilt leadership leans heavily on outside operators: COO Alan Rice arrives via xAI and Meta by way of the U.S. Navy, while new CTO Sinclair Schuller and CPO Rakesh Malhotra previously built and sold an enterprise software firm to EY.
The math is still unsettled. Nvidia has folded Groq's licensed IP into its own inference hardware, and falling per-token pricing across the industry — DeepSeek recently cut its V4 Pro rate by 75% — means Groq's comeback now hinges on execution in a market it no longer controls alone.
OpenAI's Own Researchers Are Now 56x Deeper Into Codex
OpenAI says median internal Codex output-token usage has grown 56 times in its Research function since November 2025, with Customer Support up 32 times, Engineering up 27 times, and even Legal — a department with no obvious coding workload — up 13 times.
The company frames Codex's spread as a leading indicator for how agentic tools change work generally: tasks are getting longer-running, more cross-functional, and less confined to a single department's toolkit.
Non-developer adoption is the more striking trend line. OpenAI reports individual non-developer usage of Codex up 137 times and organizational non-developer usage up 189 times since August 2025, with non-developers now making up roughly a fifth of Codex's weekly users — adopting three times faster than engineers.
The numbers are self-reported by the company selling the product, a caveat worth holding onto. But directionally, they echo what builders elsewhere are already describing: agentic coding tools quietly becoming general-purpose work tools, not just developer tools.
5 Key Signals
- Inc. — Apple's price increase, explained
- Axios — White House asks OpenAI to limit GPT-5.6
- CNBC — Anthropic accuses Alibaba of Claude extraction
- TechCrunch — General Intuition's $2.3B bet
- Scientific American — Microsoft's quantum claims questioned
- TechCrunch — Groq confirms $650M raise
- OpenAI — How agents are transforming work
- VentureBeat — Claude Tag launches in Slack
- CNBC — OpenAI & Broadcom unveil Jalapeño
- MLQ News — Groq's inference-market pricing pressure