[Daily Briefing] Ternus runs Apple, RAM up 110 percent, AI browsers hijacked

Tim Cook is gone. John Ternus runs Apple from today.

John Ternus became Apple CEO effective today, 1 September 2026, with Tim Cook moving to executive chairman after 15 years in the chair. The handover was announced back in April and approved unanimously by the board, so this is the scheduled landing rather than a shock.

The timing is the part nobody is saying out loud. Apple holds its Surprise and Shine event on 9 September, which gives Ternus eight days before his first keynote as chief executive. He joined Apple in 2001 on the Product Design team and has run Hardware Engineering since January 2021, so the company has handed the top job to a hardware person at the exact moment the pressure is all on AI software.

Underneath him, much of the leadership layer that has run Apple for 15 years is aging out alongside Cook. Ternus has to rebuild that bench while shipping a fall lineup that reportedly includes an Apple Watch Series 12 and a foldable iPhone with Pencil support.

Why it matters: Apple Intelligence has been the weak spot for two years running. Putting a hardware engineer in charge is a signal about where Apple thinks the fix lives, in silicon rather than in models.

Read more

Quick hits

• Consumer RAM has more than doubled this year. DRAM contract prices rose 90 to 95 percent in Q1 and another 58 percent in Q2, and SSDs are up 147 percent in twelve months. The buying guide, priced
• Every AI browser tested was vulnerable to prompt injection. The Opera AI browser, Perplexity Comet and ChatGPT Atlas all executed instructions hidden as white text on white backgrounds. What to stop letting agents do
• Intel put up to 480GB on a single air cooled card. Crescent Island goes LPDDR5X and 350W while AMD and Nvidia go liquid cooled HBM4. Three vendors, three answers to the same memory bound problem. The specs
• Semiconductors are heading for 1.6 trillion dollars in 2026, and memory is 54 percent of it. Memory used to be the low margin end of this industry. That inverted inside a single cycle. The numbers
• The Chinese small model race is now a weekly event. Z.ai posted GLM-5.3 Flash and Alibaba Cloud announced Qwen3-8 Flash, both pitched on price to performance rather than benchmarks.

Tool of the day: a boring DDR5 kit, bought now rather than later

Not exciting, and that is rather the point. If you are on 16GB and expect to keep the machine into 2027, memory is the one component where waiting has actively cost people money all year, and the shortage is forecast to run through next year. A 32GB DDR5 5600 kit is the sensible floor for most desktops. Current DDR5 kits

Honest caveat, because it matters more than the link: if you already run 32GB and never touch local AI models, you do not need this. Do nothing and keep your money.

Reply and tell me one thing, because I am genuinely curious how this is landing. Are you buying memory now, or waiting it out and hoping 2027 is kinder?

Nath, Future Technology

Forwarded? Get your own free briefing at futuretechnologyhq.com/newsletter

Some links may be affiliate links.