Apple's new Siri AI runs on a custom 1.2-trillion-parameter Google Gemini model licensed at roughly $1 billion per year, with an Extensions system that lets users route Apple Intelligence features through Claude or ChatGPT instead. The two signals in one keynote — Apple outsourcing its AI core to a rival and Tim Cook handing the CEO role to hardware chief John Ternus on September 1 — mark the end of Apple's self-contained platform model.

Google has placed a confirmed order for more than 3 million tensor processing units (TPUs) with Intel for 2028 production — concrete enough to lift Intel shares 9%, and the first public evidence that TSMC has a funded, large-scale foundry rival in AI chips. Nvidia is separately evaluating Intel's chip packaging for its next-gen Feynman GPU architecture, and with Tesla already committed to Intel's 14A process for Terafab, the AI chip supply chain is visibly splitting across two foundries.

NTT has assembled a ¥70 billion-plus fund with SK Group, Chunghwa Telecom, Sony, Fujitsu, and MUFG to back startups in North America, Asia, and Europe building on IOWN — its all-photonic network that eliminates photon-to-electrical conversion to cut data center power consumption. The bet is structural: as AI workloads push electricity demand sharply higher, a cross-Pacific consortium is funding the photonics path rather than waiting for incremental chip efficiency to close the gap.

During Jensen Huang's Seoul visit, NVIDIA formalized deals spanning Korea's entire AI layer: SK hynix in a multiyear memory co-development partnership across four NVIDIA platforms, SK Telecom and NAVER both targeting gigawatt-scale AI Cloud buildouts using NVIDIA DSX, and LG and Doosan committing to AI factory infrastructure for robotics and industrial automation. The breadth — from memory substrate to data center to robot endpoint — means Korea is now simultaneously one of NVIDIA's key memory suppliers and one of its largest infrastructure customers, with the full stack running on NVIDIA platforms.

Foundation Models is now a unified native Swift API that routes requests to Apple's on-device models — built in collaboration with Google — or to Claude, Gemini, or any provider implementing the new language model protocol, alongside Core AI for running full-scale LLMs locally on Apple silicon. An iOS developer can consolidate model-provider integrations to a single call today, and Xcode 27 adds agentic coding sessions with models from Anthropic, Google, and OpenAI directly in the IDE.

Hackers injected password-stealing malware into at least 70 Microsoft open source GitHub repos — Azure utilities, Claude Code integrations, Gemini CLI, and VS Code extensions — with Cloudsmith confirming this as a re-compromise of the Durable Task project, meaning the May cleanup left the attacker's foothold intact. Developers who pulled any of the affected packages before the repos were disabled may have exposed credentials to the attackers.

CVE-2026-23111 is a use-after-free in nf_tables — the Linux kernel's packet-filtering subsystem — triggered by a single misplaced exclamation point in the source; an unprivileged local process can exploit the corrupted verdict-deletion path to escalate to root. Patches are in development but not yet universally distributed, so every shared-compute or multi-tenant Linux host — cloud VMs, containerized workloads — stays exposed until the kernel is updated across the fleet.

Apple's WWDC 2026 App Bundles expansion lets developers at separate companies package their subscriptions together — including 'Suites' that have no standalone purchase option — mirroring the cross-company bundle deals common in streaming. Developers with overlapping but non-competing user bases can now build joint pricing tiers to drive cross-app retention without competing on individual subscription price.

Lovable grew from $400M to $500M in annualized revenue between February and early June 2026, with 1 million new projects per week — and its own user survey shows the builders are primarily non-technical: founders, designers, and salespeople replacing SaaS tools like CRMs, inventory systems, and HR platforms with custom builds. The build-vs-buy shift is measurably underway on the creation side; the open question is whether non-technical teams will also absorb the maintenance burden that drives most companies to buy software rather than build it.