A KV cache, or key-value cache, speeds up autoregressive transformer inference by storing attention information calculated for tokens that have already been processed. Without the cache, the model would recompute those keys and values each time it generates another token. Reusing them reduces repeated work and improves response latency, especially for long prompts and outputs. The tradeoff is memory: cache size grows with sequence length, batch size, layer count, and model dimensions. Production systems manage this cost with paging, quantization, eviction, prompt caching, and request scheduling. KV caching changes inference efficiency, not the model’s learned knowledge, and cached data must be protected when prompts contain sensitive information.
Google AI Plus adds Gemini 3.1 Pro, Omni Flash video and family sharing while undercutting pricier ChatGPT and Anthropic subscription tiers in the US market.
More than 600 OpenAI employees sold a combined $6.6 billion in shares in October, marking one of the largest pre-IPO payouts in Silicon Valley history.
OpenAI is negotiating a 10-gigawatt Ohio data center lease that could cost at least $500 billion, with Nvidia guaranteeing the financing in one of the largest AI infrastructure bets yet proposed.
SpaceX has released its most detailed look yet at a planned orbital AI data center satellite, unveiling a 70-meter-wingspan structure during the same week the company is pitching investors on a $1.75 trillion public listing.
Anthropic has released Claude Fable 5, its most capable model available to the general public, alongside Claude Mythos 5 – an identical underlying model with key safety restrictions removed, available only to approved cybersecurity and research partners.
Apple fell 3% on Tuesday in its worst single-day decline since February, as investors sold off following Monday’s WWDC announcements amid concern that Siri AI lacks a confirmed launch date and faces regulatory delays in two major markets.
JPMorgan Chase plans to roll out AI agents capable of operating autonomously for one to two hours later this year, as the bank’s chief analytics officer says the technology has moved from single-task tools to multi-step workflow managers.
OpenAI has submitted a confidential S-1 filing with the SEC, joining Anthropic and SpaceX in a rare simultaneous push toward public markets by three of the most valuable private companies in the AI sector.
Apple used its WWDC 2026 keynote to deliver a rebuilt Siri powered by Google’s Gemini, launch iOS 27 and macOS Golden Gate, and refine the Liquid Glass design system – all in Tim Cook’s final keynote as CEO before stepping down in September.
Apple’s WWDC 2026 keynote begins June 8 at 1PM ET, with a rebuilt Siri, Gemini integration, iOS 27 features, and new AI tools across apps and operating systems expected.