Large Language Models

Today's Stories

1. 1

   Broadcom to lend Anthropic up to $42 billion for chip leases

   Reuters reported, citing a filing, that Broadcom will lend Anthropic up to $42 billion to lease its chips. This ties a major chip supplier's lending directly to a customer's chip use, so Broadcom's exposure appears to rest on Anthropic's continued leasing.

   The arrangement hinges on how much of the up-to-$42 billion Anthropic actually draws, since it is a lending ceiling rather than a confirmed total.

2. 2

   NVIDIA's DIN Deploy ships C++ local AI samples

   NVIDIA published DIN Deploy, open-source C++ samples that combine ONNX Runtime with the TensorRT RTX execution provider, with support for Windows and Linux. Developers get a shared code path with vendor-specific CUDA code only in optional accelerated paths, so the same samples can run across systems that support the required tensor APIs.

   The samples are only as portable as the execution providers that support ONNX Runtime's tensor APIs. Watch the DGX Spark results, where whisper-large-v3-turbo ran 58.5x real time on GPU versus 3.8x on CPU.

3. 3

   Alphabet sends AI chips to orbit in first Project Suncatcher test

   Alphabet launched its own tensor processing units (TPUs — Google's custom AI chips) on a SpaceX Falcon 9 on Thursday, the first in-orbit test for Project Suncatcher, first revealed in November 2025. Chip performance in orbit is still unknown.

   Industry experts say orbital data centers would need cooling systems and radiation-hardened chips, and remain limited by expensive rocket launches. SpaceX COO Gwynne Shotwell said 'supercompute in space' will be deployed in 2027.

4. 4

   Charles Schwab rolls out AI assistant Charley in October

   Charles Schwab launched Charley, a generative AI-powered assistant, with a phased rollout starting in October to eligible US retail clients on schwab.com and Schwab Mobile. Retail brokerage clients get a self-service tool for market information, account and service questions, task execution, and connecting to a company representative.

   Access is limited at first to eligible US retail clients, then expands to additional client groups over time, so the test is how smoothly that wider rollout goes.

5. 5

   Meta's Muse hits 560,000 daily users, spurs Booking and Expedia bets

   Meta's Muse shopping agent reached 560,000 daily active users in 11 days, Truist cut price targets on Booking Holdings and Expedia, and Expedia partnered with Muse while Booking stayed out. Partnering with Muse could teach Meta how travelers shop and let it bypass Expedia later, while Booking's direct channel and widely processed payments may keep it essential to AI agents.

   Booking's approach hinges on whether large language model traffic stays below 1% of room nights and its consumer direct mix holds in the mid-60% range; watch for any climb above those levels.

6. 6

   Idira gives AI agents SPIFFE identities, MCP tool limits

   Palo Alto Networks' Idira shipped August features to govern AI agents — MCP access policies limiting which tools an agent can run, SPIFFE identities and short-lived tokens for Google Gemini Enterprise agents, and near-real-time account sync from Privilege Cloud. Agents become registered, auditable identities rather than shadow scripts borrowing someone else's token, and machine identity shifts from stored static secrets toward workloads proving what they are — the platform's own "identity where you can, secrets where you must" idea.

   Sync latency is described as "close to real time" and rollouts finished region by region, so whether the tightest least-privilege controls hold in day-to-day operations is likely the test. Note the APIs now hide authenticator configuration without permission.