CNBC: Broadcom stock falls on revenue miss and unchanged AI chip forecast

Gemini is Google's family of multimodal AI models designed to compete with OpenAI's GPT series. Formerly known as Bard, it rebranded to Gemini in 2024 and directly integrates with Google services. The Ultra 1.0 model achieved state-of-the-art performance on multiple benchmarks. Gemini is available through the Google AI app, web interface, and integrates with Gmail, Docs, and other Google Workspace applications.

Portkey documents an AI gateway at docs.portkey.ai that unifies access to more than 250 models through a Portkey SDK or OpenAI-compatible base URL (`PORTKEY_GATEWAY_URL`) with provider routing headers. Official quickstarts show three-line Python or TypeScript integrations that start monitoring LLM requests for resilience, security, and performance. Portkey states the open-source gateway is free to self-host while the managed service includes a free tier of 10k requests per month, edge-hosted workers adding roughly 20–40ms latency versus direct API calls, ISO 27001 and SOC 2 certifications, and optional configurations that skip storing request/response bodies.

NVIDIA NIM documents performance-optimized inference microservices at docs.api.nvidia.com/nim and docs.nvidia.com/nim that expose industry-standard APIs (OpenAI-compatible `/v1/chat/completions`, `/v1/completions`, `/v1/responses`, Anthropic-compatible `/v1/messages`) from containerized models backed by TensorRT-LLM, vLLM, or SGLang per deployment. Teams can self-host GPU-accelerated models on cloud, data center, or RTX workstations, or prototype via NVIDIA-hosted NIM API endpoints through the Developer Program. Management endpoints such as `/v1/health/ready` and `/v1/metrics` support readiness probes and Prometheus metrics on self-hosted containers per the LLM API reference.

Weights & Biases documents a hosted Model Context Protocol server at `https://mcp.withwandb.com/mcp` (recommended) and an open-source `wandb/wandb-mcp-server` package for local stdio or HTTP per docs.wandb.ai/platform/mcp-server. Authenticated clients pass a W&B API key in the `Authorization: Bearer` header. Documented tools include `query_wandb_tool` and `get_run_history_tool` for experiment metrics, `query_weave_traces_tool` and `count_weave_traces_tool` for LLM traces, `create_wandb_report_tool` for markdown reports, `search_wandb_docs_tool` for docs.wandb.ai, and `query_wandb_entity_projects` for project listings. Dedicated/on-prem deployments can set `WANDB_BASE_URL` with the local server per README.

Langfuse documents two MCP surfaces: a hosted Prompt Management server at `https://cloud.langfuse.com/api/public/mcp` (streamable HTTP, built into the platform per langfuse.com/docs/prompt-management/features/mcp-server) with tools such as `createTextPrompt`, `createChatPrompt`, `getPrompt`, and `updatePromptLabels`; and a public Docs MCP at `https://langfuse.com/api/mcp` for semantic search over Langfuse documentation (`searchLangfuseDocs`) without authentication. The legacy `langfuse/mcp-server-langfuse` Node server remains for local command-based setups with LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY. Langfuse positions MCP for agentic onboarding—IDE agents can add tracing to OpenAI, Anthropic, or LangChain codebases from the Get Started agentic tab.

Structures verification of custom-AI chip vendor earnings headlines into a finance and supply-chain checklist. The workflow separates consolidated revenue and EPS beats from AI semiconductor sub-segment growth, full-year AI revenue guidance (raised vs reiterated), and infrastructure software shortfalls cited in the same report. It references CNBC reporting on June 3, 2026 that Broadcom's fiscal Q2 revenue was $22.19 billion versus $22.27 billion estimated (48% YoY), adjusted EPS $2.44 vs $2.40, AI semiconductor revenue $10.8 billion (more than doubled YoY), Q3 revenue guidance about $29.4 billion vs $28.53 billion expected, infrastructure software revenue $7.18 billion vs $7.32 billion expected, CEO Hock Tan reiterating AI semiconductor revenue in excess of $100 billion in fiscal 2027 without raising the 2026 forecast, naming six core custom-chip customers including Anthropic, Google, Meta, and OpenAI, and saying Broadcom would offer chips only rather than complete integrated AI systems—without treating media figures as procurement commitments.

Turns public semiconductor announcements into a verification checklist when vendors claim novel scaling laws, stacked logic architectures, or nanometer-class equivalence without independent benchmarks. Teams separate marketing nomenclature from manufacturing readiness by demanding yield, thermal, packaging, and third-party validation evidence—patterns highlighted when CNBC reported Huawei's LogicFolding and τ Scaling Law claims alongside analyst skepticism about true 1.4nm-class process breakthroughs without EUV access. The skill also maps export-control context (ASML EUV restrictions) and competitive implications for GPU vendors operating in constrained geographies.