Tech Current Daily Brief — November 13, 2026

Intro

The November 12 global technology cycle brought a historic leadership change at NVIDIA, proposed AI campuses worth more than $74 billion, new AMD accelerator configurations, cooperation on open inference software and manufacturing-focused AI agents. The distinction between confirmed announcements, proposed spending and vendor performance claims remains essential.

Today’s 5 Top Stories

NVIDIA confirms Jensen Huang's resignation and appoints Colette Kress interim CEO

NVIDIA announced November 12 that co-founder Jensen Huang resigned as president, CEO and director, effective November 11. Its SEC Form 8-K confirms that CFO Colette Kress is interim CEO while retaining her finance role. Harvey C. Jones was named board chair and directors have begun a search for a permanent successor. NVIDIA said it expects operations and its product roadmap to continue without interruption; this is management's expectation, not a guarantee.

Why it matters: The first CEO transition in NVIDIA's 33-year history is consequential for the global AI accelerator ecosystem. Watch the succession process and the November 17 earnings call rather than assuming immediate disruption to product shipments.

SpaceX outlines more than $74 billion in proposed AI supercomputing campuses in Texas and Arizona

SpaceX outlined plans November 12 for more than $74 billion of investment in two US AI supercomputing campuses: over $24 billion in Texas and about $50 billion in Cochise County, Arizona. Reuters reported that the Arizona project still requires local approvals and faces community concerns. Planned investment does not mean funds have been fully spent or the facilities have been built.

Why it matters: AI capacity expansion depends on power generation, grid access, permitting and financing as much as on chip supply. Proposed campuses of this scale underline both demand and execution risk.

AMD expands Instinct MI400 family with 288 GB MI450X and MI455X inference data

AMD announced the Instinct MI450X accelerator with 288 GB HBM4 on November 12, expanding its Helios platform choices. AMD also published its own MI455X DeepSeek-R1 inference results, reporting 21,333 tokens per second per GPU under a specified offline-style configuration. Comparisons with NVIDIA benchmarks use separately reported setups and should not be treated as independent like-for-like testing.

Why it matters: Inference economics depend on usable memory, throughput and deployment fit. The product option increases customer choice, but third-party reproducibility remains critical.

SGLang Consortium unites major AI chip and cloud firms around open inference

LMSYS Org announced the SGLang Consortium on November 12 with NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk. The group aims to coordinate open-source inference runtime work across hardware, multimodal workloads, distributed serving and model support. Project maintainers retain code-acceptance authority. The launch is a governance and collaboration announcement, not proof that planned optimizations have already shipped.

Why it matters: Open inference runtimes may help reduce model-serving costs and limit dependence on one accelerator vendor. Transparent governance and actual code contributions will determine the consortium's impact.

Anthropic launches Claude for Manufacturing with 17 skills and four connectors

Anthropic introduced Claude for Manufacturing on November 12, describing a plugin with 17 task-specific skills and connectors for Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. Proposed uses include job quotations, quality investigations and supply-chain coordination. Anthropic says human employees review and approve outputs before external use; the plugin does not directly operate factory machinery. Productivity claims have not been independently established.

Why it matters: Industrial AI agents need reliable enterprise data access, audit trails and explicit approval boundaries. Domain-specific skills may make value easier to measure, but operational reliability and access controls remain open questions.

Data & Market Pulse

SpaceX described more than $74 billion of proposed AI infrastructure investment, not completed construction. AMD announced a 288 GB HBM4 accelerator option and reported 21,333 tokens per second per MI455X GPU in a vendor-run configuration. Anthropic listed 17 manufacturing skills and four enterprise connectors. Planned capital, hardware specifications, performance measurements and product feature counts are distinct metrics.

Trend Watch

The AI supply chain faces a governance test as NVIDIA begins a rare leadership transition, while rivals compete on memory configurations and measured inference economics.

Inference software is emerging as a shared battleground: major competing chipmakers are cooperating through the newly announced SGLang Consortium.

Persistent agents and System 1-style decision layers require auditable approvals and access controls. No independently verified November 12 Jev milestone qualifies for today's top five.

What to Watch Next

Watch NVIDIA's CEO search and November 17 earnings, permitting and financing for SpaceX's proposed campuses, independent AMD benchmark reproduction, SGLang engineering contributions, and field evidence for Claude for Manufacturing.