Superpower Daily
University of Hawaiʻi Gets $2M to Test AI Early Warnings for Fo...Fervo Targets October Start for Potential First U.S. Commercial...InferenceX Ranks B300 First for Qwen3.5 at 15,815 Tokens per Se...Fix Victoria Uses AI Ads Ahead of Victoria’s Election While Its...Artificial Analysis Clarifies Astra’s 67-Point Coding Score Is...More Than $45 Million in Data-Center Ads Pushes AI Infrastructu...University of Hawaiʻi Gets $2M to Test AI Early Warnings for Fo...Fervo Targets October Start for Potential First U.S. Commercial...InferenceX Ranks B300 First for Qwen3.5 at 15,815 Tokens per Se...Fix Victoria Uses AI Ads Ahead of Victoria’s Election While Its...Artificial Analysis Clarifies Astra’s 67-Point Coding Score Is...More Than $45 Million in Data-Center Ads Pushes AI Infrastructu...
Swipe

Full archive

Latest AI News

Search reporting on AI markets, models, products, companies, policy, and culture.

Category

More filters
Search: B300
Showing 1-14 of 14Results matching "B300"

Today

2 stories
BenchmarkInfrastructure

InferenceX Ranks GB200 First for MiniMax M3, While MI355X Has Lower Token Cost

The live AgentX result gives operators a comparable speed-and-price snapshot at one responsiveness target, but the ranking can shift with serving software and does not settle the wider configuration trade-offs. Infrastructure teams serving MiniMax M3 now face a clean split in InferenceX ’s latest ranking: GB200 NVL72 delivers more throughput, while MI355X carries the lower reported price per million tokens.

3 min readRead story

Earlier this week

3 stories
FundingStartups

Lambda Adds $1B in Short-Dated Debt for Nvidia Chips Bound for Microsoft

The borrowing adds another customer-specific hardware deployment to Lambda’s balance sheet, while its reported equity plans remain unfinished. Lambda has reportedly raised $1 billion in private, short-dated debt to buy Nvidia AI chips it intends to lease to Microsoft, adding another large borrowing to its AI-cloud expansion. The latest $1 billion debt is intended to buy Nvidia AI chips for lease to Microsoft.

2 min readRead story

Last week

7 stories
Government actionInfrastructure

Vineland AI Site Linked to Microsoft Faces Stop-Work Orders and Generator Permit Questions

The dispute puts a basic AI-infrastructure constraint in view: contracted computing capacity still depends on permits, local acceptance and systems that can safely support a massive site. DataOne says it is committed to meeting environmental and permitting requirements at its Vineland, New Jersey, project. The project is not a conventional cloud facility with a modest local footprint.

2 min readRead story
Open releaseTools

NVIDIA’s TensorRT Model Connect Skips ONNX, but x86_64 Users Must Build From Source

The open-source preview turns model conversion into a versioned deployment artifact, separating engine construction from native inference while imposing an ARM-first installation path. NVIDIA has released TensorRT Model Connect, or TRTMC, in public preview: an Apache-2.0 tool that takes supported Hugging Face or local model checkpoints to TensorRT inference in two commands, without an intermediate ONNX export.

3 min readRead story
Enterprise adoptionInfrastructure

Meta Tests Data-Center Repair Robots, but Cable Bots Still Need Human Supervision

Meta already uses robots for bounded logistics work. The harder test is whether supervised machines can make delicate maintenance tasks fast and reliable enough to change how data centers are staffed. Meta is pushing robots beyond rack transport and barcode scanning toward work traditionally done by technicians: swapping cables, resetting servers, and reseating parts.

3 min readRead story
LaunchProducts

Corti Launches a Governed AI Coding Layer on Denmark’s Gefion Supercomputer

The new service gives enterprises one route into multiple AI models while retaining choices over where workloads run and how coding use is controlled. Corti and the Danish Center for AI Innovation have launched Corti Models, a model-agnostic infrastructure and governance layer for European enterprises. Customers can use sovereign-cloud or on-premises deployment, rather than being confined to one delivery route.

3 min readRead story
BenchmarkInfrastructure

OpenAI’s Jalapeño Claims Up to 1.9x Efficiency Gain, but Leaves Nvidia Training Intact

The Broadcom co-developed inference processor is headed for deployment by year-end. Its published gains are meaningful, but they exclude model training and Nvidia’s Vera Rubin platform. OpenAI says Jalapeño, its custom chip co-developed with Broadcom, delivered up to 1.9 times higher throughput per kilowatt and 1.7 to 3.6 times lower end-to-end latency than Nvidia GB200 and GB300 systems in published tests.

3 min readRead story
Government actionPolicy

Taiwan Indicts Nvidia Manager Over Alleged Diversion of 74 B300 Servers to China

The case centers on an alleged mismatch between where 130 AI servers were documented as installed and where authorities say 74 ultimately went, testing export controls from paperwork to border inspection. Taiwan has indicted an Nvidia manager surnamed Chang and eight others over an alleged operation that sent 74 servers carrying high-end B300 chips to Chinese customers.

3 min readRead story
BenchmarkInfrastructure

NVIDIA Says Vera Rubin Delivers 30x More Agentic Throughput per Megawatt, Pending Review

The preliminary comparison shifts attention from fixed prompt benchmarks toward the power required to keep long, tool-using AI sessions responsive. NVIDIA says Vera Rubin NVL72 delivered up to 30 times the throughput per megawatt of GB300 NVL72 for an agentic AI workload at 160 output tokens per second per user. NVIDIA reports the DeepSeek V4 Pro 1.6T comparison on the AgentX workload.

3 min readRead story

August 2026

2 stories
PartnershipInfrastructure

Together AI Secures $240 Million IBM Cloud Buildout for Open-Source Inference

The agreement gives Together AI a planned dedicated IBM Cloud system for serving enterprise open-source models. The commercial test will be whether the 2027 deployment delivers better service performance and token economics at its reported scale. Together AI is lining up a dedicated, large-scale IBM Cloud cluster to serve open-source models to enterprise customers.

3 min readRead story