Superpower Daily

Superpower Daily

Your reading space

Sign in
China-Linked Posts Do Not Explain America’s Data-Center BacklashGoogle’s Ironwood Takes an FP8 Cost Lead in New Serving Benchma...Grove City Holds 310-Acre Data Center Plan, Exposing Limits of...Morgan Stanley Raises Oracle Target by $3, Keeps Equal Weight a...Matt Clifford to Leave ARIA After Anthropic Role Draws Conflict...TeraWulf’s $3.2B AI Campus Exposes Gaps in Safety and Power Acc...China-Linked Posts Do Not Explain America’s Data-Center BacklashGoogle’s Ironwood Takes an FP8 Cost Lead in New Serving Benchma...Grove City Holds 310-Acre Data Center Plan, Exposing Limits of...Morgan Stanley Raises Oracle Target by $3, Keeps Equal Weight a...Matt Clifford to Leave ARIA After Anthropic Role Draws Conflict...TeraWulf’s $3.2B AI Campus Exposes Gaps in Safety and Power Acc...
Swipe

Latest AI News

Showing 1-18 of 37Results matching "GPU-as-a-Service"

Today

2 stories
EarningsBusiness

Morgan Stanley Raises Oracle Target by $3, Keeps Equal Weight as GPU Margins Improve

The small target increase puts a fresh margin signal beside an unchanged rating—and an AI infrastructure plan that still requires substantial spending and new financing. Morgan Stanley sees a better margin path in Oracle ’s GPU cloud business, but its latest stock call remains deliberately modest. Morgan Stanley’s assessment suggests its view of the economics is improving as Oracle builds that business.

2 min readRead story
PartnershipInfrastructure

HPE Issues Oracle Warrants in AI Network Deal With Financing and Telemetry

The newly detailed terms add stock warrants, support, financing capabilities and network-monitoring work around Oracle’s prospective global Juniper deployment. HPE has issued Oracle warrants to buy HPE common stock under an AI networking collaboration that also includes financing capabilities, multi-year support services and joint telemetry work. Large GPU clusters rely on the network to move data among machines.

2 min readRead story

Last week

9 stories
Infrastructure

HyperVault Targets $7.4 Billion for a 1-Gigawatt AI Campus in Hyderabad

The proposed campus would give TCS’s infrastructure unit a large physical base for GPU-heavy AI workloads. The next test is converting a secured site and planned capacity into service for AI companies and hyperscalers. Tata Consultancy Services subsidiary HyperVault and its partners are targeting up to 700 billion rupees, or $7.41 billion, for a 1-gigawatt AI data-center campus in Hyderabad, Telangana.

3 min readRead story
PartnershipInfrastructure

Crusoe Signs Reported $13 Billion, Five-Year AI Cloud Deal With Jane Street

The reported agreement gives the trading firm access to Crusoe’s cloud infrastructure for AI training and inference, but it does not connect the $13 billion commitment to a specific site or amount of capacity. Bloomberg reported the private agreement September 3, citing people familiar with the matter who requested anonymity. The reported deal arrives after Crusoe’s June capacity disclosure.

2 min readRead story
FundingStartups

Aranya Raises $11M for Software It Says Can Ready GPU Clusters in 48 Hours

The startup is targeting the operational work that follows GPU procurement: configuring workloads, storage and networking into infrastructure that can run AI services. Its deployment and scale figures are company-reported. Aranya has raised $11 million to build software for a stubborn stage of AI deployment: turning delivered GPU servers into clusters that can run production workloads.

2 min readRead story
Open releaseModels

DeepSeek Releases 305B Vision Weights Under MIT License, but Local Use Needs Serious Hardware

The release gives developers the model files and serving references needed to run or adapt DeepSeek-V4-Flash-Vision-Exp independently. Its 168GB size means that freedom is aimed more at infrastructure-equipped teams than typical individual users. The shift gives developers more control over a model that accepts images, but the 168GB checkpoint sets a high infrastructure threshold for anyone seeking to run it locally.

3 min readRead story
LaunchInfrastructure

Broadcom Validates Five AI Model Families for VMware’s On-Premises Model Service

The update gives enterprises a path to run models from NVIDIA, Google, NEC, Alibaba Cloud and Z.ai inside their own infrastructure, with new sharing, routing and code-isolation services around them. Enterprises using VMware Cloud Foundation now have a validated route to run five additional AI model families on premises and offer them internally as a model-as-a-service.

2 min readRead story

August 2026

7 stories
FundingInfrastructure

Blue Owl Leads $2.4 Billion IREN GPU Financing Tied to Hardware Delivery

IREN has a large debt commitment for its Canadian AI-cloud buildout, but the amount becomes available alongside delivered and commissioned hardware—not as one immediate cash payment. Funds managed by Blue Owl Capital have committed $2.4 billion to IREN’s AI-cloud expansion at its Mackenzie campus. The headline sum is not an upfront payment: IREN will draw the facility as equipment is delivered and commissioned.

2 min readRead story
LaunchInfrastructure

OpenRelay Routes AI Inference Across GPUs, TPUs and Trainium Through One API

The early-access network aims to let developers buy performance rather than a particular accelerator, but its cost-savings claims and ability to deliver consistent service across mixed infrastructure remain unproven. OpenRelay has launched an inference API that sends AI workloads across GPU, TPU and Trainium capacity through a single endpoint. OpenRelay is in early access, though its /v1 API is listed as stable.

3 min readRead story
Enterprise adoptionInfrastructure

Salesforce Uses SageMaker Placement Controls to Meet 2-AZ Agentforce Standard, Keep 8x Savings

The configuration separates model-level resilience from an endpoint’s infrastructure footprint, but balanced placement still depends on GPU capacity in each target Availability Zone. Salesforce has used new placement controls in SageMaker Inference Components to make its production Agentforce models support two Availability Zones, a requirement that an endpoint spanning multiple zones did not previously guarantee.

3 min readRead story
Open releaseTools

Intel Releases 20 Agent Skills for Arc GPUs, With an Open-Port Security Catch

The open repository turns Intel’s GPU deployment steps into reusable instructions for coding agents, but several model-server launchers bind broadly without authentication by default. Intel has released 20 open-source Agent Skills that give coding agents procedures for getting AI workloads running on Arc and Arc Pro GPUs. XPU is Intel’s GPU software path in this collection.

3 min readRead story
EarningsBusiness

IREN Takes $639M Mining Writedown as AI Cloud Revenue Rises Nearly 8x

IREN is retiring older mining hardware and preparing sites for AI cloud growth, but contracted annualized capacity revenue remains well ahead of the cloud revenue it has recognized. IREN shares fell about 6% after its fiscal 2026 results put the cost of its business transition in plain view: a $638.8 million impairment tied largely to older bitcoin-mining equipment it plans to retire.

2 min readRead story
Open releaseInfrastructure

Snowflake Open-Sources Semi-Persistence With Sub-Second Single-GPU vLLM Swaps

The release treats GPU model copies as disposable while retaining weights in CPU memory, aiming to let many specialized models share costly accelerator capacity without long reload delays. Idle models can give back GPU memory only if they can return before the next request waits too long. Avoiding the slow path on every swap The target is a tradeoff in vLLM, the inference engine Semi-Persistence currently wraps.

3 min readRead story
BenchmarkInfrastructure

Heidi Cuts ASR GPUs From 16 to 4 With CUDA MPS, Holding Sub-Second Latency

The production design turns spare capacity in small speech-inference requests into concurrency, but it depends on workload-specific scheduling and protections against MPS failure modes. AWS, Nvidia and Heidi Health have published a speech-recognition serving design that cut Heidi’s required GPU instances from 16 to four while maintaining sub-second transcription latency. Source: aws.amazon.com.

3 min readRead story
18 more next