TLDR IT 2026-08-14

IBM is launching a dedicated OpenAI consulting practice and plans to certify tens of thousands of consultants. The companies will jointly build AI offerings for industries including finance, government, telecom, and retail.

Cisco booked $4 billion in hyperscaler AI infrastructure orders last quarter, bringing its FY2026 total to $9.3 billion. It expects AI infrastructure revenue to climb to roughly $7.5 billion in FY2027.

Enterprise AI agents fail to gain broad adoption because industry-standard interfaces force users to make too many engineering decisions. The proposed alternative is a simpler default: one persistent, general-purpose agent embedded in workplace chat, using generic API access and live data rather than complex agent configurations, knowledge bases, or fine-tuning.

Capital One built its multi-agent platform around open-weight models as it pushes toward enterprise-wide AI. The approach gives it more control over deployment, customization, and infrastructure.

Prices for 30TB enterprise SSDs have risen roughly 6.5x year over year, significantly changing the economics of AI infrastructure. Storage is becoming a major cost factor for large GPU deployments.

Oracle added NVIDIA Nemotron 3.5 Lightning, multi-node H100 serving, expanded on-demand models, and new IAM controls to OCI Enterprise AI. The updates give enterprises more flexibility over models and infrastructure.

A10 Networks has announced the general availability of the A10 AI Gateway, a centralized control plane for routing, governing, monitoring, and managing the costs of enterprise AI agents, applications, and large language models. The company says it can route requests by complexity, enforce identity-based access and usage limits, and run in customer-controlled environments including on-premises, private-cloud, and air-gapped deployments.

Writer says Palmyra X6 can cut agent costs by 52% as enterprises face growing token bills. The model is optimized for long-running agentic workloads.

AWS detailed how AgentCore Observability can monitor agents running on-premises, Azure, GCP, or AWS. OpenTelemetry telemetry feeds into one dashboard for tracing actions, token usage, and reliability.

MCP's latest architecture makes models carry session handles rather than relying on persistent server sessions. That improves scalability but leaves applications responsible for reliable correlation, authorization, and session management.

Nebius will deploy high-density NVIDIA infrastructure at Vantage's Wales campus as the first commercial capacity commitment in the South Wales AI Growth Zone.

Lovable is partnering with Cerebras to accelerate inference on its AI-powered application development platform.