Locsiclocsic.com
← Observe Developing · Medium

AI Infrastructure Rebuild

AI infrastructure is moving from chip performance to system-level competition.

2026-06-194 signals
Topic direction
Why watch

The real competition is shifting from single-chip speed to rack, network, memory, power, packaging, protocol, software stack, and deployment economics.

Current read

The next infrastructure boundary will be decided by the coupled system: accelerator, NIC, switch, protocol, memory locality, runtime, power, storage, and operations.

Last updated · 2026-06-19 Signals · 4 Entities · NVIDIA / AMD / Broadcom / Cisco / Arista / Huawei / HPE / SpaceX
AIInfrastructureNetworkingInferenceDatacenterMemoryPower
Key events
Timeline axis Latest → Earlier

Scroll right for earlier events

Reverse event list

Watching · Medium

AI factory becomes a control-plane problem

HPE positions compute, storage, network, Cray, and software as one factory stack.

4 sources 4 sources Track whether infrastructure vendors can integrate the stack without hiding economics.
Company

HPE turns AI factory into stack integration

HPE Discover ties AI factory, network, agent workload, and strategy into one systems bet.

Company Company HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.
Watching · Medium

MaaS cost stack becomes visible

Inference economics starts to expose KV cache, batching, scheduling, and token distribution as one stack.

2 sources 2 sources Track which layer captures margin.
Market

MaaS exposes inference economics

MaaS articles make the cost stack visible: KV cache, batching, routing, utilization, and token distribution.

Market Market HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.
Watching · Medium

Memory pressure rewrites inference infra

KV Cache and SSD-as-memory signals point to a new storage hierarchy for inference.

2 sources 2 sources Connect memory economics to accelerator and cloud deployment decisions.
Technology

KV Cache rewrites the memory hierarchy

Inference forces infrastructure to treat memory, SSD, and scheduling as one hierarchy.

Technology Technology HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.
operation

Colossus shows compute is not battle power

SpaceX Colossus rental signals that raw GPU count does not equal usable AI capacity.

operation operation HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.
Technology

Optics debate moves into AI fabric timing

CPO/NPO debate shows optical interconnect is now an investment and deployment timing question.

Technology Technology HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.
Watching · Medium

Network protocols enter architecture

MRC, ZCube, RailFly, Lingqu, and optical shuffle show networking is no longer plumbing.

4 sources 4 sources Keep clustering topology, transport, and chip design into one infrastructure line.
Technology

Topology and protocol become architecture

MRC, ZCube, RailFly, and Lingqu connect transport, topology, NICs, switches, and AI workload shape.

Technology Technology HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.
Technology

Power, packaging, and memory become the hidden bottleneck

800V power, 3D stacking, Rubin BOM, and CPU comeback point to non-GPU constraints.

Technology Technology HPE, MaaS, KV Cache, optical interconnect, and AI network protocols are converging into one infrastructure thesis.

Open questions

  • Q01 Trend Will infrastructure advantage shift from accelerators to coupled rack-level systems?
  • Q02 Possible path Which layer captures margin first: model, serving stack, memory, network, power, or cloud operations?
  • Q03 Constraint Do NICs, switches, and protocols become part of the accelerator architecture once memory locality dominates?
  • Q04 Validation Will enterprise AI factories be integrated products, or remain custom assembly projects?

Linked Thinking

June 202611 posts
2026-06-18HPE The $14 Billion Bet: HPE Discover 2026 Strategic Overview A decade of divestiture, then all-in on networking and AI factories. HPE's… Read 2026-06-18HPE Network as the AI Control Plane: HPE's Networking Gamble QFX six-tier coverage from training to inference, GreenLake Intelligence… Read 2026-06-18HPE Decoding the HPE AI Factory: Compute, Storage, Software, and the Cray Integration Experiment DL 394 Gen 12, Alletra MPX 10000 MCP-native storage, GreenLake full-stack… Read 2026-06-16MaaS MaaS Inference Tech Stack: How Six Levers Cut Cost by 96% Technical dissection of DeepSeek's 96% per-token cost reduction. Six… Read 2026-06-16MaaS The Token Distribution Era: MaaS Service Models and Business Anatomy China's MaaS break-even line is 5-7 yuan/million tokens, with mainstream… Read 2026-06-15AI The Tyranny of Memory: How KV Cache Is Reshaping Every Layer of AI Inference In the 1M-context era, KV Cache is the defining bottleneck of inference… Read 2026-06-14AI When SSD Becomes Memory: How AI Inference Is Rewriting the Storage Hierarchy Three facts, sitting side by side in the first half of 2026, create a… Read 2026-06-13AI基础设施 Compute Power Is Not Fighting Power: What the SpaceX Colossus Lease Tells Us About AI Infrastructure Reality 220,000 GPUs built and leased out within a year. The SpaceX Colossus lease… Read 2026-06-12CPO One Report Wiped Out Optical Stocks: Is CPO Actually Dead? On June 9, 2026, SemiAnalysis sent a research note to institutional… Read 2026-06-02AI From CLOS to ZCube: Network Topology Evolution for AI Computing Clusters From Charles Clos's non-blocking telephone switching network in 1953 to… Read 2026-06-02AI MRC: When the NIC Becomes the Brain of the Network OpenAI, together with NVIDIA, AMD, Broadcom, Arista, and Cisco, overturned… Read
May 20264 posts
2026-05-31AI Data Center The $27 Billion Hidden Thread: AI Data Center Power Distribution Revolution and the New Analog Chip Landscape As rack power consumption surges from 15kW to 1.5MW, power distribution… Read 2026-05-30AI Deep Analysis of the Lingqu Protocol The Interconnect Bet Behind China's AI Compute Breakout Read 2026-05-28AI The CPU Is Back: How Agentic AI Is Rewriting the Server Processor Landscape Three things happened almost simultaneously in May 2026. AMD's Venice… Read 2026-05-26AI Starting from ZCube: Network Topology Design for PD Disaggregated Inference As inference replaces training as the primary battlefield for AI… Read