Demand (PC, memory) and capital (bonds, equity) both dragged into the present — the whole chain's fate rests on the durability of the inference margin
The AI capital cycle is a structure that has pulled both demand (PC, memory) and capital (bonds, equity) forward from the future into the present. The whole chain's ability to repay hangs on one variable — the durability of the inference margin.
Two events signaled from both ends of the chain (capital and accounting) at once: NVIDIA's first $25 billion bond issuance since 2021, and Amazon cutting AI-asset useful life from 6→5 years against its own earnings.
Structural winners — data infrastructure / RAG pipelines (the memory mechanism separating the 5%), and own-silicon vertical integration (inference cost edge). Pressure variables — hyperscalers holding a six-year useful life (earnings cushion), and firms with off-balance-sheet exposure and widening CDS spreads (Oracle, Meta).
The point where inference-ASIC TCO breaks the older-GPU cascade (life-premise collapse), further life shortening in hyperscaler 10-Ks, Oracle/Meta CDS spreads and days payable outstanding, and the exhaustion point of pulled-forward PC/memory demand.
What forced the corporate PC refresh was not AI but a deadline. Microsoft ended Windows 10 mainstream support on October 14, 2025, and the prior major refresh — the pandemic work-from-home boom (2020–2021) — reached the end of its 4–5 year at the same moment. As a result, 2025 global PC shipments hit roughly 270 million units (+9.1%), with Q4 at 71.5 million (+9.3%) — the strongest rebound since the pandemic (Gartner).
The problem is that this is not organic . Both Gartner and IDC attribute Q1 2026 shipments (62.8 million, +4% per Gartner) to pre-buying ahead of component price increases, Windows 10 migration and new product launches — not organic demand — and warn of possible shipment declines for the rest of 2026. About 40% of the installed base was still on Windows 10 at end-2024, and migration proceeded slowly.
Refresh capacity is a function of cash flow and IT-governance maturity. Large enterprises refresh first; SMBs lag through 2027 on budget cycles. Those who cannot replace pay for ESU (Extended Security Updates) at $61 per device in year one, $122 in year two, $244 in year three (Microsoft) — a penalty that doubles every year, hitting the smallest hardest. The on-device AI (Copilot+ tier) spec floor is 40 TOPS NPU, 16GB, 256GB, but cloud AI runs fine in a browser, so "spec enforcement" is more pretext than necessity.
The key linkage is memory. As hyperscaler DRAM/HBM allocation drained PC-grade DDR/NAND supply and lifted prices, OEMs pulled orders forward ahead of 2026 price hikes — one driver of the Q4 shipment surge (Gartner). The PC cycle and the memory cycle are two sides of the same cause.
Strong PC refresh is pulled-forward demand, and its driver is a by-product of memory allocation. AI is the pretext layered on top; the real forcing function is Windows 10's end and component prices. Pulled-forward demand reverts to declining shipments once exhausted — that reversion point is the first thing to watch.
Sign in with Google to unlock the full body of every free report instantly.
Sign in with GoogleThis site runs on ads — the tier system rewards community contributions.
This report is provided for informational purposes only and does not constitute a recommendation to buy or sell any financial instrument. Investment decisions should be made based on your own judgment and responsibility. The analysis and opinions contained herein are based on information available at the time of writing and are subject to change.