Second Order by TKN
The free Daily Wire is a five-minute brief: what changed, why money moves, and what to watch next.
Everyone watches Nvidia. We map the next links in the stack—power, cooling, packaging, and neoclouds.
Get the WireLead Explained
The Wire names the move. Start here when you need the mechanism behind it.
The system
Desks publish the notes. These two indexes show where money sticks in the stack, and what changed by country.
Layers
Start where the rent is rising. Full stack later.
Open The StackMarkets
Layers explain the stack. Markets show what changed locally — and what’s coming next.
Open MarketsWire
Know what changed, why it matters, and what to watch next. That is the Daily Wire benefit, on the site and in the inbox.
Workers at Micron's Taoyuan DRAM and HBM plant voted down a new profit-sharing offer after a one-day strike, the first in Taiwan's chip industry. They want 15% of operating profit. Talks resume 27 October.
Firebird has a US export licence for 45,000 Nvidia GPUs for a $5 billion, 109 MW AI data center in Kazakhstan due in 2027. It runs about a seventh of that today, and has named no lender or customer.
Meta's signed but unstarted leases rose to $355.7 billion at 30 September, from $279 billion in June, on top of $454.5 billion of purchase commitments. Each lease starts only at handover, from late 2026 to 2036, while third-quarter free cash flow was negative $2.1 billion.
Micron showed a roadmap with vertical HBM, memory stacked straight onto the processor, aimed at 2028, and said memory supply will not catch up with demand. Heat and bonding now set the pace.
Explained
The Wire flags the move. Explained traces the link between a bottleneck, the buyer, and the cash flow.
A finance lease is booked like buying with borrowed money; an operating lease is booked as rent. Same building, same rent, different lines. Microsoft's new 25-year building life moves leases off its capex line.
SoIC is TSMC’s wafer-level 3D stack: chips bonded on top of chips, then placed in CoWoS. HBM still sits beside the GPU. Heat, not the copper join, is the bind.
AI data centers sit in a few dense clusters, led by Northern Virginia and Texas, with Johor, Singapore, Sydney and Melbourne in Asia-Pacific. They go where a utility can promise megawatts on a date, so the map of signed power is far bigger than the map of halls switched on.
Free daily brief
One five-minute brief each day: what changed across AI infrastructure, why it matters, and what to watch next.
Confirm your email to start receiving the Wire.
