Deployment & demand · United States of America

Microsoft

MSFT US Work there, own it, buy from it ↓

Hyperscale compute buyer. Holds none of the How this is computedCount of rows in the bottleneck corpus: every material plus every non-material chokepoint.Method →Code → bottlenecks tracked here; listed as context for its step of the chain.

Bottlenecks held0 of 33none of the tracked rows Directory gradeConcentrateda judgement, not a sourced fact

Why graded so · On the demand side the grade reads the other way: a handful of buyers account for most of the world’s accelerator orders.

Role cited from azure.microsoft.com ↗. The grade is this project’s judgement. Report an error

01What it rests on, and what rests on it

each row quotes the page it rests on

Depends on

Grid interconnection queues
“Limitations in power availability, delays in obtaining power connections, outages, shortages, increased energy costs, or requirements imposed by utilities, regulators, or other market participants could restrict our ability to develop or expand datacenter capacity.”
sec.gov · primary source · Fiscal 2026 10-K risk factor. Names delays in power connections, not the interconnection queue by name.
Leading-edge foundry capacity
“Fabricated on TSMC’s cutting-edge 3-nanometer process, each Maia 200 chip contains over 140 billion transistors”
blogs.microsoft.com · primary source · Microsoft’s own Maia 200 accelerator only, not the GPUs it buys.
High-bandwidth memory stacking yield
“a redesigned memory system with 216GB HBM3e at 7 TB/s”
blogs.microsoft.com · primary source · Maia 200 only. Names HBM3e, not its supplier.

A recorded dependency is a sentence, not a share, a volume or a contract. Walk it on a portfolio →

02SEC filings

3 most recent, from SEC EDGAR

The company’s own filings, from SEC EDGAR, fetched hourly. Substrata has not judged their effect on any bottleneck.

03Others at the same step of the chain

directory rows · grades are judgements

Amazon Web ServicesHyperscale compute buyer and custom siliconConcentrated
GoogleHyperscale compute buyer and custom siliconConcentrated
MetaHyperscale compute buyerConcentrated
OpenAIFrontier model developerConcentrated
AnthropicFrontier model developerConcentrated
xAIFrontier model developerConcentrated
CoreWeaveSpecialist compute providerCompetitive

04Ways to take part

how to reach it, not whether to

Not investment advice. This says how the public can reach a company — a job, its shares, its products — not whether anyone should. It takes no account of anyone’s circumstances, and a company holding a chokepoint is not a reason to own it. What this is not

Work there: It publishes its roles on its own careers page.Official careers page ↗
Own part of it: Its shares trade publicly as MSFT US, bought and sold through a broker that reaches that exchange.MSFT US listing record ↗Its SEC filings
Buy from it: What it does in the chain: Hyperscale compute buyer. At this depth of the chain most sales are to other businesses, under contract. Its own page describes what it offers.Its own page ↗

05Hiring

official careers page

Its roles are published on its own site, which Substrata links rather than reads. Official careers page ↗

06What this holds up

7 loops

Frontier training runone turn: Weeks to months
depends on Grid interconnection queues, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source
Energyone turn: Months to years
depends on Grid interconnection queues, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source
Software scaffoldone turn: Minutes
depends on Leading-edge foundry capacity, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source
Firmware and programmable logicone turn: Minutes to hours
depends on Leading-edge foundry capacity, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source
Frontier training runone turn: Weeks to months
depends on Leading-edge foundry capacity, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source
Software scaffoldone turn: Minutes
depends on High-bandwidth memory stacking yield, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source
Frontier training runone turn: Weeks to months
depends on High-bandwidth memory stacking yield, which gates it.
Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15 · this join is primary source

A loop is one design, build and measure turn. Severity is a dated judgement on a 0–12 scale, not a measurement, and the judgements are not added together — the figure shown is the worst single one.

07Sources

1 cited

azure.microsoft.com ↗ — its role in the chain: Hyperscale compute buyer

Discussion

Comments use threadkit (permission is participation). Sign in with OrangeCat to write. Fact-check is the same assistant as Ask, grounded in the corpus.

No comments yet.