AI Supply Chain
AI Supply Chain is a data page on the AI desk of Glassbox Auditing. The AI build-out runs from chips to data centres to power; spending at one layer becomes revenue at the next, so strain or slack in one layer reaches the others.
Last updated: .
What is on this page today?
- Texas: 1,730 MW of AI data-centre power operating; residential electricity up 34.6% over 5 years against up 34.4% for the US — 12 tracked sites, 5,397 MW planned · Residential price 16.0 cents per kWh against 11.9 5 years earlier (twelve-month averages) · Epoch AI tracks the largest AI campuses, not every data centre · the two measures are set… (Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices), 8 Nov 2026)
- Tennessee: 1,373 MW of AI data-centre power operating; residential electricity up 25.4% over 5 years against up 34.4% for the US — 3 tracked sites, 1,958 MW planned · Residential price 13.7 cents per kWh against 10.9 5 years earlier (twelve-month averages) · Epoch AI tracks the largest AI campuses, not every data centre · the two measures are set… (Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices), 8 Nov 2026)
- Ohio: 1,244 MW of AI data-centre power operating; residential electricity up 46.2% over 5 years against up 34.4% for the US — 6 tracked sites, 2,514 MW planned · Residential price 18.4 cents per kWh against 12.6 5 years earlier (twelve-month averages) · Epoch AI tracks the largest AI campuses, not every data centre · the two measures are set… (Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices), 8 Nov 2026)
- Indiana: 1,215 MW of AI data-centre power operating; residential electricity up 27.8% over 5 years against up 34.4% for the US — 3 tracked sites, 2,966 MW planned · Residential price 17.0 cents per kWh against 13.3 5 years earlier (twelve-month averages) · Epoch AI tracks the largest AI campuses, not every data centre · the two measures are set… (Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices), 8 Nov 2026)
- Virginia: 962 MW of AI data-centre power operating; residential electricity up 39.9% over 5 years against up 34.4% for the US — 6 tracked sites, 1,321 MW planned · Residential price 16.7 cents per kWh against 11.9 5 years earlier (twelve-month averages) · Epoch AI tracks the largest AI campuses, not every data centre · the two measures are set… (Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices), 8 Nov 2026)
- Georgia: 653 MW of AI data-centre power operating; residential electricity up 23.7% over 5 years against up 34.4% for the US — 2 tracked sites, 653 MW planned · Residential price 15.1 cents per kWh against 12.2 5 years earlier (twelve-month averages) · Epoch AI tracks the largest AI campuses, not every data centre · the two measures are set… (Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices), 8 Nov 2026)
- Memory & storage stocks rose 132% in a year, the most of the AI supply chain's layers (the whole chain +22%) — equal-weighted average of each layer's stocks' price moves, dividends not included · closes through 6 Nov 2026 · bars are one-year moves of the five best layers (Stock exchanges' closing prices, 5 Nov 2026)
- WDC: capital spending $227M in the quarter to 2 Oct 2026, +211% on a year earlier — revenue +55% on a year earlier (SEC filings, 1 Oct 2026) WDC
- BE: capital spending $153M in the quarter to 30 Sep 2026, +1148% on a year earlier — revenue +138% on a year earlier (SEC filings, 29 Sep 2026) BE
- IREN: capital spending $2.13B in the quarter to 30 Sep 2026, +1084% on a year earlier — revenue -5% on a year earlier (SEC filings, 29 Sep 2026) IREN
- SLDP: capital spending $4.68M in the quarter to 30 Sep 2026, +720% on a year earlier — revenue -73% on a year earlier (SEC filings, 29 Sep 2026) SLDP
- TLN: capital spending $167M in the quarter to 30 Sep 2026, +695% on a year earlier — revenue +105% on a year earlier (SEC filings, 29 Sep 2026) TLN
- COHR: capital spending $776M in the quarter to 30 Sep 2026, +646% on a year earlier — revenue +50% on a year earlier (SEC filings, 29 Sep 2026) COHR
- LCUT: capital spending $3.49M in the quarter to 30 Sep 2026, +636% on a year earlier — revenue +3% on a year earlier (SEC filings, 29 Sep 2026) LCUT
- CLS: capital spending $274M in the quarter to 30 Sep 2026, +635% on a year earlier — revenue +83% on a year earlier (SEC filings, 29 Sep 2026) CLS
- AAOI: capital spending $367M in the quarter to 30 Sep 2026, +630% on a year earlier — revenue +127% on a year earlier (SEC filings, 29 Sep 2026) AAOI
- FIX: capital spending $211M in the quarter to 30 Sep 2026, +496% on a year earlier — revenue +54% on a year earlier (SEC filings, 29 Sep 2026) FIX
- BL: capital spending $2.59M in the quarter to 30 Sep 2026, +386% on a year earlier — revenue +8% on a year earlier (SEC filings, 29 Sep 2026) BL
- NXDR: capital spending $801,000 in the quarter to 30 Sep 2026, +385% on a year earlier — revenue +14% on a year earlier (SEC filings, 29 Sep 2026) NXDR
- PINS: capital spending $15.5M in the quarter to 30 Sep 2026, +379% on a year earlier — revenue +12% on a year earlier (SEC filings, 29 Sep 2026) PINS
Where does the data come from?
Items on this page come from Epoch AI (Frontier Data Centers); U.S. Energy Information Administration (electricity prices); Stock exchanges' closing prices; SEC filings; SEC 10-Q and 10-K filings (cash flow statements); U.S. Census Bureau, construction spending (value put in place); SEC EDGAR (annual reports on Form 10-K); World Semiconductor Trade Statistics (WSTS), Historical Billings Report. Each item links to its own source. All sources, how items are checked.
What does this page lead to?
- Natural Gas (Macro): New data centers draw on the grid, and gas plants supply the largest share of U.S. power, so the build-out adds to gas burned for electricity. (moves the same way; years; sector scale) Source: EIA natural gas data (API v2): weekly storage; monthly production, consumption, trade and prices.
- Cloud Prices (Companies): The clouds rent out the accelerators they buy; tight chip supply tends to hold GPU instance list prices up, while new capacity lets providers cut them. (either way, depending on the cause; months; sector scale) Source: AWS Price List Bulk API (us-east-1 list prices).
- Trade Exposure (Companies): Chips, servers and networking gear are largely made in Taiwan, South Korea and Mexico, so the build-out raises these companies' imports and their exposure to tariffs and export rules. (moves the same way; quarters; company scale) Source: UN Comtrade (U.S. Census trade data).
- Resources & Energy (World Powers): Chips and data centers need copper, gallium, germanium and rare earths; China refines most of several of these and has restricted exports of gallium and germanium since 2023. (moves the same way; years; global scale) Source: USGS Mineral Commodity Summaries (mine production, reserves, U.S. net import reliance).
- Business Activity (Macro): Orders for computers and electronic products and spending on data-center construction carry the build-out into the Census business indicators. (moves the same way; quarters; economy scale) Source: U.S. Census Bureau Economic Indicators (retail, durable goods, construction, housing, business formation, inventories, trade).
- Nation (AI): Spending on chips, servers and software counts as business investment in GDP, so the build-out shows in the tech investment and growth contribution on the Nation page. (moves the same way; quarters; economy scale) Source: U.S. Bureau of Economic Analysis (NIPA).
- Big Tech Map (Options): Hyperscalers' capital-spending plans, given on earnings calls, are the chip and equipment layers' revenue, so a raised or cut plan tends to move both groups' shares and option bets. (either way, depending on the cause; same day; company scale) Source: Yahoo Finance prices.
- Live Now (Markets): The largest AI chip and cloud companies are a large share of SPY by weight, so their moves carry much of the index's daily move. (moves the same way; same day; economy scale) Source: State Street, SPY daily holdings.
- AI Models (AI): Labs train on the clusters the supply chain delivers, so more accelerators and data-center power let them run larger training runs; Epoch AI records each notable model's training compute. (moves the same way; quarters; sector scale) Source: Epoch AI notable models catalogue (CC BY 4.0).
What feeds this page?
- Progress Level (AI): Each step up the capability index comes from training on larger chip clusters, so faster progress goes with heavier orders for accelerators, memory, networking and data-center power. (moves the same way; quarters; sector scale) Source: Epoch AI Capabilities Index (CC BY 4.0).
- AI Companies (AI): Money raised by AI developers is mostly spent on compute: chips, cloud contracts and data centers, which is revenue for the supply chain's layers. (moves the same way; quarters; sector scale) Source: SEC Form D data sets: private offerings under Regulation D (DERA, 2008 on).
- AI Trajectory (AI): The compute vector reads each company's clusters and capital spending from Epoch AI and SEC filings; that spending is chip, server and data-center revenue for the supply chain. (moves the same way; quarters; sector scale) Source: Epoch AI, AI Supercomputers (CC BY 4.0).
- AI Models (AI): Training compute for notable models has grown several-fold a year (Epoch AI); each larger training run is spending on chips, power and data-center space. (moves the same way; quarters; sector scale) Source: Epoch AI notable models catalogue (CC BY 4.0).
- Nation (AI): Investment in computers and software in GDP is, on the other side, the sales of the chip, server and software companies in the supply chain. (moves the same way; quarters; sector scale) Source: U.S. Bureau of Economic Analysis (NIPA).
- Signals (Briefing): The AI capital-spending signal flags companies whose spending surges while free cash flow falls, the strain the AI supply chain is watched for layer by layer. (moves the same way; quarters; sector scale) Source: SEC EDGAR XBRL company facts (10-Q/10-K).
- Company Reports (Calendar): The largest cloud companies' capital-spending plans, given with their results, set the order books of chip, network and power suppliers. (moves the same way; the next quarters; sector scale) Source: SEC EDGAR XBRL company facts (10-Q/10-K).
- IPO Watch (Calendar): Anthropic's compute bill is paid to data-center landlords, neo-clouds and chip makers under multi-year contracts, so a new lease, a funding round or a delay tends to show first in those suppliers' shares. (moves the same way; on each contract or funding round; sector scale) Source: CoreWeave news releases.
- Company Research (Companies): The biggest companies' disclosed purchase commitments and data-center leases are orders already placed with chipmakers, server makers and builders, so rising commitments feed AI supply-chain revenue. (moves the same way; quarters; sector scale) Source: SEC EDGAR filings (10-Q/10-K text).
- Valuation (Companies): High valuations on the AI leaders let them raise money cheaply in stock and bonds, which pays for chips and data centers further down the chain. (moves the same way; quarters; sector scale) Source: Yahoo Finance quarterly statements.
- Bond Issuance (Companies): Large bond sales by the cloud companies pay for data-center building, so money raised in bonds turns into orders for chips, power equipment and construction. (moves the same way; quarters; sector scale) Source: SEC EDGAR filings (10-Q/10-K text).
- Supply Map (Companies): Chipmakers and server builders name the cloud companies among their largest customers; when those customers raise or cut spending, it runs down the chain layer by layer. (moves the same way; quarters; sector scale) Source: SEC EDGAR filings (10-Q/10-K text).
- Cloud & Hubs (Companies): Data-center floor space in annual reports is the room the chips, servers and power equipment fill, so a growing footprint means orders down the AI supply chain. (moves the same way; quarters; sector scale) Source: SEC EDGAR filings (10-Q/10-K text).
- Energy Industry (Companies): Power developers, utilities and grid engineers in the directory are a layer of the AI build-out: data centers need new generation and transmission before they open. (moves the same way; years; sector scale) Source: BTU Graph open energy directory (graph.json, companies.csv; CC BY 4.0, attribute BTU Graph).
- News (Feed): Large funding rounds at AI labs are money for compute: much of it is spent on GPUs and cloud capacity, revenue for the suppliers on AI Supply Chain. (moves the same way; quarters; sector scale) Source: Data Cube AI News API (AI-written summaries of AI industry news and funding rounds, with links to the original stories).
- Web Tracking (Feed): Search interest in AI tools and cloud services is a read on end demand; Web Tracking checks its AI indices against the suppliers' reported revenue on AI Supply Chain. (moves the same way; 1-2 quarters; sector scale) Source: SEC EDGAR XBRL company facts (10-Q/10-K).
- Events & Interviews (Feed): Capital spending plans announced at investor days and conferences, such as data-center budgets, are orders for the chip, power and construction suppliers on AI Supply Chain. (moves the same way; quarters; sector scale) Source: SEC EDGAR XBRL company facts (10-Q/10-K).
- Technology (Global Intelligence): The AI data centers on the map are where chip, server and power spending lands, so new sites and expansions feed the revenue of the chip and equipment makers on AI Supply Chain. (moves the same way; quarters; sector scale) Source: Epoch AI, AI Supercomputers (CC BY 4.0).
- Sanctions & Watchlists (Global Intelligence): Commerce's Entity List, part of the Consolidated Screening List, bars U.S. exports to listed firms without a license, which has cut chip and chip-equipment makers' sales to Chinese customers. (moves the other way; quarters; sector scale) Source: US Consolidated Screening List (Commerce, State, Treasury).
- Conflict Zones (Global Intelligence): Chinese exercises that close parts of the Taiwan Strait sit on the lane for chips made in Taiwan, where TSMC makes most leading-edge chips; chip buyers' annual reports name this as a supply risk. (either way, depending on the cause; days; sector scale) Source: SEC EDGAR filings (10-Q/10-K text).
- Business Activity (Macro): Non-residential construction includes the data centers and factories being built, so its growth runs alongside the capital spending that chip, power-equipment and cloud companies report. (moves the same way; quarters; sector scale) Source: U.S. Census Bureau Economic Indicators (retail, durable goods, construction, housing, business formation, inventories, trade).
- Electricity (Macro): Data-centre load adds to grid demand and to commercial electricity sales, where EIA counts most data centres; the power layer (electrical gear, turbines, cooling, contractors) sells into that build-out, so its orders and revenue depend on how much capacity is added. (moves the same way; quarters to years; sector scale) Source: EIA Hourly Electric Grid Monitor (Form EIA-930): daily demand by balancing authority (API v2).
- Stock by Stock (Markets): Chip, memory, networking and data-center stocks tend to move together because their revenue comes from the same AI spending; a big move in one shows up across the layers on AI Supply Chain. (moves the same way; same day; sector scale) Source: Yahoo Finance prices.
- Every Ticker (Options): When the options of a whole layer (chips, memory, data centers) lean the same way against their own histories, it is positioning on that layer's shared AI demand, which AI Supply Chain tracks in revenue. (moves the same way; same day; sector scale) Source: Robinhood options market data (OPRA).
- Big Tech Map (Options): Hyperscalers' data-center spending is revenue for chip, memory, networking and power suppliers, so a shift in how Big Tech options lean tends to show again in those suppliers' options and shares. (moves the same way; same day; sector scale) Source: SEC EDGAR XBRL company facts (10-Q/10-K).
- Alliances & Blocs (World Powers): The U.S. coordinates chip-equipment export controls with the Netherlands and Japan, so allied alignment decides whether ASML and Tokyo Electron can ship advanced tools to China. (either way, depending on the cause; months; global scale) Source: Federal Register API.