The largest cloud companies report how much they spend on data centers. Producer prices and company backlogs show whether the equipment behind that spending is getting scarce. The radar puts both next to the companies that might benefit, with the uncertainties and the data that would prove each idea wrong.
Research, not adviceCatalog RADAR-v1Sources refreshed 2026-10-23 02:25 UTC34 of 34 source series current
The lines are Boostio's hypothesis chain, not measured money flows. Numbers: SEC filings and BLS producer prices, with dates and links below. A green outline marks a paper-tested candidate; an orange one, a triggered invalidation rule. The dashed nuclear branch is research only: no paper trades.
Demand
$601.5B of cash capex in 12 months, +82.6%
Cash paid for property and equipment by the five largest cloud and AI platform companies, from their SEC filings, over each company's latest four reported quarters, against the same companies a year earlier. Sum of each company's latest trailing-12-month cash capex; fiscal quarters differ by up to two months between companies.
Bars: the last eight quarters; dimmer bars are derived by subtracting year-to-date totals in consecutive filings. Capex is not megawatts, memory or turbines: it mixes assets, leases, regions and schedules. Boostio does not convert it into physical units. Demand can be double-counted: a hyperscaler purchase and its supplier's sale are the same investment seen twice.
Supply and pricing
What the equipment costs now
Producer price indexes from the U.S. Bureau of Labor Statistics: what U.S. makers charge for this equipment, month by month. Rising prices alongside rising demand fit a shortage; they do not prove one.
Power deliveryPower, distribution and specialty transformers
+7.8%in 12 months
+13.5% in 3 years · latest Sep 1, 2026 (index 475.065)
Producer price indexes measure prices for a whole industry, not a company's orders or a shortage itself.
Bottleneck · research candidates
Power delivery
Turbines, transformers, switchgear, grid construction and data-center power and cooling. Data centers need electricity delivered, not just generated, and grid equipment has long lead times.
GEVGE Vernova
Gas and wind turbines, grid equipment
Research only
Company facts (SEC filings, Nasdaq.com)
Backlog (RPO)
$176.3B
+37.0% y/y · Jun 30, 2026
Revenue, 12 months
$41.4B
+13.0% y/y · to Jun 30, 2026
Close Oct 22, 2026
$827.42
+41.4% in a year · −29.6% vs 52-week high
Weekly points from daily closes, about one year (final bars only)
Why it might benefit · Boostio's hypothesis
Makes gas turbines and grid equipment (transformers, switchgear, HVDC). If data-center demand keeps power equipment scarce, a long order book can support prices and margins.
Uncertainties
Its remaining performance obligations cover every segment (including wind and long service contracts), so they are not an AI-specific backlog.
Turbine and transformer capacity additions across the industry can end the shortage.
Orders can be delayed or cancelled when projects slip.
The share price may already reflect the backlog.
What would invalidate it
✓Remaining performance obligations fall year over year in a new filing. (now +37.0%)
Builds and upgrades transmission, substations and data-center connections
Research only
Company facts (SEC filings, Nasdaq.com)
Backlog (RPO)
$33.5B
+75.1% y/y · Jun 30, 2026
Revenue, 12 months
$32.9B
+26.3% y/y · to Jun 30, 2026
Close Oct 22, 2026
$552.44
+26.3% in a year · −29.6% vs 52-week high
Weekly points from daily closes, about one year (final bars only)
Why it might benefit · Boostio's hypothesis
Builds and upgrades transmission lines, substations and large-load connections. More data-center load means more grid work, which shows up in its remaining performance obligations.
Uncertainties
Utility and renewable work dominate its backlog, not only data centers.
Labour, permitting and utility spending cycles can slow conversion of backlog to revenue.
Contracting margins are thin; revenue growth does not guarantee profit growth.
What would invalidate it
✓Remaining performance obligations fall year over year in a new filing. (now +75.1%)
Switchgear, power distribution and data-center electrical systems
Research only
Company facts (SEC filings, Nasdaq.com)
Revenue, 12 months
$30.0B
+15.5% y/y · to Jun 30, 2026
Close Oct 22, 2026
$368.99
−1.2% in a year · −19.8% vs 52-week high
Weekly points from daily closes, about one year (final bars only)
No machine-readable backlog for this company in SEC filings, so the evidence here is revenue only.
Why it might benefit · Boostio's hypothesis
Sells switchgear, transformers and power-distribution systems into data centers and utilities. Electrical gear is one of the slowest items to source for a new data center.
Uncertainties
Its remaining performance obligations have not been tagged in SEC XBRL since 2024-03-31, so there is no machine-readable backlog evidence here.
Data centers are one segment of a diversified industrial company.
Higher prices for electrical gear can attract new capacity.
What would invalidate it
✓Trailing 12-month revenue falls year over year. (now +15.5%)
Operating nuclear generators, fuel and enrichment suppliers and reactor developers tied to AI power demand. Approved as a research theme from a reviewed reel, not as a basket of stocks: each company sits at a different stage, and regulatory approvals, contracts and funding are shown with their own dates and conditions. Research only: no paper trades.
Sources checked Oct 13, 2026: issuer releases, counterparties, the NRC, DOE and the SEC. Starting point: reel Dd9UpDFPBae. The full 78-second reel was transcribed and its frames inspected. It pitched five stocks that "can 10x"; Boostio approved a nuclear and AI-power research theme instead, not a basket. Status labels: Confirmed (by a regulator, the SEC or the counterparty), Issuer-stated (stated by the company itself), Target (a plan or target date, not a result), Unresolved (contradicted or not validated).
Operating generatorAlready produces and sells electricity from nuclear plants.
VST
Fuel and enrichmentSupplies or enriches reactor fuel; does not run reactors.
LEU
Reactor developerDesigns or builds new reactors; no commercial reactor electricity yet.
XE · OKLO · SMR
VSTVistra Corp.
NYSE · Integrated power generation and retail electricity: nuclear, gas, coal, solar and storage
Research only · no paper trades
ExposureOperating generator and retailer (diversified)
Revenue stageSells electricity today; figures are whole-company, not nuclear-only
GAAP, same quarter, attributable to the company · $327.0M a year earlier
Whole-company figures, not nuclear-segment results; GAAP net income includes hedge marks.
Why Boostio watches it · hypothesis
Owns nuclear plants that already produce and sell electricity, and Meta named three of them in its January 2026 nuclear announcement. More data-center demand can raise the value of existing carbon-free baseload power.
Uncertainties
Whole-company results are not nuclear-segment results: the fleet also runs on gas, coal, solar and storage.
GAAP net income includes hedge marks (a $472 million unrealized loss in Q2 2026, per the issuer), so one quarter says little.
Power and capacity prices move with policy, weather and gas prices, not only AI demand.
Milestones, dated, with their status
Jan 2026ConfirmedMeta's January 2026 nuclear announcement identifies operating Vistra nuclear plants at Perry, Davis-Besse and Beaver Valley.Meta
Aug 7, 2026ConfirmedQ2 2026 results: GAAP net income of $305 million, including a $472 million unrealized loss on hedges that settle in future years.Vistra
Unresolved limits
Nuclear-segment economics are not separated in the figures here.
What would change the picture (watched by hand)
A data-center power agreement involving its nuclear plants is cancelled or not renewed.
A nuclear unit is retired or has an extended outage.
Revenue stageRevenue from fuel sales and contract services; profitable in Q2 2026
Reactor / operating stageNot a reactor operator: demonstration HALEU production under its NRC license
Latest quarter filed with the SEC
Revenue, quarter to Jun 30, 2026
$176.1M
10-Q filed Aug 6, 2026
Net income, attributable
$16.8M
GAAP, same quarter, attributable to the company · $28.9M a year earlier
Revenue is recognized sales and services; optional DOE purchases and customer prepayments are not revenue until earned.
Why Boostio watches it · hypothesis
Most advanced reactor designs are expected to need HALEU, and Centrus is producing it today under an NRC license (more than 1,900 kg of HALEU UF6 so far) with a DOE expansion contract and a signed enrichment contract with X-energy. Scarce fuel could be a bottleneck that pays.
Uncertainties
It is not the only licensed HALEU producer: NRC names two, at different maximum enrichment.
Most of the expansion value is optional DOE purchases and planned capacity, not earned revenue.
Advanced reactors may arrive later than planned, delaying HALEU demand.
Milestones, dated, with their status
Jun 2026Issuer-statedFinal 900 kg of HALEU UF6 under its DOE contract produced in mid-June, more than 1,900 kg cumulative.Centrus
Jul 1, 2026Issuer-statedSigned a $900 million DOE task order to expand HALEU production; total value $1.07 billion including discretionary options.Centrus
2029TargetFirst new enrichment capacity expected by 2029 (issuer expectation).Centrus
Aug 6, 2026Issuer-statedDefinitive contract to provide X-energy with LEU and HALEU enrichment services; customer prepayments support the expansion (posted on its site August 10).Centrus
Unresolved limits
"The only U.S. company allowed to make it" is overstated: NRC lists two licensed HALEU producers.
"Nine out of ten new designs" refers to the reactors selected by DOE's Advanced Reactor Demonstration Program, not all designs.
HALEU UF6 mass is not uranium mass, and NRC's FAQ license quantity (600 kg) is a snapshot, not the current license conditions.
What would change the picture (watched by hand)
The DOE task order or the X-energy contract is cancelled or not funded.
Another enricher is licensed for 19.75% HALEU at scale.
Nasdaq · Small modular reactor (Xe-100) and TRISO-X fuel developer
Research only · no paper trades
ExposurePre-commercial reactor and fuel developer
Revenue stageServices revenue and grant income; no reactor electricity sales
Reactor / operating stageNo operating commercial Xe-100; first projects are targets for the early 2030s
Latest quarter filed with the SEC
Revenue, quarter to Jun 30, 2026
$54.6M
10-Q filed Aug 13, 2026
Net income, consolidated
−$105.3M
GAAP, same quarter, incl. noncontrolling interests · −$88.8M a year earlier
The tagged revenues include grant income: the issuer's Q1 2026 release split $43.4 million into $39.9 million of services revenue and $3.5 million of grant income.
Since its April 2026 IPO, X-energy reports through an Up-C structure: the loss attributable to X-Energy, Inc. covers only its post-IPO share and is not comparable with earlier periods, so the consolidated net loss (including noncontrolling interests) is shown.
Why Boostio watches it · hypothesis
Develops factory-built modular reactors and their fuel, with an Amazon investment and a planned Energy Northwest project. If AI data centers need new firm power, early reactor developers with fuel supply and anchor customers could be positioned for it.
Uncertainties
No commercial Xe-100 has produced electricity; faster and cheaper modular deployment is a design claim, not a result.
Program funding is cost-shared and conditional; grants, tax credits, equity and program ceilings are not the same as government cash.
Listed only since April 24, 2026: a short public record.
Milestones, dated, with their status
Oct 13, 2020ConfirmedDOE Advanced Reactor Demonstration Program award: $80 million initially; later funding contingent on appropriations, satisfactory progress and continuation approval.DOE
undatedTargetAmazon invests in X-energy; the Energy Northwest project starts around 320 MW (optionally 960 MW) and is expected to meet regional needs in the early 2030s.Amazon
Apr 24, 2026ConfirmedBegan trading on Nasdaq as XE; the IPO closed April 27 with about $1.1 billion of net proceeds.X-energy
Aug 6, 2026Issuer-statedDefinitive LEU and HALEU enrichment contract with Centrus (Centrus announcement).Centrus
Unresolved limits
The reel's "up to $2.1 billion" of government backing is not validated: the issuer's FAQ states $1.2 billion of cost-shared program funding for the first Xe-100 deployment and the first TRISO-X fuel plant, and no current DOE award amendment was found.
The composition of the reported revenue (services versus grant income) is shown separately by the issuer; the SEC figure here is labelled as filed.
What would change the picture (watched by hand)
The Energy Northwest project or the Amazon partnership is cancelled or delayed beyond the early 2030s.
NYSE · Fast-fission power plant developer; isotope and test-reactor activity
Research only · no paper trades
ExposurePre-commercial reactor developer
Revenue stageSmall company revenue; its composition was not established and it is not reactor electricity sales
Reactor / operating stageLow-power test reactor critical under DOE authorization; no NRC commercial operating license
Latest quarter filed with the SEC
Revenue, quarter to Jun 30, 2026
$1.21M
10-Q filed Aug 7, 2026
Net income, attributable
−$48.5M
GAAP, same quarter, attributable to the company · −$24.7M a year earlier
The business-line composition of this revenue was not established; it is not reactor electricity sales.
Why Boostio watches it · hypothesis
Meta signed a development partnership for an Oklo nuclear campus in Pike County, Ohio, and Oklo's Groves test reactor reached first criticality in August 2026. Reactors built to serve data-center regions are a direct bet on AI power demand.
Uncertainties
The Meta campus is a future project (as early as 2030): no Oklo reactor delivers power to Meta today.
A DOE-authorized test reactor is not an NRC commercial license for its Aurora plants.
Large losses relative to revenue: progress depends on financing.
Milestones, dated, with their status
Jan 2026TargetMeta partnership for a Pike County, Ohio campus: up to 1.2 GW into the PJM grid, possibly online as early as 2030, with early procurement and development funding.Meta
Aug 6, 2026Issuer-statedGroves low-power isotope test reactor achieved first criticality under DOE authorization.Oklo
Unresolved limits
"Oklo has no revenue" is out of date at company level, but the business-line composition of its revenue was not established.
The reviewed sources do not establish a binding power purchase agreement or an NRC construction or operating license for the Meta campus.
What would change the picture (watched by hand)
The Meta campus partnership is cancelled or its 2030 target slips materially.
NYSE · Small modular reactor designer (NuScale Power Module)
Research only · no paper trades
ExposureSMR design company
Revenue stageMinimal revenue; no plant sales or electricity yet
Reactor / operating stageNRC standard design approval for the US460 (May 2025); no operating plant
Latest quarter filed with the SEC
Revenue, quarter to Jun 30, 2026
$75,000
10-Q filed Aug 5, 2026
Net income, attributable
−$47.5M
GAAP, same quarter, attributable to the company · −$17.6M a year earlier
Net income, consolidated
−$50.1M
GAAP, same quarter, incl. noncontrolling interests · −$37.6M a year earlier
NuScale reports noncontrolling interests (an Up-C structure): the loss attributable to NuScale Power Corporation and the consolidated loss are both shown.
Why Boostio watches it · hypothesis
Holds an NRC standard design approval for its US460 small modular reactor (six 77-MWe modules, 462 MWe), and its partner ENTRA1 is negotiating with TVA toward a large deployment program. A licensed design can shorten the path to new firm power for data centers.
Uncertainties
A design approval is not a site construction or operating authorization, and not revenue.
The TVA program is still in discussions, not a definitive power purchase agreement.
Other approval paths exist: NRC issued construction permits for the Kairos Hermes 2 test facility in November 2024.
Milestones, dated, with their status
May 2025ConfirmedNRC issued the US460 standard design approval: six 77-MWe modules, 462 MWe (NRC project page reviewed September 16, 2026).NRC
Aug 2026Issuer-statedENTRA1 continues discussions with TVA toward a definitive power purchase agreement (Q2 2026 release); not a signed agreement.NuScale
Unresolved limits
"The only small reactor approved" holds narrowly for its design approval; it is not universal regulatory exclusivity or an operating approval.
What would change the picture (watched by hand)
The ENTRA1/TVA discussions end without a definitive agreement.
The design approval is not followed by a site construction application.
ConfirmedNRC: two companies hold licenses to produce HALEU (uranium enriched to between 5% and 20% U-235). Centrus may produce up to 600 kg enriched up to 19.75%; Louisiana Energy Services may produce HALEU enriched up to 5.5%.NRC
ConfirmedA lower-assay HALEU license is not the same as 19.75% capability: enrichment level, licensee, facility and process are tracked apart.NRC
Issuer-statedCentrus reports more than 1,900 kg of HALEU UF6 produced cumulatively under its DOE contract, beyond the NRC FAQ's 600 kg snapshot: FAQ pages cannot substitute for current license conditions and production records.Centrus · NRC
Issuer-statedCentrus: "9 out of the 10 reactors selected by the Department of Energy's Advanced Reactor Demonstration Program" are expected to run on HALEU. A selected cohort, not all new designs.Centrus
Left out on purpose
Return promises in the reel ("10x", "generational wealth", "six figures a month") are unverified promotion: no audited performance, valuation or forecast method was given. They are not shown as signals.
A claim that Peter Thiel "sold out completely" and returned with $490 million "mostly in power" is unverified: no exact fund, reporting quarter or holdings filing was identified, so it is not in the fund-holdings feed.
Valuations: no dated price and share-count source was validated, so Boostio does not compute them for these companies.
What the reel claimed, and what the sources show
Claim
What the sources show
Verdict
ClaimVistra owns nuclear plants, sells the electricity and makes money today.
SourcesQ2 2026 GAAP net income $305 million; an operating fleet across nuclear, gas, coal, solar and storage; Meta names three operating Vistra nuclear plants.Vistra · Meta
Supported
A diversified operating generator, not a nuclear-only developer.
ClaimX-energy builds factory-made reactors and is partnered with Amazon.
SourcesListed on Nasdaq (XE) since April 24, 2026; Amazon invested and the Energy Northwest project targets the early 2030s.X-energy · Amazon
Partly supported
Ticker and partnership hold; "they built reactors" does not: no commercial Xe-100 output yet.
ClaimThe government put up to $2.1 billion behind X-energy.
SourcesThe issuer FAQ states $1.2 billion of cost-shared program funding; DOE's 2020 award was $80 million initially, later funding conditional.X-energy · DOE
Not validated
Kept open until a current DOE award amendment is found.
ClaimCentrus is the only U.S. company allowed to make HALEU, and 9 of 10 new designs need it.
SourcesNRC names two licensed HALEU producers (up to 19.75% and up to 5.5%); "9 of 10" refers to DOE's selected demonstration reactors.NRC · Centrus
Overstated
Real fuel exposure; no monopoly and no market-wide 90%.
ClaimOklo puts reactors next to data centers; Meta signed an Ohio deal.
SourcesMeta confirms a Pike County, Ohio campus of up to 1.2 GW into PJM, as early as 2030; Q2 2026 revenue $1.21 million, net loss $48.5 million.Meta · SEC EDGAR
Partly supported
A future campus, not an operating data-center reactor; "no revenue" is out of date.
ClaimNuScale has the only small reactor approved by the U.S. regulator; everyone else is waiting.
These would matter, and they are not here. Boostio lists them instead of estimating them.
Grid interconnection queues (Berkeley Lab "Queued Up")Not machine-retrievable here (the site answers automated requests with an access challenge); it was reviewed by hand in the source research. It counts proposed power plants seeking a grid connection, most of which withdraw, so it is neither completed supply nor a data-center connection queue.
Data-center connection and energization queuesNo consolidated public dataset; utilities report them inconsistently.
Backlogs of non-US equipment makers (Siemens Energy, Hitachi Energy, Mitsubishi Heavy)Not SEC filers, so no comparable XBRL data.
Vertiv backlog; Eaton remaining performance obligations after 2024-03-31Reported in earnings materials, not tagged in SEC XBRL.
HBM supply from SK Hynix and SamsungNot SEC filers; HBM output and contract prices are not published as official data.
Korean and Taiwanese chip export statisticsNot connected yet. When added they need interpretation: product category, price and mix, not just totals.
Finance leases and AI share of capexCash capex excludes data centers leased through finance leases, and no filing splits AI from other capex.
Nuclear valuations and current DOE award termsNo dated price and share-count source was validated for these companies, and no current DOE award amendment for X-energy was found, so neither is shown or estimated.
Paper test
No radar idea has been paper-tested yet
Radar ideas become paper trades only when Boostio's house research agent recommends one and an admin approves it. Until a simulated order fills, every candidate above is research only.
Forward paper record of the general stock book, after modeled slippage and fees. Not a backtest.
Does the radar work?
Not proven. The rule, fixed in advance: at least 30 closed paper trades over at least 90 days, and an edge only if the whole 95% interval of the excess return over the S&P 500 (SPY) is above zero. No backtest is shown as a result.
Method
Rules written down before the first scan
When an idea qualifies
Combined capex growing, with at least four of five filings less than 150 days old; the company's own filing less than 135 days old; a current final close; backlog or revenue up 10% or more in a year; and no invalidation rule triggered. A price that doubled in a year halves the suggested size.
From idea to paper trade
The house research agent turns a qualifying candidate into a sized recommendation; an admin approves or rejects each one with a reason. Approval re-checks the data and becomes one simulated market-on-open order in the general stock paper book (5% of the book by default, capped by theme and in total). Nuclear names are research only and never become paper trades. No broker, no real money.
Caveats that always apply
A shortage is not automatically profit: the question is who has pricing power, what new capacity removes the shortage and how much optimism the share price already holds. Remaining performance obligations are contracted revenue not yet recognized; they are company-wide, not AI-specific, and can be cancelled or delayed.
Freshness
Sources and their state
34 of 34 source series were read successfully at the last refresh (2026-10-23 02:25 UTC). A source that fails keeps its last good data, shown with its own older dates.
No source failed at the last refresh.
Questions
The radar, answered
What is an AI bottleneck?
A physical input the AI build-out needs faster than it can be supplied: electricity delivered to data centers (turbines, transformers, switchgear, grid work) and the high-bandwidth memory stacked next to AI chips. Boostio watches the demand side in the hyperscalers' SEC filings and the supply side in producer prices and company backlogs.
Is this a list of stocks to buy?
No. The candidates are research: each comes with Boostio's labelled hypothesis, the uncertainties against it and the data that would invalidate it. A shortage is not automatically profit, and the share price may already reflect it. Nothing here is investment advice.
Where does the data come from?
SEC EDGAR XBRL for capital spending, backlog (remaining performance obligations), revenue and gross margin; U.S. Bureau of Labor Statistics producer price indexes, read through FRED; and final daily closes from Nasdaq.com. Each number links to its source and shows its period and filing date.
Why are the nuclear companies research only?
They were approved as a research theme from a reviewed reel, not as stocks to buy. Each sits at a different stage (an operating generator, a fuel supplier and three reactor developers), several milestones are plans or company statements rather than results, and valuations were not validated. Boostio shows the evidence and its status; it places no paper trades on them.
What does paper-tested mean?
Boostio's house research agent can turn a qualifying candidate into a recommendation that an admin approves or rejects. An approved one becomes a simulated market-on-open order in Boostio's general stock paper book, filled at the real opening price with modeled costs. A candidate is paper-tested once such an order has filled; results are forward only.
Research, not advice. Boostio shows dated public data with its source and labels its own hypotheses as such. Paper only: nothing here is a real trade, and no result here is a promise.
Follow the research as it updates
Boostio is a research-first platform in private beta: crypto candidates, daily market stories, public paper trading, rival stocks and AI-infrastructure research. These research pages stay free and open.
