A $4.8 Million Veterans Giveaway, and the $10 Billion Microsoft Quarter Behind It insight cover
Private CompanyMSFT · ORCL · ACN•18 min read

A $4.8 Million Veterans Giveaway, and the $10 Billion Microsoft Quarter Behind It

OpenAI's Nov. 11 announcement — more than 20,000 servicemembers and veterans on free ChatGPT Plus years, verified through ID.me and delivered with veteran nonprofits — is a consumer-side constituent channel, not a federal contract: eligibility and redemption run through personal accounts, and no Veterans Affairs or Defense Department agreement exists. The federal AI channel is a separate and much larger machine, and it runs through Azure Government, where the compliance authorizations sit; the cash follows the gate, with Microsoft's September-quarter filing showing $10.0 billion of OpenAI-related revenue, a ~24% equity stake and a $7.2 billion receivable. With Democrats now controlling both chambers, the repricing lands on the debt-funded buildout of megawatts — Oracle's $28.5 billion of quarterly capex, CoreWeave's $60.7 billion of net debt, SoftBank's ¥1.1 trillion quarterly loss — while federal demand for AI software keeps compounding.

Published Nov 12, 2026Updated Nov 12, 2026

Veterans on free ChatGPT Plus years

20,000+

Claimed over the past year, per OpenAI on Nov. 11, 2026

List-price value of those free years

$4.8M

20,000 accounts at $20 a month, foregone at list price

Microsoft revenue from OpenAI arrangements

$10.0B

Quarter ended Sept. 30, 2026, including revenue sharing

OpenAI receivable at Microsoft

$7.2B

Up from $6.0B three months earlier

Veterans on free ChatGPT Plus years

20,000+

Claimed over the past year, per OpenAI on Nov. 11, 2026

List-price value of those free years

$4.8M

20,000 accounts at $20 a month, foregone at list price

Microsoft revenue from OpenAI arrangements

$10.0B

Quarter ended Sept. 30, 2026, including revenue sharing

OpenAI receivable at Microsoft

$7.2B

Up from $6.0B three months earlier

ChatGPT paying subscribers

50M

Alongside 900 million weekly users, as of Feb. 27, 2026

Federal ChatGPT Enterprise price

$1 per agency

One year via GSA's OneGov schedule, announced Aug. 6, 2025

Start with what OpenAI did not do. It did not sign a contract with the Department of Veterans Affairs or the Defense Department. Its published terms bar enterprise workspaces from redeeming the offer, verification runs through ID.me, and military partners are nonprofits, not procurement offices. On Nov. 11, 2026 the company marked the first anniversary of a program that opened on Nov. 10, 2025 for anyone within 12 months of separation or retirement: more than 20,000 claims, military spouses added, and seven more OpenAI Academy events scheduled from St. Louis on Nov. 12 to San Antonio on Nov. 21. The free year routes through personal accounts and ID.me military verification, which is why this is positioning rather than a distribution channel.

The offer, in plain terms

A consumer giveaway sized to be political, not commercial

The economics are trivial against OpenAI's scale, which is exactly why the program is worth reading as something other than customer acquisition. Twenty thousand free years at the $20 monthly list price are worth about $4.8 million of foregone subscription revenue — against a revenue run rate approaching $70 billion, 50 million paying subscribers and 900 million weekly users. Spread against a cohort of several hundred thousand annual service-leaver transitions, the take-up is still a sliver. What the company buys is standing: a verified identity graph of veterans and military families, delivered by organisations they already trust, at a moment when the political weather for frontier AI turned.

  • Not a contract: no Veterans Affairs or Defense Department agreement exists, and at launch OpenAI said collaboration with both agencies was "not out of the question" rather than agreed.
  • One offer per subscriber, no stacking with other promotions, and the year expires 12 months after the claim, after which it reverts to $20 a month unless the user cancels.
  • Reach is built through nonprofits, not procurement: Veterans Forge, NPower, the Wounded Warrior Project and AI Collective Hampton Roads co-hosted more than a dozen trainings for nearly 1,000 people.
  • Renewal is the only real metric here — 20,000 claims against 50 million paying subscribers says the channel buys constituency, and the next twelve months say whether it converts.
The offer as published, and what each term actually does
TermPublished ruleInvestor read
EligibilityVerified veterans and retirees who left active duty in the past 12 months; service members transitioning out within the next 12 monthsA rolling 365-day intake window that refills continuously rather than a one-time cohort
VerificationID.me confirms military status and shares the result with OpenAIIdentity, not purchase history, is the gate — the channel's real asset
Where it livesPersonal ChatGPT accounts only; enterprise workspaces cannot redeemDeliberately outside enterprise and government contracting channels
Price and durationFree for 12 months from the claim, then the standard monthly rate unless cancelledA subsidised trial that converts into a $20 subscription or churns
DistributionOpenAI Academy with Veterans Forge, NPower, the Wounded Warrior Project and AI Collective Hampton RoadsBorrowed trust: the constituencies that shape the post-sweep Congress

Where federal OpenAI actually runs

The federal channel is a compliance gate, and the gate has an owner

The federal AI market OpenAI is actually chasing works like a licence market with an approval step. Agencies buy ChatGPT Enterprise through GSA's OneGov schedule at $1 per agency for a year, plus a 60-day window of unlimited use of advanced models. The Pentagon put ChatGPT in front of all three million Department personnel, and GenAI.mil passed one million unique users within two months of deployment. Accenture Federal Services is the paid implementation layer, with 15,000 professionals enabled on OpenAI's systems and FedRAMP-aligned pathways running through Impact Level 6. And the deployments themselves sit in Microsoft's Azure Government, because that is where the authorizations live. OpenAI's federal route runs through Microsoft's cloud rather than its own sales force, and the paperwork is what makes it possible.

Six links in the federal AI chain, and who holds each
LinkWho holds itEvidenceInvestor read
Constituent channelOpenAI, with ID.me and veteran nonprofits20,000+ free ChatGPT Plus years claimed in a year; spouses added Nov. 11, 2026Cheap political positioning with no procurement dependency
Institutional channelGSA's OneGov schedule$1 per agency for one year, plus 60 days of unlimited advanced modelsA price anchor that pushes agencies toward self-serve AI over bespoke builds
Defense scaleGenAI.milChatGPT available to all 3 million Department personnel; 1M+ unique users in two monthsDemand that no procurement cycle has to approve
Compliance gateMicrosoft Azure GovernmentAzure OpenAI Service authorized at FedRAMP High and DoD Impact Level 6; ChatGPT Gov ran inside agencies' Azure Government cloudsThe authorised cloud is both the moat and the margin
ImplementationAccenture Federal Services15,000 staff enabled, 3,000+ on Codex, 1,500 on Enterprise ChatGPT; IL6-aligned pathwaysIntegrators are paid for delivery, not for licences
Care settingsVA GPT and Microsoft 365 Copilot ChatAbout 100,000 VA GPT users; VA inspector general found neither tool classified high-impact AI in June 2026Federal clinical AI still runs on Microsoft's and the VA's own tools

One more link sits underneath all of it: the megawatts. OpenAI and Oracle agreed 4.5 gigawatts of additional Stargate capacity in July 2025, and five more sites brought planned capacity toward 7 gigawatts and more than $400 billion of investment. That layer is now financed with debt and equity rather than operating cash flow, which is why the post-election repricing of rates and permitting lands here rather than on the demand side.

The middleman arithmetic

Microsoft books the toll twice, and now carries the credit

One line in Microsoft's September-quarter filing captures the arrangement. Revenue from commercial arrangements with OpenAI, inclusive of revenue-sharing payments, came to $10.0 billion — roughly 11% of Microsoft's $92.3 billion of quarterly revenue — after $24.1 billion for all of fiscal 2026. Microsoft's as-converted interest in OpenAI slipped to about 24% from roughly 25% at June 30 after the October 2025 recapitalisation and new funding rounds, and the quarter still produced $272 million of net gains on OpenAI investments in other income, which implies OpenAI's private valuation rose faster than fresh capital diluted Microsoft's holding.

  • Revenue: $10.0 billion of OpenAI-related revenue in the quarter against $24.1 billion in fiscal 2026, on total Microsoft revenue of $92.3 billion and $331.8 billion respectively.
  • Credit: OpenAI's receivable at Microsoft rose to $7.2 billion from $6.0 billion in three months — the hyperscaler is financing the revenue ramp of the equity it already owns.
  • Ownership: about 24% as-converted, diluted from roughly 25% at June 30, 2026, alongside $272 million of net investment gains booked in the quarter.
  • Commitments: $13.0 billion of total funding commitments with $12.0 billion already funded, so the toll is increasingly paid by the middleman rather than the vendor.
Annualise the September quarter and Microsoft's OpenAI-related revenue runs at roughly $40 billion, against OpenAI's own revenue run rate approaching $70 billion. Most of OpenAI's revenue now transits Microsoft's accounts — through Azure consumption and revenue sharing — before a single federal contract is signed.

Post-election read

The sweep repriced the buildout, not the demand for federal AI

Democrats took both chambers on Nov. 3, and the market repriced the complex within days. Reuters reported that the Goldman Sachs broad AI basket fell 15% in October and more than 24% from its peak, with regulation risk and a slower infrastructure rollout cited as the drivers, and that markets pulled back expectations for further fiscal stimulus. The transmission mechanism matters more than the headline. Analysts pointed to the Texas and Georgia governor races, which together describe about a third of all AI data-centre spend and projects. So the sweep hits the physical buildout and the financing of it, while the buying side keeps running.

The pace of AI infrastructure rollout could slow, potentially quite markedly in Texas.

Reuters, Nov. 4, 2026, on the midterms' read-through for AI infrastructure

Demand did not soften with the vote. Palantir's U.S. government revenue nearly doubled year over year to $953 million in the September quarter, Oracle's cloud infrastructure revenue rose 121% to $7.39 billion, and CoreWeave's revenue rose 168% to $3.66 billion. What softened was the willingness to fund megawatts on trust, which is why the balance sheets carrying that buildout are where the damage lands.

Year-over-year revenue growth across the federal AI chain, latest reported quarter

Software and infrastructure demand kept compounding after the midterms; hourly federal services revenue did not.

Unit: % YoY

CoreWeave revenue

Q3 2026: $3.66B vs $1.36B a year earlier

168%

Oracle cloud infrastructure revenue

Quarter ended Aug. 31, 2026: $7.39B vs $3.35B

121%

Palantir U.S. government revenue

Q3 2026: $953M vs $486M

96%

Microsoft Azure revenue

Quarter ended Sept. 30, 2026: up $10.8B

48%

Booz Allen Hamilton total revenue

Quarter ended Sept. 30, 2026: $2.78B vs $2.89B

-3.9%

Who carries the buildout, and what the latest reported quarter cost them
CompanyLatest reported quarterCapacity or investmentProfit or lossLeverage
OracleQ1 FY2027, quarter ended Aug. 31, 2026$28.5B quarterly capex; $7.39B cloud infrastructure revenue$4.76B net incomeFree cash flow -$5.4B in the quarter
CoreWeaveQ3 2026, quarter ended Sept. 30, 2026$75.7B net property and equipment-$709M net loss$66.2B total debt, $5.5B cash; $846M interest expense in the quarter
SoftBank GroupFiscal Q2 2026, quarter ended Sept. 30, 2026$64.6B cumulative OpenAI investment for about 13%-¥1.1T net loss, against +¥2.5T a year earlierFinancing costs ¥462B; $3.7B loss on US data-centre power-equipment prepayments

Horizons

Two clocks: renewals and revenue now, the compliance moat later

  • Renewal season starts now: the 2025 cohort's free years begin expiring in November, and each converted account turns a giveaway into a $20 subscription at zero incremental acquisition cost.
  • Veterans Affairs procurement: the solicitation for third-party enterprise AI support services, expected around the start of fiscal 2027, is where integrators rather than model vendors get paid for the federal estate.
  • Microsoft's fiscal Q2 report in December refreshes both the $10.0 billion OpenAI revenue line and the $7.2 billion receivable; the ratio between them shows who is financing whom.
  • Permits: Texas and Georgia govern roughly a third of planned AI data-centre spend, so gubernatorial outcomes there set the pace for every capacity-backed name in the chain.
  • Compliance sets the federal price: Azure Government is the authorised substrate for OpenAI's federal offerings at every federal classification level, so the authorised cloud captures the margin whichever vendor wins the model race.
  • Seat prices are being floored: $1 per agency and $20 per veteran year are the same strategy at different layers — give the seat away, monetise the deployment.
  • Mix has already flipped: the revenue run rate is up more than 70% since early July on enterprise sales more than doubling, so the veterans channel is positioning rather than growth.
  • Governance is the ceiling in care settings: the VA inspector general found in June that the department never classified its own tools, or Microsoft's Copilot Chat, as high-impact AI, and that gap has to close before OpenAI's offerings can carry that load.
The thesis in one line: the veterans programme behaves as a $4.8 million hedge against a political cycle rather than a distribution channel, while the federal AI money keeps flowing through Azure Government and into Microsoft's revenue. The sweep hurts the borrowers of that trade — leverage funding megawatts in Texas and Georgia, and federal services sold by the hour — not the owners. Watch the receivable and the permits; the sign-up counts are a rounding error either way.

Where the federal AI dollar lands, and who carries the risk

MMicrosoftMSFT--
--Vol --
-
Bullish
  • About 11% of Microsoft's revenue came from OpenAI in a single quarter, $10.0 billion including revenue sharing, after $24.1 billion in fiscal 2026.
  • Azure Government is the compliance gate for federal OpenAI workloads at FedRAMP High and DoD Impact Level 6, so Microsoft owns the authorised substrate.
  • The $7.2 billion OpenAI receivable, up from $6.0 billion in three months, makes the position credit as well as equity.
  • Horizon, 1-3 years: a ~24% stake diluted toward lower percentages, plus revenue sharing on a ~$70 billion run rate, keeps OpenAI the largest swing factor in Microsoft's growth mix.
OOracleORCL--
--Vol --
-
Bullish
  • Cloud infrastructure revenue rose 121% to $7.39 billion in the August quarter, with remaining performance obligations of $664 billion.
  • Only about 13% of that backlog converts to revenue within a year, so $28.5 billion of quarterly capex precedes the receipts.
  • OpenAI's 4.5 gigawatt Stargate agreement and five further sites take planned capacity toward 7 gigawatts and more than $400 billion of investment.
  • Horizon, 1-3 years: if Texas and Georgia permits slow, Oracle's cash flow bends before its backlog does, with free cash flow already -$5.4B in the quarter.
AAccentureACN--
--Vol --
-
Bullish
  • Enabled 15,000 federal professionals on OpenAI systems, with 3,000+ on Codex, 1,500 on Enterprise ChatGPT and FedRAMP-aligned paths through Impact Level 6.
  • Revenue is earned on implementation and compliance rather than licence resale, which holds up even at $1 per agency.
  • Fiscal 2026 revenue of $74.18 billion and a 15.8% operating margin give it the balance sheet to absorb federal AI workflow build-outs.
  • Horizon, 12-24 months: the Veterans Affairs enterprise AI support-services solicitation, due around the start of fiscal 2027, is the next identifiable pipeline.
PPalantirPLTR--
--Vol --
-
Mixed
  • U.S. government revenue nearly doubled to $953 million in the September quarter, on total revenue of $2.31 billion, up 96%.
  • Commercial revenue of $1.17 billion now exceeds government revenue of $1.15 billion, so federal exposure is roughly half the business.
  • Counterweight: a $1-per-agency federal price floor and subsidised enterprise seats compress per-seat pricing for every federal software vendor.
  • Horizon, 1-3 years: net income of $1.13 billion and net cash protect it if procurement slows, but federal scrutiny hits the half of revenue that is government.
CCoreWeaveCRWV--
--Vol --
-
Bearish
  • Revenue grew 168% to $3.66 billion, but the quarter produced a $709 million net loss and $846 million of interest expense.
  • Carries $60.7 billion of net debt and $75.7 billion of net property and equipment against $5.5 billion of cash.
  • Full-year 2026 guidance of $13.0-13.3 billion and a $20-21 billion exit run rate depend on data centres that still need permits in the states where the sweep bites.
  • Horizon, 6-12 months: financing terms and energisation schedules matter more than the demand backdrop, which is not the problem.
BBooz Allen HamiltonBAH--
--Vol --
-
Bearish
  • Revenue fell 3.9% year over year to $2.78 billion in the September quarter, with net income of $179 million and $334 million of cash on hand.
  • The divergence from federal AI software growing 96% to 168% suggests federal spending is rotating from hourly services to licensed tools.
  • With a services model exposed to that substitution and little cash, pricing pressure shows up in operating income before it shows up in guidance.
  • Horizon, 1-3 years: veterans' employment services and federal IT modernisation both compete with the same free tools now reaching the transition cohort.
9SoftBank Group9984.T--
--Vol --
-
Bearish
  • Financing costs more than doubled to ¥462 billion in the September quarter as the group funded the US data-centre build-out.
  • The quarter swung to a ¥1.1 trillion net loss, against a ¥2.5 trillion profit a year earlier, after $64.6 billion went into about 13% of OpenAI.
  • A $3.7 billion loss on advance payments for power-generation equipment for US AI data centres shows a financing company now carrying construction risk.
  • Horizon, 12 months: with shares nearly halved since early June, the next OpenAI mark and any further funding round decide whether the 13% stake is support or funding gap.