AI exposure can be a plausible thematic allocation, but not through an undifferentiated "AI is profitable / not profitable" filter.
Independent AI Investment Memo - As of June 22, 2026
Is AI Profitable Yet?
This page answers the original question with more nuance than a simple No/Yes counter: as an independent investment proposal sheet for AI exposure, cash flow, CapEx, revenue, opex, profit, loss and source quality.
Reference
A counter-model and clarification of the broad claim made by isaiprofitable.com.
ROI.
Not just "No." or "Yes."
Investment Proposal Sheet This page does not make a buy recommendation. It structures which AI exposures appear underwritable, which mostly carry momentum, and which data is still missing for a rigorous investment committee.
The relevant measure is the conversion of today's infrastructure investment into future free cash flows.
Hardware life, depreciation and demand ramp need to be evaluated together.
Chips, cloud backlog, enterprise products and embedded productivity with credible monetization.
Private labs and aggressive CapEx cases without segment margins, utilization evidence and financing clarity.
An investment case should show how CapEx, OpEx, depreciation, utilization, pricing power and source quality combine into a plausible payback path.
Financial Run-rate Monitor These counters are modeled run rates since page load. They are not real-time accounting; they show order-of-magnitude movement for CapEx, opex/costs, revenue, profit, loss and cash flow.
$563B/year forward envelope
Amazon, Meta, Microsoft annualized, Oracle and Anthropic. Timing band: 70-130% deployment pacing.
$606.2B/year mixed revenue anchors
AWS, Google Cloud, Microsoft AI ARR, Oracle Cloud, Nvidia Data Center annualized, Anthropic ARR.
$651.9B/year cash-generation baseline
Amazon FY, Alphabet FY, Microsoft Q3 x4, Meta Q1 x4 and Oracle FY. Mixed periods.
$173.6B/year FCF signal
OCF minus CapEx or reported FCF signals. Shows how strongly infrastructure compresses cash flow.
$438.2B/year net-income signal
Amazon FY net income, Microsoft Q3 x4 and Nvidia Q1 x4. Not AI-only, but relevant for financing capacity.
$20.9B/year OpenAI operating loss
Media/leak-based figure. Deliberately kept separate from official Big Tech filings.
$151.7B/year mixed cost base
Meta FY costs & expenses plus reported OpenAI 2025 costs. Not AI-only, but visible as a cost-pressure proxy.
Why counters instead of one number?
CapEx, revenue, opex, profit, loss and free cash flow do not move in the same way. A single score would hide investment phase, utilization and source quality.
How is it counted?
Each value uses: annual run rate * seconds since page load / 31,557,600. The values are diagnostic magnitudes, not real-time payment flows.
| Counter | Annual run rate | Composition | Source quality | Primary sources |
|---|---|---|---|---|
| Infra-CapEx deployment | $563B | Amazon 2026 $200B + Meta 2026 midpoint $135B + Microsoft Q3 CapEx x4 $128B + Oracle FY26 $50B + Anthropic infra $50B. | Guidance | Amazon - Meta - Microsoft - Oracle - Anthropic |
| Revenue anchors | $606.2B | AWS FY + Google Cloud FY + Microsoft AI ARR + Oracle Cloud FY + Nvidia Data Center Q1 x4 + Anthropic ARR. | Mixed | Amazon - Alphabet 10-K - Microsoft - Oracle - Nvidia - Anthropic |
| Operating cash-flow firepower | $651.9B | Amazon FY OCF + Alphabet FY OCF + Microsoft Q3 OCF x4 + Meta Q1 OCF x4 + Oracle FY OCF. | SEC/IR | Amazon - Alphabet - Microsoft - Meta - Oracle |
| Free cash-flow after buildout | $173.6B | Amazon FY FCF + Alphabet FY OCF-CapEx + Microsoft Q3 FCF x4 + Meta Q1 FCF x4 + Oracle FY FCF. | Derived | Amazon AR - Alphabet - Microsoft - Meta PDF - Oracle |
| Public profit signal | $438.2B | Amazon FY net income + Microsoft Q3 net income x4 + Nvidia Q1 net income x4. | IR | Amazon - Microsoft - Nvidia |
| Opex / cost pressure | $151.7B | Meta FY costs & expenses + reported OpenAI 2025 costs. This is a cost-pressure proxy, not AI-only Opex. | Mixed | Meta FY2025 - Fortune |
| Reported private loss signal | $20.9B | Reported OpenAI 2025 operating loss. Kept separate because it is not a public filing. | Leak | Fortune - The Information |
Company Financial Comparator This table compares real disclosed or explicitly reported financial signals by company. Periods differ by company, so the basis column is part of the data, not a footnote. Public-company figures use official IR/SEC sources where available; private-lab rows are clearly marked as lower-auditability.
Filter
Sort
Visible revenue / ARR
$0
Visible CapEx / spend
$0
Visible FCF signal
$0
Visible profit / loss
$0
| Company | Basis | Revenue / ARR | CapEx / spend | FCF signal | Profit / loss | CapEx / revenue | FCF margin | Evidence |
|---|---|---|---|---|---|---|---|---|
| AMZNAmazon / AWSHyperscaler, marketplace, cloud | FY2025 actuals | $716.9B | $128.3Bcash CapEx | $11.2B | $80.0Boperating income | 18% | 2% | IR/AR |
| GOOGLAlphabetSearch, ads, cloud, Gemini | FY2025 actuals | $402.8B | $91.4Btechnical infra CapEx | ~$73.3BOCF - CapEx | $132.2Bnet income | 23% | 18% | SEC |
| MetaAds, social, AI infrastructure | Q1 2026 annualized + 2026 CapEx guide midpoint | $225.2B | $135.0B2026 guide midpoint | $49.5BQ1 FCF x4 | $91.5Boperating income x4 | 60% | 22% | IR |
| MSFTMicrosoftCloud, software, AI run rate | Q3 FY26 annualized | $331.6B | $127.6BCapEx x4 | $63.2BFCF x4 | $127.1Bnet income x4 | 39% | 19% | IR |
| ORCLOracleCloud infra financing case | FY2026 actuals + Q3 CapEx guide | $67.4B | $50.0BFY26 guide cited in Q3 | -$23.7B | $20.6Boperating income | 74% | -35% | IR |
| NVDANvidiaAI chip and systems supplier | Q1 FY27 actuals | $81.6B | $1.8BPPE/intangibles purchases | $48.6B | $58.3Bnet income | 2% | 60% | IR |
| PRIVATEAnthropicModel lab, run-rate disclosure | May 2026 run-rate + infrastructure commitment | $47.0Brun-rate revenue | $50.0Binfra commitment | n/anot public | n/anot public | 106% | n/a | Private |
| PRIVATEOpenAIModel lab, leaked financials | 2025 reported leaks/media | $13.1B | $34.0Bcosts & expenses | n/anot public | -$20.9Boperating loss | 260% | n/a | Leak |
Negative or unavailable private-company free cash flow is excluded from the aggregate FCF tile. Mixed periods are deliberately visible: this comparator is for investment diligence, not a single leaderboard.
Investment Dimensions A defensible investment case should not only ask whether AI revenue is larger than AI spending today. The key is which economic mechanism is being measured.
Accounting
CapEx, OpEx, depreciation, financing costs and free cash flow belong in different lines. Mixing them creates false precision.
Business model
Chip suppliers, cloud providers, model labs, app vendors and enterprise software companies sell different products with different margins.
Revenue attribution
Copilot, AWS, Gemini, ad ranking, search and internal automation create value in different ways. Not all of it appears as direct AI revenue.
Time horizon
Infrastructure is built before demand, utilization and depreciation fully show up. A snapshot can heavily distort free cash flow.
Data quality
Audited filings, official releases, management commentary, leaks and analyst estimates should not be weighted equally.
Dependencies
AI ROI can show up in retention, pricing power, productivity, security, search quality or cloud utilization, not only in chatbot subscriptions.
Investment Evidence Pack These charts show which comparisons are useful for an investment committee and which only provide context. Periods and source quality are made explicit instead of forcing everything into one false "AI profit" number.
AMZNAmazon 2025 FCF
$11.2B
Despite $139.5B operating cash flow; FCF fell because infrastructure investment rose sharply.
Meta 2026 CapEx Guidance
$125-145B
Q1 2026 guidance was revised upward; drivers include component prices and data-center capacity.
MSFTMicrosoft AI Run Rate
$37B
Annual revenue run rate in Q3 FY26; alongside $31.9B quarterly CapEx.
PRIVATEAnthropic Run Rate
$47B
Official May 2026 announcement; private company, no full public P&L view.
Reading rule for the memo
FY figures, quarters, annualized run rates, guidance and commitments are not the same thing. The charts are due-diligence tools: they show magnitude, capital pressure, source quality and payback logic, but not one single "AI is profitable" score.
CapEx intensity
comparableCapEx relative to revenue: Microsoft Q3 FY26 31.9 / 82.9; Amazon 2025 128.3 / 716.9; Alphabet 2025 91.4 / 402.8; Meta 2025 72.22 / 200.97. This shows capital intensity, not AI profit.
Margins as context
not AI-onlyAmazon total = operating margin 2025; AWS = operating margin 2025; Microsoft = Q3 FY26 net margin; Nvidia = Q1 FY27 net margin. High company-level margins do not prove every AI initiative, but they do refute blanket "everyone is losing money" claims.
Cash-flow pressure from infrastructure
FCF visibleAMZNAmazon 2025
GOOGLAlphabet 2025
MSFTMicrosoft Q3
Meta Q1
This chart shows the real tension: operating cash flows are strong, but AI/cloud infrastructure can compress free cash flow sharply in the short term. *Alphabet FCF is simplified here as OCF minus CapEx.
Forward Infrastructure Envelope
mixed periodsNot everything is identical: Amazon/Meta/Oracle are guidance, Microsoft is an annualized Q3 rate, Anthropic is an infrastructure commitment. That difference is exactly what should be visible.
Private labs: growth vs. transparency
low auditabilityPrivate-lab figures are directionally important but weakly auditable. Run rate, annual revenue, operating loss and cash burn are different metrics.
Scaling pressure vs. revenue anchors
ratio lensIntentionally imperfect but useful stress test: which revenue anchor sits next to the infrastructure buildout in rough magnitude? Microsoft = Q3 CapEx annualized / AI run rate; Amazon = 2026 CapEx guidance / AWS 2025 revenue; Alphabet = 2025 CapEx / Google Cloud 2025 revenue; Oracle = FY26 CapEx guidance / FY26 revenue; Meta = 2026 CapEx midpoint / Q1 revenue annualized. This is not a profit comparison, but a capital-pressure indicator.
CapEx payback sensitivity
model$100B infra, 3 years
$55.6B
Annual revenue required just to cover $33.3B depreciation at 60% gross margin.
$100B infra, 4 years
$41.7B
Annual revenue required to cover $25.0B depreciation at 60% gross margin.
$100B infra, 5 years
$33.3B
Annual revenue required to cover $20.0B depreciation at 60% gross margin.
Model calculation, not a company disclosure: required revenue = annual depreciation / gross margin. Excludes power, operations, financing, R&D, sales, underutilized capacity and replacement investment. That is why AI ROI depends heavily on utilization, margins and useful life of the hardware.
ORCLOracle as an infrastructure financing case
official dataOracle shows why "high CapEx = unprofitable" is too blunt: very high backlog/RPO, a negative free-cash-flow phase, external financing and partly customer-financed hardware can exist at the same time. RPO is not profit; timing, margin, utilization and financing structure matter.
Revenue visibility vs. capital pressure
Matrix
top: high revenue visibility
right: high capital pressure
NVDANvidia seller
AMZNAWS / Cloud
MSFTMicrosoft AI
METAMeta AI infra
PRIVATEOpenAI / labs
Internal AI value
The farther right, the higher the risk of misinvestment and depreciation pressure. The higher up, the more visible revenue is today. Private labs can grow fast and still lack financial transparency.
Who earns where in the AI stack?
model logicChips
Clear revenue, visible margins
Cloud
High CapEx, strong contracts
Models
High growth, unclear cash burn
Apps
Possible margin, strongly use-case dependent
Internal value
Often valuable, rarely reported separately
From top to bottom in each column: revenue visibility, capital intensity, data quality. That is why one single profit counter is misleading.
Why CapEx-minus-revenue is too shallow
AccountingYear 0: cash out
Data centers, GPUs, networks and power contracts hit free cash flow immediately.
Asset is created
Part of the spending is capitalized. That is not the same as an immediate loss.
Depreciation
The income statement is affected over several years, not necessarily in full at the start.
Monetization
Revenue emerges through utilization, cloud contracts, product features, ads, search and productivity.
Source quality for financial figures
source hygiene
Audited
10-K, 10-Q, audited financial statements, official earnings releases
Official
Management comments, investor presentations, official guidance
Modeled
Research groups with methodology, modeled uncertainty, reproducible assumptions
Leak
Leaks, paywalled reports, ARR estimates, private investor documents
Weak
Blogs, viral charts, aggregated search results without clean methodology
A good dashboard should make source quality visible instead of making audited filings, official guidance, models and leaks look equivalent.
Three economic scenarios
ROI instead of yes/noBear Case
Underutilization
Capacity is built, but demand or pricing is insufficient. Result: FCF pressure, depreciation, contract risk, margin compression.
Base Case
Absorption
Cloud and enterprise demand absorb capacity over several years. Result: expensive in the short term, viable over the long term.
Bull Case
Platform Leverage
AI increases productivity, pricing power, retention and new workloads. Result: CapEx becomes a strategic moat.
Diligence & Source Quality Good sources exist, but not equally for every company. An investment proposal should make that visible.
| Company / area | What is well supported | What remains uncertain | Source quality | Better source |
|---|---|---|---|---|
| MSFTMicrosoft | Q3 FY26: $82.9B revenue, $37B AI run rate, $31.9B CapEx, $15.8B free cash flow. | AI margin and incremental AI profit are not fully visible. | SEC/IR | Earnings Release - Earnings Call |
| GOOGLAlphabet | 2025: $91.4B CapEx, primarily technical infrastructure; significant increase expected in 2026. | Direct Gemini/AI share and internal search/ads value are not cleanly isolated. | SEC | Alphabet 2025 Form 10-K |
| Meta | Q1 2026: $56.31B revenue, $19.84B CapEx, $12.39B free cash flow; 2026 CapEx guidance of $125-145B. | AI improves ads and ranking, but direct AI product revenue is not separately visible. | IR | Meta Q1 2026 Results - Meta FY2025 Results |
| AMZNAmazon / AWS | 2025: $716.9B revenue, $128.7B AWS revenue, $45.6B AWS operating income, $11.2B free cash flow; about $200B CapEx expected in 2026. | AI share of AWS revenue, margin and future utilization remains partly open. | IR | Amazon Q4/FY2025 Results - Annual Report |
| ORCLOracle | FY2026: $67.4B revenue, $34.0B cloud revenue, $32.0B operating cash flow, -$23.7B free cash flow; Q3 guidance cited $50B FY2026 CapEx. | RPO, prepayments and customer-supplied GPUs reduce capital risk but do not replace visible segment margin; timing and concentration of large AI contracts remain central. | IR | Oracle Q4/FY2026 Results - Oracle Q3 Guidance |
| NVDANvidia | Q1 FY27: $81.6B revenue, $75.2B data-center revenue, $58.3B net income. | Demand cycle, export controls and customer concentration remain risks. | IR | Nvidia Q1 FY2027 Results |
| PRIVATEAnthropic | Run-rate revenue per company statements: $30B in April 2026, $47B in May 2026; $50B infrastructure commitment announced in 2025. | Private company: no full public P&L, no audited segment margin, run rate is not an annual report. | Private | Anthropic Google/Broadcom Update - Anthropic Series H |
| PRIVATEOpenAI | Reported leaks: 2025 about $13.07B revenue, $20.92B operating loss; Q1 2026 per The Information: $5.7B revenue and $3.7B cash burn. | No public audited financials; figures are based on leaks/media reports and could change with IPO filings. | Leak | The Information Q1 2026 - Fortune Leak Summary - Epoch AI Revenue Estimates |
| PRIVATEMistral / Cohere / xAI | Funding rounds, ARR estimates and individual media reports provide some signals. | Profitability, gross margins, compute commitments and cash burn are barely public in a reliable way. | Estimate | Sacra Mistral - Sacra Cohere |
| PRIVATEDeepSeek | R1 training costs were discussed in a peer-reviewed context. | Total costs, pretraining, infrastructure, research, talent and ongoing inference costs are not covered by that figure. | Paper | Nature - arXiv Paper |
Why isaiprofitable.com is not an investment case The critique of overheated AI investment on isaiprofitable.com is useful as a warning signal. As an investment thesis, however, the broad framing is too blunt.
Weak framing
- CapEx is treated like an immediate loss.
- Direct AI revenue is compared with broad AI infrastructure spending.
- Big Tech and startup economics are compressed into one table.
- Audited filings and speculative leaks appear visually equivalent.
- Internal productivity, ad optimization, search and retention are largely missing.
Stronger framing
- Show CapEx, OpEx, depreciation, financing and free cash flow separately.
- Differentiate companies by role: lab, cloud, chip, app, enterprise software.
- Label sources by evidence quality.
- Assess direct revenue and indirect value contributions separately.
- Explicitly show where the data is insufficient for a firm conclusion.
Methodological conclusion: AI profitability cannot be assessed by subtracting estimated AI revenue from broadly aggregated AI spending. A robust view separates cash flow, accounting profit, asset buildout, utilization, pricing power, indirect value creation and source quality.