BiggerPockets Money · Valuation Lab · The AI Complex

Mega-Cap Workbook

Eleven companies now carry more market value than most national stock markets combined. This worksheet reverse-engineers their prices: what do you have to believe about future free cash flow to justify what the market is paying today?

Data snapshot June 10, 2026
Default WACC 10.0%
Every number below is editable

Assumptions

Inputs
%
%
!
What's live, what's not. Market caps refresh hourly when the live feed is connected (see masthead). Financials are a snapshot updated each earnings season. Edit any figure in “The Data” table below - everything recalculates instantly.

The Result

Reverse DCF
Required FCF growth - 2027→2036, after the guided 2026–27 capex
-
Combined market cap-
Share of S&P 500-
Share of global equities-
From price to multiple
Market cap
-
included companies
Net cash
-
cash less debt
=
Enterprise value
-
price of the businesses
÷
FCF (TTM)
-
reported basis
EV / FCF
-
vs ≈20–25× for the median large cap
Where the value sits
Aggregate free cash flow today (selected basis)-
Wall Street consensus FCF, 2026E / 2027E-
Reference: implied growth if compounding started from today’s FCF (ignores the 2026–27 capex dip)-
FCF the price implies by 2036-
…that’s this multiple of today’s FCF (nominal)-
…as a share of today’s entire S&P 500 earnings (in today’s dollars)-
…as a share of all U.S. corporate profits today (in today’s dollars)-
A decade from now these companies must throw off cash at a scale comparable to large slices of the entire current profit pool of corporate America - and 2036 isn’t the finish line. The model assumes the cash keeps growing forever after that.

What You Have to Believe

Belief vs. price
-5%10%20%30%45%
Your growth belief-
Fair value of the complex at your belief-
Market price today-
Implied upside / (downside)-
Market-implied growth (break-even belief)-
Fair value = street 2026–27 FCF, then your growth rate (fading to terminal), discounted at your WACC. If your estimate sits below the red marker, the model’s fair value is below today’s price; above it, the reverse.

The Required Cash Flow Path

Built company by company
How to read this chart
Each bar stacks every company’s operating cash flow - the cash the businesses generate before investment, color-coded by company.
The rust slice at the top of each bar is capex - cash spent on data centers and chips before anything reaches shareholders.
The black tick is what’s left over: net free cash flow. Notice that through 2027 it falls even as the bars grow - the capex bill is growing faster than the cash.
The rust line (right-hand axis) is combined revenue - the top line all of this cash has to come from.
The chart changes meaning at the dashed line. Through 2027 the bars are actuals and Wall Street consensus estimates. From 2028 on, the bars show the cash flow today’s prices require - a reverse DCF run for every company - with capex assumed to fade back toward pre-AI levels. The gray cap is the market value of companies whose 2027 consensus free cash flow is zero or negative, so no per-company path can be computed; it is carried as a residual.
Hover over (or tap) any year in the chart for the full dollar breakdown.
Two-year capex bill, 2026–27 - vs. how much FCF is expected to change over the same span-
Market value sitting in companies expected to have zero or negative FCF in 2027-
Required FCF growth per year, from the 2027 consensus base to 2036-
Single-year FCF the price requires by 2030-
Operating cash flow required by 2036 (before assumed steady-state capex)-
Required revenue by 2036 at the steady-state margin-
This chart and the headline in The Result show the same consensus-anchored requirement; the percentages match. The smaller Reference row in The Result instead starts growth from today’s cash flow and skips the 2026–27 capex years; its figure is a growth-phase rate rather than an average through the fade, so the two percentages are not directly comparable. Two assumptions here work in the direction of lower required growth: free-cash-flow margins expand smoothly to your steady-state setting, and capex intensity fades from today’s ≈33% of revenue back to the pre-AI norm of ≈11%. With both in place, the operating bars still grow - from the trailing-twelve-month bar.

Scale Checks

The requirement, in context

Under these assumptions, the market is valuing this group as if it will produce - of free cash flow in 2036 - - in today’s dollars. That is - of the - earned by every publicly listed company on Earth this year, and - of the entire world’s projected listed-company profit pool in 2036.

The world’s top line
The required 2036 revenue of - would equal - of the revenue booked by every listed company on the planet this year (in today’s dollars), and - of the projected 2036 global top line. For scale: the Fortune Global 500 - the five hundred biggest companies on Earth - collectively books about $41.7T of revenue. These eleven would need to capture revenue equal to roughly a third of that entire list.
The base-rate problem
The path requires - compound FCF growth for nine years from a base of -. The most celebrated cash-flow decade in corporate history - Apple’s iPhone run, roughly 2010–2020 - compounded free cash flow at about 16% per year, starting from a base of roughly $17B. This ask is faster growth from a starting point - that size. Decade-long growth persistence at mega-cap scale is historically rare; the price assumes it for an entire group simultaneously.
If growth never comes: the yield floor
Strip away all growth and these businesses currently hand their owners a free-cash-flow yield of - on enterprise value (- on street 2027 estimates). Against the - return you asked for in Assumptions, growth must supply everything else. A 10-year Treasury pays more with zero execution risk - the premium above that depends entirely on future growth.
One point of rate
The requirement is highly sensitive to the discount rate. At a discount rate one point below your setting, the required growth eases to -; one point above, it steepens to -. This is why the entire complex trades like a long-duration bond around every Fed meeting - the cash is far away, so the discount rate does the heavy lifting.
Each module recomputes live from the chart above. World-pool figures are derived, editable estimates - the global profit and revenue pools are not officially tallied anywhere; the derivations (global market cap divided by global P/E and price-to-sales) and the Fortune Global 500 reference points are the honest scaffolding.

Company X-Rays

One reverse DCF each

The Bull Case, Company by Company

Steelman first

Most valuation critiques attack individual companies. This section does the opposite: it presents the strongest honest case for each name - no strawmen - alongside what has to be true, what slice of the future economy each story implies, and the terminal free cash flow each price requires.

The Aggregation Question

Sum of the parts

Taken one at a time, most of these stories hold together on their own. NVIDIA could work. So could Microsoft, Alphabet, Amazon, TSMC, Meta. The hard part is finding a future where all of them at once earn the profits implied by their current valuations.

Individual bull cases
Evaluated one at a time, most narratives pass.
Assumed simultaneously
-
The market prices every story succeeding at once - a joint outcome that single-stock analysis does not test.
Required aggregate 2036 FCF
-
A future where NVIDIA wins infrastructure and Microsoft wins software and Alphabet wins search and Amazon wins cloud and Meta wins advertising and TSMC wins manufacturing is the future today’s prices assume. The Planet Test below measures that requirement against the whole economy.

Analysts grade each story on its own. Almost no one asks whether the collection of stories can coexist - these companies compete for the same enterprise budgets, the same ad dollars, the same compute contracts, and each other’s margins.

The Claims Ledger

Bottom-up, market by market

Each bull case above claims a slice of a real end market. This ledger sizes each market’s bull-case 2036 revenue, converts it into a maximum free-cash-flow pool at a stated margin, and compares that pool to the terminal cash flow the claimants’ prices require, allocated across each company’s business mix. The dashed tick on each bar marks the full pool; rust is the portion of the claims the pool cannot cover. All figures are nominal 2036 dollars, and every input is editable.

Across all seven markets
Terminal FCF the prices require (sum of company solves)-
FCF pools at 100% capture of every market-
Cross-market netting: cloud–silicon contradiction · transfers & overlap -
Upper bound: full capture, net of the adjustments-
Central case: reachable shares, net of the adjustments-
Central case vs requirement-
Upper bound vs requirement-
What is fixed and what is editable. Business-mix allocations per company (the chips on each market) are fixed in this version and stated so they can be argued with; TAMs, margins, reachable shares, and the two netting figures are editable judgments. The semiconductor pool includes foundry profit: TSMC’s revenue sits upstream of NVIDIA’s and Broadcom’s, so it is additive in cash but not in end-market share. The AI-applications line is the most speculative input; that TAM is itself the bull case. The cloud–silicon netting exists because one market’s revenue is the other’s capex: both pools cannot be maximal at once.

The Circular Economy

Who pays whom

A lot of the revenue inside this complex is a cost somewhere else in the complex. NVIDIA’s revenue is hyperscaler capex; hyperscaler capex becomes NVIDIA revenue. Cloud revenue from the AI labs is partially funded by equity investments from those same hyperscalers. One company’s success shows up as another company’s expense. The system can create real value - it just can’t create infinite value by recycling spending inside the group. The revenue is real. What matters is whether the ultimate source of cash sits outside the ecosystem.

THE BUYERS MSFT GOOGL AMZN META ORCL THE SELLERS NVDA AVGO TSM THE AI LABS OpenAI · Anthropic · xAI capex ≈$879B ’26E → chip & fab revenue equity investments fund the labs labs spend it back as cloud revenue labs buy chips too funded by ads, software, retail - the outside economy
The CapEx Dial
Adjust any buyer’s capex and see the effect on the sellers’ revenue and on the buyers’ own free cash flow.
Scenarios:
The buyers’ capex budgets
Outside buyers step in OPTIONALshare of cancelled chip orders replaced by parties not on this page - sovereign AI programs, enterprises, externally-funded clouds0%
The sellers’ revenue from this group
Estimates, not disclosures. The capex-to-vendor splits are assembled from customer-concentration disclosures and supply-chain reporting. TSMC feels NVIDIA and Broadcom cuts at one remove - fewer chip orders mean fewer wafer orders. Directionally right, precisely unknowable. Dashed tick = current plan.

The sellers of compute

NVIDIA, TSMC, and Broadcom book revenue that is, to a striking degree, the capital expenditure of the other names on this page. NVIDIA’s data-center revenue is hyperscaler capex; TSMC fabricates the chips NVIDIA, Apple, and Broadcom design; Broadcom’s custom accelerators are built for Google and Meta. Their growth is a derivative of the buyers’ willingness to keep spending.

The buyers - and the loop

Microsoft, Alphabet, Amazon, Meta, and Oracle fund that capex from advertising, software, and retail profits. The loop closes a second time through the AI labs: hyperscaler cloud revenue from OpenAI and Anthropic is partly funded by equity investments from those same hyperscalers and NVIDIA. Money that leaves as capex or investment and returns as revenue makes the group’s combined top line larger than its sales to the outside economy.

Estimates, not disclosures. Companies rarely break out customer-level revenue. The percentages above are assembled from 10-K customer-concentration disclosures, reported capex budgets, and supply-chain reporting - treat them as directional, and edit them in the data table if you have a better view. And circularity doesn’t mean the revenue is fake - it’s real. The risk is correlation: if the buyers cut capex, the sellers’ revenue, earnings, and the multiples on both fall together.

The Planet Test

The aggregate check

The point of this section is scale, not prediction. The figures state the requirement even after allowing for substantial growth. Every figure below is the 2036 requirement implied by current prices, at your assumptions, measured against the whole planet.

Required 2036 revenue
-
Required 2036 free cash flow
-
Share of global corporate profits
-
Share of global corporate revenue
-
Revenue per human
-
per year, every person alive
Profit per human
-
free cash flow extracted per person
Revenue per affluent consumer
-
per year
Profit per affluent consumer
-
per year
“Can NVIDIA do this?” Any single company might.
The aggregate question: “Can all of them extract this much revenue and profit from the same future economy?”
Individual plausibility does not aggregate automatically: each story can be reasonable on its own while the combined requirement is not. The strongest companies in the world can all be excellent businesses and still be disappointing investments if too much collective success has already been priced in. This tool won’t tell you which company falls short. It shows how much success the whole system requires. The aggregate question: can all of these stories be true at the same time?

The Data

Edit anything · $ billions
CompanyMkt capRevenueNet incOp. CFCapExSBCNet cashFwd NIFCF ’26EFCF ’27ECapEx ’26ECapEx ’27ENorm capex %Circular %
All figures in $ billions, trailing twelve months unless noted. Fwd NI and the FCF / CapEx ’26E–’27E columns are Wall Street consensus-style estimates by calendar year (these feed the chart). Norm capex % is the steady-state capex-to-revenue ratio used by the “Normalized” FCF basis - roughly each company’s pre-AI-boom intensity. Circular % is the estimated share of revenue paid by other companies in this set (or by AI labs they fund).

Show Your Work

Methodology
1 · The reverse DCF solve for g, not for value

A normal DCF asks “given growth, what’s it worth?” This tool asks the inverse: given the price, what growth is required? The canonical model is consensus-anchored: 2026 and 2027 free cash flow are pinned to Wall Street estimates (which include the guided 2026–27 capex), and the solver finds the growth the years after must deliver. The classic smooth-path version - growth starting immediately from today’s TTM - is shown in The Result as a reference line; it answers a cleaner textbook question but contradicts known guidance. For each company we take free cash flow today, grow it at rate g for the high-growth runway, fade linearly to the terminal rate, capitalize the terminal year at WACC − gT, discount everything at WACC - and solve (by bisection) for the g that makes the present value equal today’s price.

EV  =  Σt=1..N FCF₀(1+g)… / (1+WACC)t  +  [ FCF₁₀ × (1+gT) / (WACC − gT) ] / (1+WACC)N   →  solve for g

Companies with negative free cash flow (Oracle and SpaceX on the reported basis) have no solvable implied growth rate - you can’t compound from a negative base, so their X-ray cards read n/m. Their bull-case cards still state a required terminal cash flow, solved directly from the price: 2026–27 stay pinned to consensus, then a straight-line ramp to the end year.

2 · The capex normalization the entire bull/bear debate in one toggle

Reported FCF for the hyperscalers is currently reduced by the AI build-out - capex budgets have roughly doubled or tripled from pre-2023 intensity. Bulls argue this is growth capex that will normalize; bears argue it is the new cost of staying in business. The Normalized basis recomputes FCF as operating cash flow minus steady-state capex (each company’s editable “Norm capex %” of revenue), which is the most favorable commonly argued basis. The gap between the two bases shows how much of the requirement capex normalization alone closes.

3 · Market cap vs. enterprise value net cash matters at this scale

A DCF values the operating business, so the cleaner comparison is enterprise value (market cap minus net cash). Apple, Alphabet, and TSMC hold meaningful net cash; Broadcom and Oracle carry meaningful net debt. The toggle lets you see both; multiples on the cards show P/E against market cap and FCF yields against EV.

4 · Index & profit-pool anchors editable denominators

Share-of-index figures compare full company market caps to the index total; the official S&P weights are float-adjusted, so published weights run slightly different. TSMC (Taiwan-listed) and SpaceX are not S&P 500 members and are excluded from the “share of S&P 500” numerator, but included in the global figure.

Sources snapshot June 10, 2026
Educational tool, not investment advice. This worksheet exists to make valuation assumptions explicit, not to tell you what to buy or sell. Figures are point-in-time estimates assembled from public sources and simplified for clarity: TTM periods differ across fiscal calendars, forward estimates are approximations of Wall Street consensus, circularity percentages are directional estimates, and the DCF handles stock-based compensation as a discount-rate adjustment rather than explicit share counts, and ignores buybacks, taxes on repatriated cash, and option value. Net income is TTM as reported, with two flags: GOOGL excludes a large Q1’26 unrealized gain on equity securities, and META’s figure includes a one-time 2025 tax charge and a 2026 transitional tax benefit that partly offset. Markets reprice daily; the snapshot date is June 10, 2026. Do your own work, and consider consulting a fee-only fiduciary advisor before acting on anything here. BiggerPockets Money is a media company, not a registered investment advisor.
BiggerPockets Money · Valuation Lab