Brandon Leon

Hyperscaler capital spending · October 21–29, 2026 · Macro · No rating

A Mixed Answer From the Buyers

Three of the four buyers committed to spending more in 2027 and the fourth would not say; what none of them settled is who pays for it.

Posted November 3, 2026Author Brandon LeonData as of company releases and calls through October 29; index closes through Friday, October 30

The verdict

No buyer stepped back. Meta, about a fifth of the four plans at the midpoint of its 2026 range, gave no 2027 direction at all: “we expect to continue investing aggressively,” with the budget still being built. Our own wording from October 19 requires restraint here: one company guiding lower needs its size and reasons examined before it means anything about the cycle.[15] That is neither a build pause nor a demand statement; it is a budget not yet set, and the disclosures do not allow a determination.

The other three pointed higher, so the demand case is intact on the majority of the spending and unproven on the rest. That is a narrower conclusion than either of the clean outcomes would have allowed, and it leaves the financing question exactly where October 19 left it.

1. What the buyers had already said

Everyone was waiting for the 2027 numbers. The buyers had already answered the question underneath them: each of the four raised its plan for the current year at least once in 2026, and that record is the baseline the new guides are judged against.

Alphabet opened the year at $175 to $185 billion and had reached $195 to $205 billion by July, having told shareholders in June that 2027 would “significantly increase” from there.[1][2][3] Amazon said about $200 billion in February and about $220 billion in July, with the chief executive attributing the increase to the cost of memory and adding that demand would exceed capacity “in 2027, too.”[11][13] Meta started at $115 to $135 billion, raised the range in April on component prices and has since narrowed it upward twice, to $135 to $145 billion last week.[7][8][9][10] Microsoft, on a June fiscal year, said in July that fiscal 2027 spending “will grow year-over-year” and put the figure at about $175 billion after moving new data-center leases off the capital-spending line.[6]

Chart 1: The Plans Kept Rising

Every buyer raised its plan at least once this year; Meta did it three times.

Capital-spending plans for the current year, midpoint of each guide as given, 2026 ($bn) 100 120 140 160 180 200 220 240 Alphabet 2026 Amazon 2026 Meta 2026 Microsoft FY27 175-185 180-190 195-205 195-205 unchanged (Oct 28) ~200 ~220 115-135 125-145 130-145 135-145 FY27 ~175 prints Oct 28-29 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Alphabet and Amazon: cash purchases of property and equipment. Meta: including finance-lease principal. Microsoft: fiscal year to June 2027, after moving new data-center leases off the capital-spending line.

Sources: company releases and calls, each quoted in the text and listed in the sources.[1]–[13] Plotted at the midpoint of each range on the date it was given.

Add the four current-year plans and the figure is about $735 billion, against roughly $410 billion the same companies spent in their last completed year.[1][5][7][11] The definitions differ, and Microsoft’s fiscal year straddles two calendar years, so treat the sum as a measure of scale and nothing more precise. Even as scale, it says the buyers were committed to nearly twice last year’s total before anyone asked them about 2027.

Chart 2: The Scale of the Commitment

The current-year plans sum to roughly $735 billion, against about $411 billion spent last year.

Last year spent, this year planned, next year guided ($bn) 0 40 80 120 160 200 240 280 last completed year, actual current plan, midpoint next year, as guided this week 91 200 words Alphabet 2025 / 2026 plan 132 220 words Amazon 2025 / 2026 plan 72 140 words Meta 2025 / 2026 plan 116 175 words Microsoft FY26 (to Jun 2026) / FY27 plan Four plans summed: about $735 billion for the current year against about $411 billion spent in the last completed year, on each company's own definition.

Sources: cash flow statements in each company's results release for the actuals; guides as quoted in the text. Definitions differ by company and the sum mixes Microsoft's fiscal years with calendar years; read it as scale only.

So the live question was narrower than the headlines suggested: would the buyers commit to more in 2027 than the roughly $735 billion already planned, and how would they pay for it. The first half is the demand case the October 19 note said this week would settle. The second half is the financing question it said this week could not.

2. How the market has been grading the raises

Which of the two halves the market cares about is also already on the record. The July reports are the best guide, because all four companies raised or held their plans within nine days and the reactions split cleanly. Microsoft left its calendar-2026 plan unchanged, backed it with Azure growth of 43% and a contracted backlog, and rose 15.5% the next session. Amazon raised to about $220 billion with AWS accelerating to 37% growth and rose 15.3%. Alphabet raised to $195 to $205 billion in the same quarter it reported its first negative free cash flow and after raising $49.6 billion of equity and $56.2 billion of debt in six months, and fell 7.1%. Meta raised its range alongside free cash flow of under $1 billion and fell 8.0%.[4][5][12][14]

October has so far delivered only the sold half of that pattern. Meta narrowed its range upward again on October 21, reported negative free cash flow for the quarter, and fell 7.0% the next day. Lam Research, which sells the equipment this spending buys, guided its December quarter to $8.6 billion against the $8.1 billion we had registered as the proof level, and fell 11.8%.[10][14][17] Good news on the spending line has been sold for two weeks. The market is grading two other things: whether the spending is paid for from operations, and whether the revenue it is meant to produce is showing up.

Chart 3: How the Market Has Graded the Raises

Since July the market has paid the plans funded from operations and fined the ones that arrived with negative free cash flow.

Next-session move after each capital-spending message, close to close (%) -15 -10 -5 0 +5 +10 +15 +20 Alphabet, Jul 22: raised to $195-205B -7.1% Meta, Jul 29: raised to $130-145B -8.0% Microsoft, Jul 29: 2026 plan unchanged +15.5% Amazon, Jul 30: raised to ~$220B +15.3% Meta, Oct 21: raised to $135-145B -7.0% Lam, Oct 21: guided above its gate -11.8% Microsoft, Oct 28 -0.2% Alphabet, Oct 28 -12.6% Amazon, Oct 29 -1.4% Green: the market paid the message. Terracotta: it fined it. Dashed: this week's reports, filled after the next close.

Source: Yahoo Finance daily closes;[14] the move is from the close before the release to the next close. Messages as quoted in the text.

That is why each of this week’s reports carries a proof test we set before it, each at a 55% prior: Azure growth above the 45% Microsoft guided, Google Cloud backlog above $560 billion, and AWS growth above the 37% of the second quarter.[16] A raised 2027 plan that arrives with its proof test passed is the July Microsoft and Amazon pattern. A raised plan with the proof missed, or with new external financing, is the Alphabet and Meta pattern, and on the evidence of the last three months the shares will say so by the next close.

One accounting point decides how to read Microsoft’s number. The ~$175 billion for fiscal 2027 is after a change that moves new data-center leases from finance to operating treatment; finance leases count as capital spending and operating leases do not, and the company said the shift lowered the figure it would otherwise have given.[6] A reported number that comes in under what the Street carried may therefore be an accounting effect, with the build unchanged. The operating-lease commitments in the quarterly filing are where the difference shows, and we read that line before calling anything a cut.

3. The guides

In the October 19 outlook we wrote: “if the combined guidance from the large cloud buyers points higher than 2026, the demand case survives the credit scare and we treat the drawdown as a repricing of leverage, not of orders. If the guidance points flat or lower across more than one major buyer, we treat the cycle as turning and say so; a single company guiding lower needs its size and reasons examined before it means anything about the cycle.”[15] The table scores each company on that sentence.

CompanyLast year, actualThis year, as planned2027Funding lineProof test
Microsoft
FY to June; Oct 28		FY26 $115.9B additions to property and equipment[5]FY27 about $175B after the lease shift; “will grow year-over-year”[6]No calendar-2027 figure, and the July ~$175B was not restated. “In Q2 we expect CapEx spend will be over $50 billion. Therefore for the calendar year, spend approximately $170 billion. Due to the 3P capacity delays noted earlier, in the second half of the fiscal year we currently expect CapEx spend to accelerate.”[18] Direction: higher into the first half of 2027.FY26 free cash flow about $67B; no external raise. Q1 FY27 free cash flow $18.6B on $45.2B of capex including $7.4B of finance leases; “we will access additional sources of funding when we believe it is attractive to do so”; operating-lease commitments are in the 10-Q.[18]Azure above 45% cc: passed, 48% (48% in constant currency); Q2 guided to 49–50%[18]
Alphabet
calendar; Oct 28		2025 $91.4B purchases[1]$195–205B, raised twice[3]June: “significantly increase”[2] Oct 28: “We continue to expect our CapEx to increase significantly in 2027 and will provide further details on our Q4 call”; 2026 held at $195–205B.[17]Q2 free cash flow −$5.9B; $49.6B equity and $56.2B debt raised in H1; $811B of purchase commitments[4] Q3 free cash flow −$9.1B on $51.4B of capex (a TPU-system inventory build and a $5.2B EC fine also weighed); purchase commitments $924B, from $811B; long-term debt $125.1B; the quarter’s issuance is in the 10-Q.[17]Cloud backlog above $560B: passed, $561B at September 30, one billion over the line[17]
Meta
calendar; reported Oct 21		2025 $72.2B incl. finance-lease principal[7]$135–145B, raised three times[10]No figure. “We expect to continue investing aggressively to support both our core business and emerging AI initiatives.”[10]Q3 free cash flow −$2.1B on $34.9B of capex[10]2026 midpoint above $137.5B: passed at $140B; shares −7.0%
Amazon
calendar; Oct 29		2025 $131.8B purchases[11]About $220B, from about $200B[13]July: demand exceeds capacity “in 2027, too”[13] Oct 29: “expect to spend meaningfully more in 2027 with a clear line of sight to attractive financial returns”; 2026 held at about $220B, the majority for AWS.[20]Trailing free cash flow −$42.4B at September 30, from −$7.6B at June 30; $75.9B of cash capex in the quarter; “We can fund this plan.”[12][20]AWS growth above 37%: passed, 39.3%, a sixth straight quarter of acceleration; shares −1.4% close to close[20]
ReadThree of the four pointed higher for 2027: Alphabet in words (“increase significantly”), Amazon in words (“meaningfully more”), and Microsoft by a second-half acceleration with no 2027 figure. Meta gave words without a direction. None pointed flat or lower. By the October 19 sentence that is branch C, a majority higher and one unproven; it becomes A only if Meta’s “investing aggressively” is read as higher, and a budget not yet set is not a direction.

Meta is the buyer that did not point higher, and it did not point lower either. Its October 21 release gave no 2027 figure and no direction, only “we expect to continue investing aggressively to support both our core business and emerging AI initiatives,” with the 2027 budget still in development. At the midpoint of its 2026 plan it is about 19% of the four. Nothing in the disclosures allows a cycle reading from it either way.

The demand case stands on the buyers who pointed higher and is unproven on the one who did not. The financing question is unchanged by either. Both proof tests at the buyers who pointed higher passed: Azure grew 48% against the 45% line and Google Cloud’s backlog reached $561 billion against the $560 billion line, by one billion.

What it means: Alphabet said it will spend significantly more next year and Microsoft said its spending accelerates into the first half of 2027; Meta would only say it keeps investing aggressively. The same calls showed the cost: Alphabet’s free cash flow was negative $9 billion for the quarter and Microsoft said it will tap outside funding when it is attractive.

4. What it does to the covered names

Four published positions sit downstream of this week’s numbers, and each carries a test written before the reports.

NamePublished stanceThe test it carriesRead after the guides
MicronHold, $1,090 (October 19)The thesis breaks on the first cut to fiscal 2028 consensus earnings; there is no price level in the testLess likely: three-quarters of Microsoft’s quarter went to short-lived assets, “primarily CPUs and GPUs, including higher memory and storage costs,” with spending accelerating in the second half, and Alphabet’s plan rises again in 2027. The memory line in the plans points up.
ASMLAccumulate, about $2,075 (October 22)Lithography orders follow this line with a lag; one of the four tests in the October 22 noteTwo buyers pointing higher is the input that note’s order test wants; graded there once Amazon has reported.
NvidiaHold, about $250 (August 26); re-underwrite due November 11Two or more buyers cutting 2027 plans is a preregistered kill; the quarter lands around November 17, date unconfirmedDid not fire: no buyer cut a 2027 plan; two pointed higher and one gave no direction.
DellHold, $340 (June 9); re-underwrite stagedAI-server orders are this spending one step down the chain; the funding of those orders was the July note’s subjectSupportive one step down: Microsoft’s second-half acceleration and Alphabet’s build of TPU-system inventory are server orders in waiting, and nothing said about their funding changed.

The index these names trade with has its own rule, registered on October 23 after the fourth straight down week: no bottom call until the market buys a capital-spending beat, meaning a weekly close back above the October 16 close after a beat the index finishes up on.[16] Meta’s and Lam’s beats were sold: Meta closed Wednesday at $612.49, 12% below its pre-report close, while Lam has recovered to $253.82 from $230.15. This week’s were sold too. Alphabet beat on revenue and cloud and closed down 12.6% the next day; Microsoft beat on Azure and closed flat, then fell 1.6% Friday; Amazon’s seven percent after-hours drop faded to −1.4% by Friday’s close. SMH went in at 567.67 and finished the week at 522.49, a fifth straight down week, back below the 539.79 October 16 close the rule needs.[14][16] Five beats now, none bought. The rule still says no bottom call.

Chart 4: The Semiconductor Index, Waiting for a Beat to Be Bought

No bottom call until the market buys a capital-spending beat.

VanEck Semiconductor ETF (SMH), weekly closes, Jul 10 - Nov 6 ($) 480 500 520 540 560 580 600 620 640 660 Oct 16 weekly close 539.79 Oct 23 weekly close 527.20 week in progress 510.58 (Nov 2) Oct 2 630.60 Jul 10 Jul 31 Aug 21 Sep 11 Oct 2 Oct 23 Nov 6 Last full week 522.49, -17.1% below the Oct 2 weekly high. The test: a weekly close above the Oct 16 close, after a capex beat the index closes up on.

Source: Yahoo Finance daily closes, taken as the last close of each week.[14] The rule was registered on October 23, before this week's reports.[16]

The book. One thing changed, and it was mechanical. No published level keys off a capital-spending guide directly; the levels that can move are the ones in the names that sell to these buyers, and those were set before the week. Bloom Energy remains restricted while a target review is pending. An order to buy Alphabet at $319.78, resting since August 18 under the resting-order exception, filled on October 26, and the position was sold before this note was published; the author holds no Alphabet at publication. This note rates no security and opens no window of its own.

5. What would make us reconsider

  1. The proof tests, graded five sessions after each report. Azure above 45%, Cloud backlog above $560 billion, AWS above 37%, each registered at a 55% prior. Azure 48%, passed; Google Cloud backlog $561 billion, passed by one billion; AWS 39.3%, passed, the fastest in nearly five years. All three passed. The five-session grades follow on November 4 and 5.
  2. A new test from what the guides revealed. The aggregate rose, so the next question is funding. The read changes from “demand intact, financing open” to “demand bought with leverage” if Alphabet’s fourth-quarter free cash flow is negative again alongside new issuance in the 10-Q, or if Microsoft’s “additional sources of funding” becomes a raise of $20 billion or more before its fiscal second quarter closes. If neither happens, the demand case stands as written.
  3. CoreWeave credit, standing. A five-year default-swap spread back below 800 basis points says the credit scare is healing. At or above 1,140, the October high, says it is deepening.
  4. The ten-year, standing. A Friday close at or above 5.31%, October’s cycle high, keeps the pressure on long-duration equity; sustained below it eases one source of it.
  5. The semiconductor index. A weekly close above the October 16 close, after a capital-spending beat the index finishes up on, ends the no-bottom-call rule. Until then the rule stands.

Housekeeping

Scope. Guidance figures are taken from each company’s own release or call and quoted as given; where a figure comes from a call and not from the release, the source says so. Prior-year actuals are from the cash flow statements in the results releases. Share moves are close to close from Yahoo Finance. The four companies define capital spending differently and the text keeps each on its own definition.

Amazon. Amazon reported Thursday, October 29, after the close, between this note’s writing and its publication, so its answer is folded in above. 2026 cash capital spending stays at approximately $220 billion, the majority for AWS, and 2027 got words rather than a figure: “expect to spend meaningfully more in 2027 with a clear line of sight to attractive financial returns.”[20] AWS grew 39.3%, a sixth straight quarter of acceleration, which passes the third proof test. The cost sits where this note keeps looking: trailing free cash flow was −$42.4 billion at September 30, against −$7.6 billion at June 30, on $75.9 billion of cash capex in the quarter, with trailing purchases of property and equipment at $215.2 billion against the roughly $220 billion plan. Asked about it, the chief executive said “we can fund this plan.”[20] The market marked the shares down seven percent after hours and settled at −1.4% by Friday’s close. The branch does not move: three of four higher with Meta unproven is still a majority higher. On the latest readings, free cash flow is negative at three of the four buyers, Microsoft the exception.

Trading window. Our policy restricts trading in a covered name from ten days before a publication to five days after it. This note rates no security. Bloom Energy’s October 23 window ran through October 28, and a pending target review keeps the name restricted until that review publishes; Vistra is inside the window ahead of its November 6 report note. The Alphabet order that filled on October 26 predated every window and was logged under the resting-order exception; the sale of that position is logged in our trade log and reported in the Alphabet note. Protective stops and kills at previously approved levels remain available at any time; the Coherent and ATI kill levels were breached on October 22 and belong to those names’ own records; nothing in this note changes a level.

Pre-registration. The direction call in section 3 was logged on October 26 with a 75% prior that every buyer who spoke would point higher for 2027. The three proof tests were registered on October 20 and 21 with their thresholds and 55% priors, before any of the reports. All four are internal records; the thresholds and priors are quoted here exactly as written.

Sources

  1. Alphabet, fourth-quarter and fiscal-year 2025 results, February 4, 2026: purchases of property and equipment $91,447M (2025) and $52,535M (2024); “our 2026 CapEx investments are anticipated to be in the range of $175 to $185 billion.” SEC EDGAR
  2. Alphabet, Form 8-K of June 1, 2026 (equity offerings): “its 2026 capital expenditures are expected to be $180-$190 billion” and “it expects 2027 capital expenditures to significantly increase compared to 2026.” SEC EDGAR
  3. Alphabet, second-quarter 2026 results, July 22, 2026: purchases of property and equipment $44,924M; free cash flow −$5,855M; Google Cloud revenue $24,768M, up 82%. The $195–205 billion range was stated on the call. SEC EDGAR; call via abc.xyz/investor
  4. Alphabet, Form 10-Q for the quarter ended June 30, 2026: cash from financing, net equity and debt issuance, purchase commitments of $811.0 billion. SEC EDGAR
  5. Microsoft, fiscal 2026 fourth-quarter results, July 29, 2026: additions to property and equipment $115,948M (FY26) and $64,551M (FY25); cash from operations $182,935M; Azure and other cloud services growth 43%. SEC EDGAR
  6. Microsoft, fiscal 2026 fourth-quarter earnings call, July 29, 2026, Amy Hood: “we expect FY ’27 capital expenditures will grow year-over-year given demand signals across our portfolio. However, the shift from finance to operating leases adjust our expectation to approximately $175 billion”; useful lives extended from 15 to 25 years; first-quarter capital spending “over $50 billion”; calendar-2026 expectations “unchanged” outside the useful-life effect. Transcript via The Motley Fool, a secondary source.
  7. Meta, fourth-quarter and full-year 2025 results, January 28, 2026: capital expenditures including finance-lease principal $72.22B (2025); purchases $37,256M plus finance-lease principal $1,969M (2024); 2026 outlook “in the range of $115-135 billion.” SEC EDGAR
  8. Meta, first-quarter 2026 results, April 29, 2026: 2026 outlook raised to $125–145 billion on component prices. investor.atmeta.com
  9. Meta, second-quarter 2026 results, July 29, 2026: “in the range of $130-145 billion, narrowed from our prior outlook of $125-145 billion”; capex $31.08B; free cash flow $784M. SEC EDGAR
  10. Meta, third-quarter 2026 results, October 21, 2026: “in the range of $135-145 billion, narrowed from our prior outlook of $130-145 billion”; capex $34.91B; free cash flow −$2.12B; cash from operations $32.79B; revenue $64.60B, up 26%; “we expect to continue investing aggressively to support both our core business and emerging AI initiatives.” SEC EDGAR
  11. Amazon, fourth-quarter 2025 results, February 5, 2026: purchases of property and equipment $131,819M (2025) and $82,999M (2024); trailing free cash flow $11,194M; “we expect to invest about $200 billion in capital expenditures across Amazon in 2026.” SEC EDGAR
  12. Amazon, second-quarter 2026 results, July 30, 2026: purchases $54,208M in the quarter and $173,028M trailing; trailing free cash flow −$7,604M; AWS net sales $42,232M, up 37%; third-quarter guidance. SEC EDGAR
  13. Amazon, second-quarter 2026 earnings call, July 30, 2026, Andy Jassy: “We now believe we will spend approximately $220 billion in cash CapEx in 2026”; “The higher cost of memory pushing this number up from our prior estimate of about $200 billion”; demand to exceed capacity “in 2027, too”; “You’ve seen us issue debt this year.” Transcript via Webull, a secondary source.
  14. Yahoo Finance daily closes for META, MSFT, GOOGL, AMZN, LRCX and SMH. Price history
  15. Our October 19 outlook: “Insurance on a Company You Don’t Own”, section 4, where the capital-spending test and the single-company caveat were set.
  16. Our internal print-gate register and ledger: the three proof tests (registered October 20 and 21), the direction call (October 26) and the index rule (October 23). Internal records without public URLs; thresholds and priors are quoted exactly as written.
  17. Lam Research, results for the quarter ended September 2026, October 21, 2026: September-quarter revenue $8.22 billion; December-quarter revenue outlook $8.60 billion plus or minus $400 million. investor.lamresearch.com
  18. Alphabet, third-quarter 2026 results, October 28, 2026: revenues $127,577M; Google Cloud revenue $30,709M, up 103%; capital expenditures $51,355M; free cash flow −$9,065M. SEC EDGAR. Call, October 28, 2026, Anat Ashkenazi: “Google Cloud’s backlog was $561 billion in the third quarter”; “our full-year 2026 CapEx guidance range of $195 to $205 billion remains unchanged. We continue to expect our CapEx to increase significantly in 2027”; future purchase commitments $924 billion. Transcript via Benzinga, a secondary source.
  19. Microsoft, fiscal 2027 first-quarter results, October 28, 2026: revenue $92.3 billion; Azure and other cloud services revenue growth 48% (48% in constant currency); Microsoft Cloud revenue $63.9 billion. microsoft.com/investor. Call, October 28, 2026, Amy Hood: “In Q2 we expect CapEx spend will be over $50 billion. Therefore for the calendar year, spend approximately $170 billion”; “in the second half of the fiscal year we currently expect CapEx spend to accelerate”; Azure guided to 49% to 50% in constant currency; “we will access additional sources of funding when we believe it is attractive to do so.” Transcript via Benzinga, a secondary source.
  20. Amazon, third-quarter 2026 results, October 29, 2026: AWS revenue $46.0B, up 39.3%; cash capex $75.9B in the quarter; 2026 about $220 billion and 2027 “meaningfully more” (Andy Jassy); trailing twelve-month free cash flow −$42.4B per the release. ir.aboutamazon.com; transcript via Investing.com, a secondary source; CNBC.

Disclosures. The author holds long positions in AI-infrastructure securities, including Nvidia, Micron, Marvell, AMD, Credo and Bloom Energy. A resting order to buy Alphabet at $319.78, entered August 18, filled on October 26, and the position was sold before publication, confirmed against brokerage records on the publication date. The author holds no position in Alphabet, Microsoft, Amazon or Meta. None of the securities discussed is rated in this note.

Not investment advice. Brandon Leon is a publisher, not a registered investment adviser, and nothing here is a recommendation to buy or sell any security or a solicitation to do so. This is independent research prepared for general readership without regard to any individual’s circumstances, objectives or risk tolerance. Forecasts are estimates and will be wrong in detail. Past performance of any call described here does not indicate future results. Do your own work or consult a licensed adviser before acting.

Methodology. Guidance is quoted from each company’s release or call and attributed to the document or speaker. Actuals are from the cash flow statements in results releases. Each company’s own definition of capital spending is kept; the sum of the four is presented as scale only. Share moves are close to close. The direction call and the three proof tests were set and logged before the reports, with their priors; the text says which were met.