Big-five tech capex exceeds $1 trillion in a year

Will the combined capital expenditure of Alphabet, Amazon, Meta, Microsoft, and Oracle exceed $1 trillion for a calendar year?

Chance by Q1 2028

76%

Cumulative probability

0%25%50%75%100%50% · Q1 2028
  • Latest forecast
Chance by the end of each quarter
YearQ1Q2Q3Q4
2026<1%
2027<1%<1%<1%3%
202876%76%76%77%
202987%87%87%87%
203090%90%90%91%
203192%92%92%93%

Forecast history

Forecast made on 16 Oct 2026

Latest of 5 forecasts · Methodology v2.1

Model reasoning

The five companies will spend roughly $790 billion in 2026 as the question counts it, so 2026 cannot qualify. Analysts expect about $1.04–1.07 trillion in 2027, and spending only needs to edge up slightly from late-2026 levels to pass $1 trillion. That result would be published in early 2028. The development that would most change this forecast is a broad pullback in spending, driven by rising interest rates, funding strain or an OpenAI shock.

Where things stand (as of 16 Oct 2026)

I pulled the qualifying cash-flow lines from SEC EDGAR's XBRL API (data.sec.gov/api/xbrl/companyconcept/...). Alphabet, Meta, Microsoft and Oracle use the tag PaymentsToAcquirePropertyPlantAndEquipment. Amazon's gross "Purchases of property and equipment" uses PaymentsToAcquireProductiveAssets. I took separate quarters by subtracting year-to-date figures:

  • First half of 2026: Alphabet $80.6B, Amazon $98.4B, Meta $49.1B, Microsoft (Jan–Jun 2026) $66.7B. Together that is $294.8B.
  • Oracle: Mar–May 2026 was $16.5B and Jun–Aug 2026 was $28.5B.
  • Full-year guidance: Alphabet $195–205B, Amazon $220B, Meta $130–145B, Microsoft about $175B for calendar 2026 including finance leases (roughly $158–163B in cash), Oracle $90–95B for its fiscal 2027.

That puts calendar 2026 at about $0.79T under the question's rules. Oracle counts only for March–November. So 2026 cannot qualify, and nothing can resolve before the 2027 numbers are complete. Those land between Oracle's December 2027 10-Q and the four companies' fourth-quarter results in early February 2028, which falls in the quarter ending 31 March 2028.

Consensus for 2027

  • FactSet consensus as of 31 Aug 2026 (Morgan Stanley Counterpoint, "To Free or Not to Free", 17 Sep 2026), calendarized cash capex for 2027: Amazon $272.9B, Alphabet $304.0B, Microsoft $208.5B, Meta $188.7B, Oracle $92.1B. Total $1,066B. The same source has 2028 at $1,159B and 2030 at $1,305B.
  • Bloomberg consensus as of 30 Sep 2026 (J.P. Morgan Guide to the Markets, 4Q26): 2025 $416B, which matches my cash-basis total, so this series is cash-basis; 2026 $799B; 2027 $1,093B (+37%); 2028 about +10%.
  • Goldman Sachs (25 Sep): consensus $1.1T; its own forecast $1.2T for 2027 and $1.4T for 2028.

Under the question's rules I drop about 25% of Oracle (December–February cannot be isolated) and remove Meta's finance-lease principal. That leaves consensus for 2027 at roughly $1.04–1.07T, about 4–7% above the threshold.

Run-rate check

At guidance midpoints, the four companies' fourth-quarter 2026 spending is about $226B, or roughly $0.90T a year. Adding Oracle's March–November share (about $66–69B), 2027 clears $1T if quarterly spending averages only about 3% above the fourth-quarter 2026 level. Consensus implies about 11%.

How often has the next year averaged at least 1.03 times the prior fourth quarter? In 2024–2026 it was 1.3–1.5 times. In 2022 it was about 1.03 times, borderline. In 2023, when rates rose and Amazon and Meta cut spending, it was about 0.92 times.

Signals for 2027 point upward:

  • Nvidia guided 70% revenue growth for fiscal 2028.
  • Amazon ordered 2 million GPUs for 2027–28.
  • Alphabet expects capex to "increase significantly" in 2027.
  • Amazon says it will lack capacity through 2027.
  • Meta is "maximizing 2026 and 2027 capacity."
  • Microsoft expects further growth in fiscal 2027.
  • Backlogs total about $2.35T.
  • Consensus estimates 15 months out were too low by 25–50% in each of 2024–2026.

Main risks to YES (strongest case for NO)

  • The Fed resumed hiking in September 2026, and the 10-year yield is about 5.3%.
  • The five companies' combined free cash flow is forecast to bottom near –$265B in 2027.
  • Oracle was cut to BBB-, and there are doubts about OpenAI's revenue and funding.
  • Microsoft's stock was rewarded for holding spending flat, which encourages restraint elsewhere.
  • Spending is shifting to leases, joint ventures and special-purpose vehicles.
  • A 2023-style retrenchment would mean a miss.

Forecast

  • Calendar 2027 above $1T: about 0.76, from a lognormal around consensus with a small upward tilt and a modelled sensitivity range of about 0.6 to 0.9.
  • By end-2027 via a partial-year floor: about 0.03. This would need first-to-third-quarter 2027 spending about 24% above consensus.
  • If 2027 falls short: chance 2028 clears $1T about 0.45; then 0.25 for 2029 and 0.2 for 2030 if earlier years miss.

Cumulating these year-by-year chances gives the series below, about 0.93 by the end of 2031. Higher capex outlooks for 2027 at the late-October results or January guidance would raise these numbers. A broad retrenchment or an OpenAI funding shock would lower them.

Resolution criteria

Question. By the last day of each quarter, has the combined capital expenditure of Alphabet, Amazon, Meta, Microsoft, and Oracle exceeded $1 trillion for a calendar year?

Definitions.

  • Companies: Alphabet Inc., Amazon.com, Inc., Meta Platforms, Inc., Microsoft Corporation, and Oracle Corporation, on a consolidated worldwide basis. Include all business segments, not just AI or cloud operations. Renamed legal successors count; do not substitute other companies. Follow the consolidation perimeter in the financial statements and count any overlapping consolidated expenditure only once.
  • Capital expenditure: gross cash payments for property, plant, and equipment reported in investing cash flows: Alphabet, Amazon, and Meta's “Purchases of property and equipment”; Microsoft's “Additions to property and equipment”; and Oracle's “Capital expenditures,” including equivalent successor line items. Treat cash outflows as positive amounts. Do not subtract separately reported proceeds from asset sales or incentives. Exclude noncash asset additions, assets obtained through leases, lease principal and interest payments, operating expenses, acquisitions of businesses, purchases of securities, and separately reported purchases of intangible assets. Capitalized costs already included in the specified PP&E cash-flow line count.
  • $1 trillion: $1,000,000,000,000 in nominal US dollars, with no inflation adjustment. Use the highest precision published in the qualifying financial statements, not rounded headline figures. Exactly $1 trillion does not qualify.
  • Calendar year: January 1 through December 31 of one year, beginning with 2026. The Observation period convention applies. Different companies' fiscal years must not be summed as though they covered the same calendar year.
  • Calendar alignment: use calendar-year totals where available, or derive nonoverlapping quarters by subtracting successive fiscal year-to-date cash-flow figures. Microsoft's June fiscal year-end does not prevent reconstruction of calendar quarters. Oracle's May fiscal year-end and February, May, August, and November quarter-ends do not align with calendar quarters: use separately disclosed calendar-year or monthly figures if available; do not prorate fiscal quarters or assign a whole cross-year quarter to the year in which it ends.

Resolves YES if qualifying published actual financial results establish a sum strictly greater than $1 trillion for a completed calendar year. A documented lower bound also suffices: sum only nonoverlapping reporting periods wholly inside that calendar year, assigning zero to unallocated portions. For example, Oracle's March–November expenditure may establish a lower bound when January–February and December cannot be isolated. An estimated allocation of cross-year spending cannot establish YES.

Use the most recent qualifying restatements available at the resolution date. A restatement that first establishes the threshold counts on its publication date; a subsequent downward revision does not reverse an established YES under the Resolution window convention.

Resolves NO at a horizon if no qualifying publication has established the threshold by then. Guidance, budgets, commitments, analyst estimates, annualized quarterly run rates, and trailing-twelve-month periods other than January–December do not count. Neither a lease-inclusive capex headline nor total investing cash outflow substitutes for the defined measure. Incomplete calendar-year coverage counts only when the documented lower bound itself exceeds the threshold.

Resolving source. The five companies' SEC Forms 10-K and 10-Q, including amendments and notes to consolidated cash-flow statements, and company-issued earnings releases or financial supplements containing the corresponding actual cash-flow tables. SEC EDGAR issuer CIKs are Alphabet 1652044, Amazon 1018724, Meta 1326801, Microsoft 789019, and Oracle 1341439. Official investor-relations sites are abc.xyz/investor, ir.aboutamazon.com, investor.atmeta.com, microsoft.com/en-us/investor, and investor.oracle.com. These company-authored financial documents are the narrower resolving sources under Public confirmation (A2); third-party estimates cannot resolve the question. Apply the Resolution window and Resolution date (A3) conventions: resolution occurs when the last publication needed to establish the qualifying combined total becomes public, not on the year's final day.

Download all forecasts (JSON)