Update: Big-5 AI Capex Guidance Hits $803B — Up $43B From the July Vintage

What changed since the July spend map

In late July we published a scope map of AI infrastructure dollar totals: Big-5 guidance midpoints near $760B for 2026, CreditSights around $750B, Goldman Sachs Global Institute all-in AI infra at $765B, and Investment Research’s 2027 hyperscaler base/bull pair at $1.14T / $1.4T. That post answered how large is the stack, and which stack? This update answers the capital-markets question that moves after every earnings week: what moved in the newest official vintage, and how large is the YoY / vintage delta in dollars?

Three things force a refresh. Amazon, Alphabet, Meta, and Microsoft all reset calendar-2026 capex language in late July. Oracle’s cash outlay guidance is now commonly cited near ~$70B net of customer prepayments (gross still ~$90–95B). And the Q2’26 cash print — Big-4 capex absorbing ~99% of operating cash flow — turns the annual guidance story into a financing story. The dashboard above is built as a vintage delta: company raises and the one headline “cut,” dual 2026 bars on the stacked path, research-house fans, and a Q2 absorption scatter.

The headline table: prior print vs Aug 2026 update

That +$43B / +5.6% revision is the clean vintage delta versus our July research print. Strip Microsoft’s accounting move and the economic stack is closer to ~$818B — a +$58B raise against the same prior baseline. Either framing beats the Street’s pre-Q2 neighborhood (~$725–750B) and keeps 2026 hyperscaler gross spend in the high-seven-hundreds to low-eight-hundreds billion band.

Apply the CreditSights-style ~75% AI-attributed haircut and the AI-specific slice of the Big-5 stack moves from roughly $570B → $602B. That is still a convention, not a 10-K line item — toggle it in the dashboard so you can see both perimeters without mixing them in a single headline.

Amazon and Alphabet did the heavy lifting

Amazon’s jump from ~$200B to ~$220B is the largest absolute company raise in the vintage. It also keeps Amazon the single largest program in the stack — larger, on a standalone basis, than many entire industry capex cohorts. The Q2 cash print reinforces the annual guide: Amazon alone spent ~$53B in the quarter against ~$45B of operating cash flow (absorption >100%).

Alphabet’s second raise — midpoint from $185B to $200B — is the cleanest “guidance only goes one direction” story in the set. The company has now stepped the calendar-year bar twice since February. Combined with Amazon, those two names contribute +$35B of the +$43B Big-5 delta even before Oracle’s net-cash step-up.

Meta’s move looks small in dollars (+$2.5B at the midpoint) but matters in structure: the floor rose from $125B to $130B while the ceiling held at $145B, and the disclosed total continues to fold in finance-lease principal. For desks that compare Meta to cash-only peers, that definitional gap still bites — another reason the dashboard keeps company toggles and source notes adjacent to every panel.

Microsoft’s −$15B is not a build cut

Coverage that treated Microsoft’s calendar-2026 print as a spending slowdown got the economics backwards. The Jul 29 language put CY2026 capex near ~$175B, down from the ~$190B figure markets had been carrying after Q1 — but the company has been shifting more data-center capacity toward operating rather than finance leases. That moves dollars off the capex line without canceling servers, power, or campus construction.

For vintage analysis that means two parallel numbers:

1. Accounting vintage: Big-5 midpoints $760B → $803B (+$43B).
2. Economic vintage: hold Microsoft near the prior ~$190B run-rate → stack nearer $818B (+$58B).

Credit and equity desks that underwrite physical build should prefer the economic frame. Desks that underwrite reported free cash flow and PP&E growth should keep the accounting frame — and then add lease commitments back in the footnotes. Our companion capex intensity update shows the same firm’s FY26 intensity still climbing even as the calendar headline softens: the build is not pausing.

Oracle’s net vs gross split still confuses totals

Oracle’s ~$70B figure circulating after the summer print is net of customer prepayments. Gross guidance in the $90–95B range remains the better comparator to peers who do not net the same way. We carry $70B in the Big-5 midpoint sum for cash-comparability with the Axis / Motley Fool synthesis many desks now quote — and we flag the gross stack in the table and caveats so readers do not undercount OCI / Stargate-linked build by $20B+.

That definitional gap is exactly why our July research post refused a single “true” AI spend number. The update does not invent one. It shows which perimeter moved and by how much.

Research houses: Street climbs; GS GI holds the layer path

Company guidance is the observed near-term object. Research scenarios remain the forward object — and they do not all re-cut on the same week.

• Street consensus for 2026 hyperscaler gross has climbed from the mid-$720Bs toward roughly ~$790B, tracking the Q2 raises (Reuters-compiled paths and sell-side blends sit in the same neighborhood).
• CreditSights-style aggregates that sat near $750B after Q1 now print closer to ~$800B.
• Goldman Sachs Global Institute’s Tracking Trillions all-in AI infra baseline ($765B in 2026, $1.01T in 2027, compounding toward ~$7.6T cumulative 2026–2031) is a layer framework, not a company-guidance rollupit did not need to jump just because Amazon raised $20B. The dashboard’s GI panel keeps that path visible so readers do not force a fake contradiction.
• Goldman Sachs Investment Research’s hyperscaler 2027 base (~$1.14T) and bull (~$1.4T) remain the trillion-dollar headline pair. Public secondary reporting has edged the 2026 IR path up with guidance; the 2027 fan is still the scenario that matters for “does consensus stay too low?” debates.

Toggle the scenario year control between 2026 and 2027. On 2026 you see prior-vs-new bars. On 2027 you see the longer fan where IR’s hyperscaler base can still sit above GI’s all-in total because the perimeters differ — the same paradox we documented in the July spend map and the chips-and-data-centers breakdown.

Q2 cash absorption turns guidance into a financing story

Annual guidance answers “how many dollars will be authorized.” The Q2 print answers “who funds them this quarter.” Across Alphabet, Amazon, Meta, and Microsoft, combined capex of about $170B ran against about $172B of operating cash flow — a cohort absorption ratio near 99%. Alphabet and Amazon both printed >100% absorption in the quarter; Meta was close to flat; Microsoft retained the widest OCF cushion.

That is the hinge for capital markets. When absorption hugs 100%, incremental servers compete with buybacks, dividends, and balance-sheet capacity. The same week’s financing lines — equity raises, senior notes, paused buybacks — are not side stories; they are the residual of the spend map. Pair this update with the AI financing research for the credit-channel view, and with the intensity update for the revenue-share view. Dollars, intensity, and funding are three faces of one cycle.

YoY context still dwarfs the vintage delta. Against our research post’s 2025 Big-5 stack (~$344B), the new ~$803B 2026 guide is roughly a +133% step-up. The +$43B August revision is large in absolute dollars and small next to that year-over-year cliff — which is why markets can treat a mid-single-digit vintage raise as “more of the same” even while credit spreads and FCF prints scream that the same is no longer self-funding.

Caveats and methodology

1. Guidance ≠ audited spend. Midpoints and call language can miss by tens of billions by year-end; Meta and Alphabet still publish ranges.
2. Definitions differ. Meta and Microsoft fold finance leases differently than Amazon/Alphabet cash PP&E. Oracle’s net-of-prepay figure is not peer-comparable to gross guides without adjustment.
3. Microsoft’s −$15B is treated as an accounting vintage move. Economic build is closer to unchanged-to-higher; do not narrate it as a demand rollover without lease footnotes.
4. AI-attributed (~75%) is a research convention applied uniformly for interactionactual AI shares differ by company and year.
5. GS Global Institute figures are a sensitivity framework, not Goldman Sachs Investment Research point forecasts. GS IR 2027 figures come from public secondary reporting of research notes.
6. Q2 absorption uses each company’s own cash-flow definitions (Axis cohort method); it is a quarterly snapshot, not a full-year FCF forecast.
7. Totals may not sum across houses because perimeters differ (leases, power, non-hyperscaler buyers, geography).
8. This post is explanatory data journalism, not investment advice.

Primary synthesis: company Q2 2026 earnings releases and calls (late Jul 2026); Axis Intelligence Research AI Capex Tracker (Aug 2026 synthesis of primary filings); prior theme baseline in AI capex spend research; Goldman Sachs Global Institute Tracking Trillions; Street / CreditSights aggregate commentary via public reporting.

What to watch into year-end 2026

Three coincident signals will tell you whether the vintage delta stabilizes or accelerates again: (1) full-year prints vs these midpoints — especially whether Amazon’s ~$220B and Alphabet’s $195–205B hold after GPU delivery timing; (2) Microsoft lease mix — if operating-lease migration continues, reported CY capex can keep undershooting economic build; (3) absorption and issuance — if the Big-4 cohort stays near 100% of OCF while absolute guides rise, the cycle’s marginal dollar is a credit-market dollar. Until the next official vintage, the live Big-5 midpoint sum is not July’s $760B. It is ~$803B — and on an economic lease-adjusted read, closer to ~$818B.