Nvidia Earnings: Wall Street's $700bn AI Capex Reckoning

On Wednesday evening, one earnings call will do more to set the direction of global equity markets than any central bank meeting scheduled this month. After the closing bell on 26 August, Nvidia reports its results for the second quarter of fiscal 2027, and it does so into a shareholder base that has spent the past week selling. The stock has fallen for seven consecutive sessions, its longest losing streak since September 2022, closing at $208.48 on Monday (StartupHub). What is being judged is no longer a single semiconductor company. It is the credibility of the largest corporate spending programme in history.

What is actually being reported#

Nvidia has guided to revenue of roughly $91bn for the quarter, plus or minus two per cent, while analysts sit slightly higher at a consensus of about $92.07bn and adjusted earnings of $2.09 a share, up from $1.05 a year earlier (Yahoo Finance). If the company simply meets its own guidance, revenue will have grown by around 95 per cent year on year. In any ordinary week, a near-doubling of sales at a firm this size would be the headline. This is not an ordinary week.

The market has stopped treating Nvidia's own numbers as the point. Investors already expect the beat. The tension sits one layer down, in the guidance and in chief executive Jensen Huang's commentary on the spending plans of Nvidia's largest customers. As Bloomberg framed it on Monday, the report will "test a stock market that's in mid-rotation" (Bloomberg). The concern is not demand today. It is who pays for the demand tomorrow.

The circular economy behind the boom#

To see why, it helps to understand where Nvidia's revenue comes from. The overwhelming majority is its data-centre segment, which sells the graphics processing units, or GPUs, that train and run artificial-intelligence models. In the previous quarter, reported on 20 May, data-centre revenue reached $75.2bn, up 92 per cent year on year and the single largest quarterly result in semiconductor history (Nvidia Q1 FY2027 earnings call; CNBC). Almost all of that comes from a handful of buyers.

Those buyers are the hyperscalers: Microsoft, Alphabet, Amazon and Meta, joined increasingly by Oracle. Their combined capital expenditure for calendar 2026 is now tracking somewhere between $700bn and $725bn, with Amazon near $200bn, Alphabet guiding to about $195bn to $205bn, Microsoft around $175bn and Meta between $130bn and $145bn (StartupHub). A large slice of that flows to Nvidia. Nvidia's growth is, in effect, the hyperscalers' spending, and the hyperscalers' spending is a bet that AI services will eventually generate matching revenue.

Estimates compiled from company guidance as reported by StartupHub; figures are approximate and subject to revision.

Why the market is nervous#

The worry crystallising in August is a simple accounting one. For most of the past decade the hyperscalers funded their investment from their own operating cash flow. That is no longer true. Aggregate cash capital expenditure across the group is on track to overtake operating cash flow around the third quarter of 2026, according to research from Epoch AI, a shift FactSet has documented as a move from self-funding to raising external capital at scale. Incremental annual debt rose from about 9 per cent of capex in fiscal 2024 to 32 per cent on a trailing basis by mid-2026, and equity has re-entered the mix: Alphabet priced an $84.75bn equity raise in June 2026 (Global Data Center Hub).

Oracle illustrates the strain in isolation. Its fiscal 2026 free cash flow was negative $23.7bn against $32bn of operating cash flow, the gap driven by capital spending that surged 162 per cent, while depreciation doubled to $7.1bn (Global Data Center Hub). Depreciation matters more than it sounds. GPUs are refreshed every 18 to 24 months, yet the bonds financing them can run for ten or thirty years. Firms are borrowing on long maturities to buy assets that lose their edge in under two (Footnote Brief). That mismatch is the quiet fault line beneath the boom.

The return side of the equation remains unproven. A widely cited study from MIT's Project NANDA, The GenAI Divide, found that 95 per cent of enterprise generative-AI pilots delivered no measurable profit-and-loss impact despite $30bn to $40bn of spending, with value concentrated in a narrow 5 per cent of deployments (Fortune). If corporate customers cannot show a return on the software, the case for hundreds of billions in hardware becomes harder to sustain.

Market implications across asset classes#

The reach of a single results call now extends well beyond technology equities. The "Magnificent Seven" reached roughly 35 per cent of S&P 500 market capitalisation at their June 2026 peak. That is the highest concentration in the index's history, against a long-run average nearer 17 per cent and a previous record of about 26 per cent in March 2000 (The Motley Fool; Forbes). Index investors are therefore more exposed to AI sentiment than most realise. A passive S&P 500 tracker is, today, a concentrated AI position wearing the costume of diversification.

The knock-on effects are already visible in less obvious corners. When AI shares wobbled in late June and early July, systematic and quantitative hedge funds suffered their worst run in a year; Goldman Sachs estimated that trend-following managers gave back roughly a quarter of their year-to-date gains as volatility jumped and leverage came down (Interactive Brokers / Traders' Insight). In fixed income, the surge in AI-related bond issuance is testing appetite for long-dated corporate credit. And because the hyperscalers are among the largest buyers of their own suppliers' equity and debt, a repricing here ripples into investment-grade spreads, options volatility and, ultimately, the cost of capital for the whole complex.

A technical read: from cash flow to valuation#

For readers less steeped in the mechanics, the valuation debate reduces to one comparison: free cash flow versus capital expenditure. Free cash flow is the money a business generates after paying to maintain and expand its asset base. When capex exceeds operating cash flow, the shortfall must be funded externally, through debt or new shares. That is sustainable if the assets earn a return above their cost of capital. It becomes fragile when the assets depreciate faster than the debt matures, because the borrower must keep refinancing into an uncertain revenue stream.

Analysts value Nvidia and its customers on discounted future cash flows, so the market is really pricing a probability: that AI-generated revenue will eventually catch up with AI capital spending. A line from several strategists in July, that "a great technology and a great investment are not the same thing", captures the distinction (StartupHub). Nvidia can keep posting record numbers even as the aggregate investment case weakens, because its revenue is its customers' spending, not their profit.

Strengths, risks and the counter-argument#

The bullish case is not flimsy. Demand for Nvidia's Blackwell systems is real, backlogs are long, and inference workloads (running models in production rather than training them) are a genuine, growing revenue stream that could underwrite spending for years. Cloud providers argue, reasonably, that under-building capacity would be the costlier mistake.

The risks are equally concrete. Concentration means a disappointment at one company transmits directly into pension and retirement portfolios worldwide. The financing structure embeds a duration mismatch that history treats unkindly. And the demand signal is partly circular: vendor financing, equity stakes and supply agreements between chipmakers, cloud providers and model developers can inflate apparent demand beyond end-customer reality. The honest position is that no one yet knows which way the return-on-investment question resolves; Wednesday's guidance is a data point, not an answer.

How this compares with past cycles#

Regulators are treating the episode as structural rather than a passing rally. The Bank of England's July 2026 Financial Stability Report warned that rapid advances in frontier AI have raised risks to cyber and operational resilience, and flagged stretched AI-related equity valuations alongside AI firms' accelerating reliance on debt financing in the first half of 2026 (Bank of England; Central Banking). The Bank for International Settlements, in its June 2026 Annual Economic Report, placed the sustainability of the AI boom among the principal threats to global stability, alongside a new "sovereign-financial nexus" of record public debt (CNBC).

The obvious comparison is March 2000, when index concentration last approached these levels. But the differences matter. The dot-com leaders often had negligible earnings; today's hyperscalers are among the most profitable enterprises ever built, and Nvidia sells a product with immediate cash margins. This looks less like a mania with no underlying business and more like a genuine platform shift whose financing has run ahead of its proven returns. The closer parallel is the late-1990s telecoms build-out, where the technology transformed the world but many of the companies that laid the fibre did not survive the bill.

Key takeaways#

1. Nvidia's fiscal Q2 2027 results on 26 August are being read as a verdict on roughly $700bn of 2026 hyperscaler AI capital spending, not merely as a company update.
2. The pivotal variable is guidance and customer-capex commentary, not the headline revenue, which the market already expects to beat.
3. Hyperscaler capital spending is on track to exceed operating cash flow around Q3 2026, pushing the group toward debt and equity financing for the first time in a decade.
4. Index concentration near record highs means AI sentiment now drives whole-market and cross-asset outcomes, from quant funds to investment-grade credit.
5. The central bank and BIS warnings, plus weak enterprise AI returns, distinguish verified strain from speculation, but the return-on-investment question remains genuinely open.

Frequently asked questions#

When exactly does Nvidia report? After the US market close on Wednesday 26 August 2026, with the earnings call at 5:00pm Eastern Time (RexShares).

What are analysts expecting? Around $92bn in revenue and adjusted earnings near $2.09 a share, against company guidance of roughly $91bn (Yahoo Finance).

Why has the stock been falling if results are strong? The recent seven-session slide reflects anxiety about whether customers can keep funding capital spending, not weakness in Nvidia's own demand (StartupHub).

What is "AI capex" and why does it matter? It is the capital hyperscalers spend on data centres and chips. In 2026 it is on course to outrun their cash flow, forcing external financing (FactSet).

Is this an "AI bubble"? Regulators including the Bank of England have flagged stretched valuations and rising leverage, but today's leaders are highly profitable, unlike many dot-com names (Bank of England).

How does it affect ordinary index investors? With the Magnificent Seven near 35 per cent of the S&P 500 at their June peak, a passive tracker carries substantial concentrated AI exposure (The Motley Fool).

What should I watch on the call? Guidance for the next quarter, gross margins, and any comment on customer spending durability and inference demand.

Glossary#

Capital expenditure (capex): Money a company spends to acquire or upgrade long-lived physical assets, such as data centres and chips.

Free cash flow: Operating cash flow minus capital expenditure; the cash left over to repay debt, pay dividends or build reserves.

Hyperscaler: A very large cloud-computing provider, chiefly Microsoft, Alphabet, Amazon and Meta, operating data centres at massive scale.

GPU (graphics processing unit): A specialised chip well suited to the parallel calculations behind AI training and inference; Nvidia's core product.

Inference: Running a trained AI model to produce outputs, as opposed to training, which builds the model. Inference is the recurring, revenue-bearing workload.

Market concentration: The degree to which a small number of companies account for an index's total value; higher concentration raises single-name risk.

Duration mismatch: Financing long-dated liabilities (bonds of 10 to 30 years) against assets that depreciate quickly (GPUs refreshed every 18 to 24 months).

Discounted cash flow: A valuation method that estimates worth as the present value of expected future cash flows.

References#

• Bank of England, Financial Stability Report, July 2026: bankofengland.co.uk
• Central Banking, "BoE warns of AI bubble in equity markets": centralbanking.com
• Bank for International Settlements coverage, Annual Economic Report 2026: CNBC
• Nvidia, Q1 FY2027 Earnings Call transcript, 20 May 2026: q4cdn.com
• CNBC, "Nvidia earnings takeaways, Q1 2027": cnbc.com
• Yahoo Finance, "Nvidia's Q2 Earnings: AI Data Center Chip Demand": finance.yahoo.com
• StartupHub, "AI Stocks Daily, 24 August 2026": startuphub.ai
• Bloomberg, "Nvidia and Warsh Will Test a Stock Market That's in Mid-Rotation": bloomberg.com
• Epoch AI, "Hyperscaler Capex vs Cash Flow": epoch.ai
• FactSet, "Hyperscalers Tap External Financing as AI Capex Outruns Cash Flow": insight.factset.com
• Global Data Center Hub, "Oracle Q4 FY2026: Hyperscaler Capex": globaldatacenterhub.com
• Footnote Brief, "Hyperscaler Depreciation and AI Capex Circularity": footnotebrief.com
• MIT Project NANDA, The GenAI Divide, via Fortune
• The Motley Fool, "The Magnificent Seven's Market Cap vs the S&P 500": fool.com
• Forbes, "S&P 500 Weight in Mag 7 Stocks Passes 30%": forbes.com
• Interactive Brokers / Traders' Insight, "AI Trade Unwind Hands Quant Funds Their Worst Run": interactivebrokers.com
• The Motley Fool, "Will Nvidia Stock Soar After Aug. 26?": fool.com

This article is for information only. It is not investment advice, a recommendation, or a forecast. Figures described as estimates, guidance or projections are not certainties. Verified facts, market interpretation and forward-looking speculation are distinguished in the text.