Apollo finds no broad AI payoff: NVDA stock forecast before earnings

Apollo’s chief economist argued in an October 24 research note that broad US companies still show little measurable financial payoff from artificial intelligence. That is a warning about the economics of the buyers at the far end of the AI chain, not a new earnings report from Nvidia. For NVDA holders, the immediate question is whether the market will lower the price it pays for Nvidia’s expected earnings before those earnings themselves turn down.

There is already a measurable gap. From September 24 to the October 23 close, TECHi’s NVDA forecast data show the same-period FY2027 EPS estimate edging from $9.3071 to $9.3124, while the share price fell from $224.58 to $210.96. TECHi calculates a 0.06% estimate increase against a 6.06% price decline. The implied price-to-estimated-earnings ratio slipped from about 24.13 to 22.65. That compression, rather than a documented collapse in Nvidia profit expectations, is the starting point for this forecast.

The market has already cut its multiple

Our calculation uses two dated observations of the same fiscal-year EPS estimate and the corresponding daily closes. The September 24 observation shows $224.58 divided by $9.3071, or 24.13 times. The October 23 close of $210.96 divided by $9.3124 yields 22.65 times. The ratio fell roughly 6.12%; EPS changed by less than one tenth of one percent. These are estimates and historical market prices from TECHi’s data providers, not audited FY2027 results. Price adjustments were not confirmed by the quote page, and the estimate date’s close is not necessarily the price at the exact moment the estimate was issued.

The distinction matters because a valuation multiple is the market’s willingness to pay for a dollar of forecast earnings. It can shrink when investors ask whether end customers will ultimately earn enough on AI to sustain infrastructure spending, even while current orders remain strong. Apollo’s note gives that question a new data frame: it compares revenue per worker and profit margins for the seven biggest technology companies with the other 493 S&P 500 constituents and finds improvement concentrated in the tech group. Apollo also points to a productivity series and token-spending data. Its argument is about broad diffusion of AI benefits, not a census of every business using Nvidia hardware.

Apollo’s language is forceful, but the evidence supports a narrower investor conclusion than “AI demand has ended.” Flat aggregate margins outside big tech cannot identify which individual firms used AI, when they adopted it, what they paid, or whether a particular project saved costs. A benefit can also arrive with a lag or be competed away in lower prices. The S&P 493 comparison is a reason to scrutinize the next layer of spending and pricing; it is not proof that Nvidia’s next reported quarter will miss guidance.

Why Nvidia can keep growing while buyers wait for returns

Nvidia’s own August 26 results provide the counterweight. Revenue for the quarter ended July 26 was $96.2 billion, up 106% from a year earlier. Data Center revenue reached $89.0 billion, up 117%. Using the more precise figures in its Form 10-Q, Data Center supplied about 92.5% of total quarterly revenue: $89.023 billion divided by $96.221 billion. This is a company whose current sales are deeply tied to the AI buildout, regardless of what Apollo’s economy-wide charts say about customers’ current returns.

The company guided for fiscal third-quarter revenue of $108.0 billion, plus or minus 2%, and a 74.0% gross margin, plus or minus 0.5 percentage point. The revenue midpoint is about 12.2% above the July quarter’s $96.221 billion. Guidance is a management forecast, not a result. Nvidia specifically excluded Data Center compute revenue from China in that outlook. A quarter near the midpoint would show that equipment demand remained robust through the reporting period, though it would not establish that enterprises buying AI services are themselves earning attractive returns.

The route from enterprise budgets to Nvidia revenue is indirect. Nvidia sells to manufacturers, integrators, cloud providers, AI labs and others; a business can rent computing from a cloud provider rather than purchase Nvidia systems itself. The 10-Q says one direct customer accounted for 16% of second-quarter revenue. Nvidia also estimates that one AI research and deployment company contributed a meaningful amount of revenue by purchasing cloud services from its customers. It does not name that indirect customer in this disclosure. Investor exposure therefore depends on a chain of customers’ funding, utilization and pricing decisions, not a single chart of non-tech corporate margins.

There is a plausible delay between Apollo’s concern and a chipmaker’s results. Hyperscalers may continue building capacity already ordered or contracted even if some enterprise customers are still testing applications. AI labs may use more compute to train and serve models before the broader economy shows a productivity gain. Conversely, a long spell of poor downstream returns can eventually limit customers’ willingness to pay, weaken cloud pricing and slow the next round of equipment orders. The investment question is the timing and strength of that feedback loop.

A conditional NVDA forecast through the next report

Friday’s $210.96 regular-session close on October 23 is the price baseline; US shares do not trade in their regular session on Saturday. TECHi’s earnings calendar lists November 17, 2026 after market close as the next confirmed Nvidia report. The following are TECHi valuation sensitivities around that event, not analyst targets, probabilities, or a claim that these prices will trade on November 18. Each calculation starts from the October 24 FY2027 EPS estimate of $9.3124 shown on TECHi’s forecast page and changes only the multiple. Results are rounded to cents after multiplication.

• Bear case, 18 times: $9.3124 × 18 = $167.62, about 20.5% below the $210.96 Friday close. This scenario represents a materially lower market appetite for AI infrastructure earnings, perhaps because guidance, customer funding or margins weaken. It is a multiple stress test; if the EPS estimate also falls, the simple $167.62 calculation would understate the downside at an unchanged 18-times multiple.
• Base case, 22.65 times: $9.3124 × 22.65 = $210.93, essentially flat against Friday’s close. This represents continued growth near current expectations without a major change in how the market prices those earnings. It is a reference line, not a prediction of a zero return.
• Bull case, 26 times: $9.3124 × 26 = $242.12, about 14.8% above Friday’s close. A stronger multiple would require investors to grow more confident in sustained shipments, gross margin and customers’ ability to monetize AI. It would not follow mechanically from one quarter of high revenue.

The spread between $167.62 and $242.12 is $74.50 per share with the EPS input held fixed. That width shows how sensitive an expensive growth stock can be to the narrative about future demand. It does not measure expected volatility or substitute for an options market estimate. TECHi’s live forecast page displays a mean analyst target of $328.95, but that target has an unspecified horizon and target sample count; it is neither a November target nor the source of our three cases.

A more direct way to read the model is to ask what multiple the market assigns after seeing the next report. If Nvidia reports near its $108 billion revenue midpoint and around the guided 74% gross margin, investors could still refuse to restore September’s 24.13-times multiple if downstream ROI remains uncertain. At the current $9.3124 estimate, 24.13 times would imply roughly $224.72, about 6.5% above Friday’s close. That is the amount of valuation recovery required merely to revisit September’s multiple using today’s slightly higher EPS estimate. The earnings report can change the EPS input as well, so this calculation should not be read as a price target.

The evidence that could overturn either case

Apollo’s research is most useful as a testable warning. The bearish feedback loop would become more persuasive if cloud providers reported weaker AI-service pricing or utilization, if large buyers deferred infrastructure commitments, or if Nvidia’s own guidance and gross margin fell below its August outlook. A broad-market productivity chart alone would not establish any of those company-specific developments. Nvidia’s filings also identify power availability, land, capital and deployment constraints as reasons customers may postpone purchases; even healthy end demand could be delayed by physical infrastructure.

The bullish counterargument is substantial. Nvidia has already shown extraordinary reported revenue growth, a 75% July-quarter gross margin, and a $108 billion next-quarter revenue midpoint. Management expects to sell both Blackwell and Rubin systems and says supply constraints remain. Demand can stay ahead of supply for a time even when average enterprise ROI is unclear. Strong cloud contracts, better utilization and pricing, or profitable deployment by customers would weaken Apollo’s concern and support a higher multiple. None of those outcomes can be inferred from share-price movement alone.

There is also a risk of measuring the wrong end of the chain. Apollo’s S&P 493 series asks whether ordinary listed companies are seeing higher revenue per employee or margins. Nvidia’s near-term customers include cloud and AI firms that may sell capacity to startups and enterprises worldwide. The public S&P 493 is not the full set of Nvidia’s end users. Its companies also differ in accounting, sector mix and spending cycles. Treat Apollo’s aggregate result as a challenge to the long-run AI demand thesis, not a substitute for Nvidia’s reported sales or the next round of customer disclosures.

For a holder deciding what to watch next, the order is practical. The October 26 market open will show how investors first price Apollo’s dated research after the weekend; there is no Saturday regular-session NVDA move to report. Then the November 17 earnings release will test the $108 billion revenue midpoint, the 74% gross-margin guide, and management’s comments on customer mix and capacity. Afterward, updates from the large cloud buyers will help determine whether rented AI compute is turning into durable, profitable demand. The forecast changes if those observations move the EPS estimate or justify a different multiple. Until then, the most defensible reading is that Nvidia’s earnings expectations have held up far better than its valuation since late September.